{ "best_metric": null, "best_model_checkpoint": null, "epoch": 1.49447819420453, "global_step": 77000, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.0, "learning_rate": 4.999935303974277e-05, "loss": 9.6033, "step": 2 }, { "epoch": 0.0, "learning_rate": 4.999870607948554e-05, "loss": 9.5619, "step": 4 }, { "epoch": 0.0, "learning_rate": 4.999805911922831e-05, "loss": 9.4581, "step": 6 }, { "epoch": 0.0, "learning_rate": 4.999741215897107e-05, "loss": 9.0576, "step": 8 }, { "epoch": 0.0, "learning_rate": 4.999676519871385e-05, "loss": 9.1234, "step": 10 }, { "epoch": 0.0, "learning_rate": 4.999611823845661e-05, "loss": 8.9434, "step": 12 }, { "epoch": 0.0, "learning_rate": 4.9995471278199386e-05, "loss": 8.8236, "step": 14 }, { "epoch": 0.0, "learning_rate": 4.999482431794215e-05, "loss": 8.6414, "step": 16 }, { "epoch": 0.0, "learning_rate": 4.999417735768492e-05, "loss": 8.5509, "step": 18 }, { "epoch": 0.0, "learning_rate": 4.999353039742769e-05, "loss": 8.592, "step": 20 }, { "epoch": 0.0, "learning_rate": 4.9992883437170456e-05, "loss": 8.6178, "step": 22 }, { "epoch": 0.0, "learning_rate": 4.9992236476913225e-05, "loss": 8.2506, "step": 24 }, { "epoch": 0.0, "learning_rate": 4.9991589516655994e-05, "loss": 8.2152, "step": 26 }, { "epoch": 0.0, "learning_rate": 4.999094255639876e-05, "loss": 8.1218, "step": 28 }, { "epoch": 0.0, "learning_rate": 4.999029559614153e-05, "loss": 7.9923, "step": 30 }, { "epoch": 0.0, "learning_rate": 4.99896486358843e-05, "loss": 7.8494, "step": 32 }, { "epoch": 0.0, "learning_rate": 4.998900167562707e-05, "loss": 7.5102, "step": 34 }, { "epoch": 0.0, "learning_rate": 4.998835471536984e-05, "loss": 7.7483, "step": 36 }, { "epoch": 0.0, "learning_rate": 4.998770775511261e-05, "loss": 7.531, "step": 38 }, { "epoch": 0.0, "learning_rate": 4.998706079485537e-05, "loss": 7.4975, "step": 40 }, { "epoch": 0.0, "learning_rate": 4.9986413834598146e-05, "loss": 7.4737, "step": 42 }, { "epoch": 0.0, "learning_rate": 4.998576687434091e-05, "loss": 7.5211, "step": 44 }, { "epoch": 0.0, "learning_rate": 4.998511991408368e-05, "loss": 7.188, "step": 46 }, { "epoch": 0.0, "learning_rate": 4.998447295382645e-05, "loss": 7.1897, "step": 48 }, { "epoch": 0.0, "learning_rate": 4.9983825993569216e-05, "loss": 6.66, "step": 50 }, { "epoch": 0.0, "learning_rate": 4.998317903331199e-05, "loss": 6.9516, "step": 52 }, { "epoch": 0.0, "learning_rate": 4.9982532073054754e-05, "loss": 7.0331, "step": 54 }, { "epoch": 0.0, "learning_rate": 4.998188511279752e-05, "loss": 6.7742, "step": 56 }, { "epoch": 0.0, "learning_rate": 4.998123815254029e-05, "loss": 7.029, "step": 58 }, { "epoch": 0.0, "learning_rate": 4.998059119228306e-05, "loss": 6.4881, "step": 60 }, { "epoch": 0.0, "learning_rate": 4.997994423202583e-05, "loss": 6.7053, "step": 62 }, { "epoch": 0.0, "learning_rate": 4.99792972717686e-05, "loss": 6.342, "step": 64 }, { "epoch": 0.0, "learning_rate": 4.997865031151137e-05, "loss": 6.1993, "step": 66 }, { "epoch": 0.0, "learning_rate": 4.997800335125413e-05, "loss": 6.4181, "step": 68 }, { "epoch": 0.0, "learning_rate": 4.997735639099691e-05, "loss": 5.8703, "step": 70 }, { "epoch": 0.0, "learning_rate": 4.997670943073967e-05, "loss": 6.1984, "step": 72 }, { "epoch": 0.0, "learning_rate": 4.9976062470482445e-05, "loss": 5.9197, "step": 74 }, { "epoch": 0.0, "learning_rate": 4.997541551022521e-05, "loss": 5.885, "step": 76 }, { "epoch": 0.0, "learning_rate": 4.9974768549967976e-05, "loss": 5.587, "step": 78 }, { "epoch": 0.0, "learning_rate": 4.9974121589710745e-05, "loss": 5.7716, "step": 80 }, { "epoch": 0.0, "learning_rate": 4.9973474629453515e-05, "loss": 5.8239, "step": 82 }, { "epoch": 0.0, "learning_rate": 4.9972827669196284e-05, "loss": 5.9782, "step": 84 }, { "epoch": 0.0, "learning_rate": 4.997218070893905e-05, "loss": 5.8922, "step": 86 }, { "epoch": 0.0, "learning_rate": 4.997153374868182e-05, "loss": 5.5118, "step": 88 }, { "epoch": 0.0, "learning_rate": 4.9970886788424584e-05, "loss": 5.3661, "step": 90 }, { "epoch": 0.0, "learning_rate": 4.997023982816736e-05, "loss": 5.472, "step": 92 }, { "epoch": 0.0, "learning_rate": 4.996959286791012e-05, "loss": 5.2819, "step": 94 }, { "epoch": 0.0, "learning_rate": 4.99689459076529e-05, "loss": 5.0728, "step": 96 }, { "epoch": 0.0, "learning_rate": 4.996829894739567e-05, "loss": 5.0135, "step": 98 }, { "epoch": 0.0, "learning_rate": 4.996765198713843e-05, "loss": 5.2769, "step": 100 }, { "epoch": 0.0, "learning_rate": 4.9967005026881205e-05, "loss": 5.043, "step": 102 }, { "epoch": 0.0, "learning_rate": 4.996635806662397e-05, "loss": 5.1687, "step": 104 }, { "epoch": 0.0, "learning_rate": 4.996571110636674e-05, "loss": 4.6953, "step": 106 }, { "epoch": 0.0, "learning_rate": 4.9965064146109506e-05, "loss": 4.9736, "step": 108 }, { "epoch": 0.0, "learning_rate": 4.9964417185852275e-05, "loss": 5.3861, "step": 110 }, { "epoch": 0.0, "learning_rate": 4.9963770225595044e-05, "loss": 5.1309, "step": 112 }, { "epoch": 0.0, "learning_rate": 4.996312326533781e-05, "loss": 4.5333, "step": 114 }, { "epoch": 0.0, "learning_rate": 4.996247630508058e-05, "loss": 4.7239, "step": 116 }, { "epoch": 0.0, "learning_rate": 4.996182934482335e-05, "loss": 4.6453, "step": 118 }, { "epoch": 0.0, "learning_rate": 4.996118238456612e-05, "loss": 4.7872, "step": 120 }, { "epoch": 0.0, "learning_rate": 4.996053542430888e-05, "loss": 4.7244, "step": 122 }, { "epoch": 0.0, "learning_rate": 4.995988846405166e-05, "loss": 4.1793, "step": 124 }, { "epoch": 0.0, "learning_rate": 4.995924150379442e-05, "loss": 4.4499, "step": 126 }, { "epoch": 0.0, "learning_rate": 4.995859454353719e-05, "loss": 4.599, "step": 128 }, { "epoch": 0.0, "learning_rate": 4.9957947583279966e-05, "loss": 4.2811, "step": 130 }, { "epoch": 0.0, "learning_rate": 4.995730062302273e-05, "loss": 4.4827, "step": 132 }, { "epoch": 0.0, "learning_rate": 4.9956653662765504e-05, "loss": 3.9349, "step": 134 }, { "epoch": 0.0, "learning_rate": 4.9956006702508266e-05, "loss": 4.4451, "step": 136 }, { "epoch": 0.0, "learning_rate": 4.9955359742251035e-05, "loss": 4.2531, "step": 138 }, { "epoch": 0.0, "learning_rate": 4.9954712781993804e-05, "loss": 4.3826, "step": 140 }, { "epoch": 0.0, "learning_rate": 4.9954065821736574e-05, "loss": 4.0539, "step": 142 }, { "epoch": 0.0, "learning_rate": 4.995341886147934e-05, "loss": 4.2217, "step": 144 }, { "epoch": 0.0, "learning_rate": 4.995277190122211e-05, "loss": 4.2532, "step": 146 }, { "epoch": 0.0, "learning_rate": 4.995212494096488e-05, "loss": 4.1193, "step": 148 }, { "epoch": 0.0, "learning_rate": 4.995147798070764e-05, "loss": 4.0205, "step": 150 }, { "epoch": 0.0, "learning_rate": 4.995083102045042e-05, "loss": 4.1066, "step": 152 }, { "epoch": 0.0, "learning_rate": 4.995018406019318e-05, "loss": 3.9215, "step": 154 }, { "epoch": 0.0, "learning_rate": 4.994953709993596e-05, "loss": 4.0196, "step": 156 }, { "epoch": 0.0, "learning_rate": 4.994889013967872e-05, "loss": 3.6452, "step": 158 }, { "epoch": 0.0, "learning_rate": 4.994824317942149e-05, "loss": 4.1924, "step": 160 }, { "epoch": 0.0, "learning_rate": 4.9947596219164264e-05, "loss": 3.6099, "step": 162 }, { "epoch": 0.0, "learning_rate": 4.994694925890703e-05, "loss": 3.843, "step": 164 }, { "epoch": 0.0, "learning_rate": 4.9946302298649796e-05, "loss": 3.22, "step": 166 }, { "epoch": 0.0, "learning_rate": 4.9945655338392565e-05, "loss": 3.5336, "step": 168 }, { "epoch": 0.0, "learning_rate": 4.9945008378135334e-05, "loss": 3.5497, "step": 170 }, { "epoch": 0.0, "learning_rate": 4.9944361417878096e-05, "loss": 3.925, "step": 172 }, { "epoch": 0.0, "learning_rate": 4.994371445762087e-05, "loss": 3.5096, "step": 174 }, { "epoch": 0.0, "learning_rate": 4.994306749736364e-05, "loss": 3.7757, "step": 176 }, { "epoch": 0.0, "learning_rate": 4.994242053710641e-05, "loss": 3.6245, "step": 178 }, { "epoch": 0.0, "learning_rate": 4.994177357684918e-05, "loss": 3.816, "step": 180 }, { "epoch": 0.0, "learning_rate": 4.994112661659194e-05, "loss": 3.7088, "step": 182 }, { "epoch": 0.0, "learning_rate": 4.994047965633472e-05, "loss": 3.5397, "step": 184 }, { "epoch": 0.0, "learning_rate": 4.993983269607748e-05, "loss": 3.6075, "step": 186 }, { "epoch": 0.0, "learning_rate": 4.993918573582025e-05, "loss": 3.3753, "step": 188 }, { "epoch": 0.0, "learning_rate": 4.993853877556302e-05, "loss": 3.1994, "step": 190 }, { "epoch": 0.0, "learning_rate": 4.993789181530579e-05, "loss": 3.1617, "step": 192 }, { "epoch": 0.0, "learning_rate": 4.9937244855048556e-05, "loss": 3.4593, "step": 194 }, { "epoch": 0.0, "learning_rate": 4.9936597894791325e-05, "loss": 3.354, "step": 196 }, { "epoch": 0.0, "learning_rate": 4.9935950934534094e-05, "loss": 3.3414, "step": 198 }, { "epoch": 0.0, "learning_rate": 4.9935303974276863e-05, "loss": 3.0447, "step": 200 }, { "epoch": 0.0, "learning_rate": 4.993465701401963e-05, "loss": 3.0906, "step": 202 }, { "epoch": 0.0, "learning_rate": 4.9934010053762395e-05, "loss": 3.297, "step": 204 }, { "epoch": 0.0, "learning_rate": 4.993336309350517e-05, "loss": 2.8983, "step": 206 }, { "epoch": 0.0, "learning_rate": 4.993271613324794e-05, "loss": 3.023, "step": 208 }, { "epoch": 0.0, "learning_rate": 4.99320691729907e-05, "loss": 3.0626, "step": 210 }, { "epoch": 0.0, "learning_rate": 4.993142221273348e-05, "loss": 3.1461, "step": 212 }, { "epoch": 0.0, "learning_rate": 4.993077525247624e-05, "loss": 2.8002, "step": 214 }, { "epoch": 0.0, "learning_rate": 4.9930128292219016e-05, "loss": 3.3363, "step": 216 }, { "epoch": 0.0, "learning_rate": 4.992948133196178e-05, "loss": 3.0975, "step": 218 }, { "epoch": 0.0, "learning_rate": 4.992883437170455e-05, "loss": 3.2126, "step": 220 }, { "epoch": 0.0, "learning_rate": 4.9928187411447317e-05, "loss": 3.1103, "step": 222 }, { "epoch": 0.0, "learning_rate": 4.9927540451190086e-05, "loss": 2.9334, "step": 224 }, { "epoch": 0.0, "learning_rate": 4.9926893490932855e-05, "loss": 2.7802, "step": 226 }, { "epoch": 0.0, "learning_rate": 4.9926246530675624e-05, "loss": 2.894, "step": 228 }, { "epoch": 0.0, "learning_rate": 4.992559957041839e-05, "loss": 3.1093, "step": 230 }, { "epoch": 0.0, "learning_rate": 4.9924952610161155e-05, "loss": 3.1455, "step": 232 }, { "epoch": 0.0, "learning_rate": 4.992430564990393e-05, "loss": 2.7299, "step": 234 }, { "epoch": 0.0, "learning_rate": 4.992365868964669e-05, "loss": 2.6996, "step": 236 }, { "epoch": 0.0, "learning_rate": 4.992301172938947e-05, "loss": 2.7217, "step": 238 }, { "epoch": 0.0, "learning_rate": 4.992236476913224e-05, "loss": 2.762, "step": 240 }, { "epoch": 0.0, "learning_rate": 4.9921717808875e-05, "loss": 3.1362, "step": 242 }, { "epoch": 0.0, "learning_rate": 4.9921070848617776e-05, "loss": 3.1559, "step": 244 }, { "epoch": 0.0, "learning_rate": 4.992042388836054e-05, "loss": 2.94, "step": 246 }, { "epoch": 0.0, "learning_rate": 4.991977692810331e-05, "loss": 2.8363, "step": 248 }, { "epoch": 0.0, "learning_rate": 4.991912996784608e-05, "loss": 2.7469, "step": 250 }, { "epoch": 0.0, "learning_rate": 4.9918483007588846e-05, "loss": 2.8318, "step": 252 }, { "epoch": 0.0, "learning_rate": 4.9917836047331615e-05, "loss": 2.658, "step": 254 }, { "epoch": 0.0, "learning_rate": 4.9917189087074384e-05, "loss": 2.5842, "step": 256 }, { "epoch": 0.01, "learning_rate": 4.991654212681715e-05, "loss": 2.4782, "step": 258 }, { "epoch": 0.01, "learning_rate": 4.991589516655992e-05, "loss": 2.5736, "step": 260 }, { "epoch": 0.01, "learning_rate": 4.991524820630269e-05, "loss": 2.6445, "step": 262 }, { "epoch": 0.01, "learning_rate": 4.9914601246045454e-05, "loss": 2.5047, "step": 264 }, { "epoch": 0.01, "learning_rate": 4.991395428578823e-05, "loss": 2.4318, "step": 266 }, { "epoch": 0.01, "learning_rate": 4.991330732553099e-05, "loss": 2.5641, "step": 268 }, { "epoch": 0.01, "learning_rate": 4.991266036527376e-05, "loss": 2.8337, "step": 270 }, { "epoch": 0.01, "learning_rate": 4.991201340501653e-05, "loss": 2.4834, "step": 272 }, { "epoch": 0.01, "learning_rate": 4.99113664447593e-05, "loss": 2.7468, "step": 274 }, { "epoch": 0.01, "learning_rate": 4.9910719484502075e-05, "loss": 2.8172, "step": 276 }, { "epoch": 0.01, "learning_rate": 4.991007252424484e-05, "loss": 2.2347, "step": 278 }, { "epoch": 0.01, "learning_rate": 4.9909425563987606e-05, "loss": 2.5303, "step": 280 }, { "epoch": 0.01, "learning_rate": 4.9908778603730376e-05, "loss": 2.4038, "step": 282 }, { "epoch": 0.01, "learning_rate": 4.9908131643473145e-05, "loss": 2.5724, "step": 284 }, { "epoch": 0.01, "learning_rate": 4.9907484683215914e-05, "loss": 2.3371, "step": 286 }, { "epoch": 0.01, "learning_rate": 4.990683772295868e-05, "loss": 2.2351, "step": 288 }, { "epoch": 0.01, "learning_rate": 4.990619076270145e-05, "loss": 2.3604, "step": 290 }, { "epoch": 0.01, "learning_rate": 4.9905543802444214e-05, "loss": 2.2751, "step": 292 }, { "epoch": 0.01, "learning_rate": 4.990489684218699e-05, "loss": 2.3773, "step": 294 }, { "epoch": 0.01, "learning_rate": 4.990424988192975e-05, "loss": 2.1117, "step": 296 }, { "epoch": 0.01, "learning_rate": 4.990360292167253e-05, "loss": 2.3509, "step": 298 }, { "epoch": 0.01, "learning_rate": 4.990295596141529e-05, "loss": 2.2596, "step": 300 }, { "epoch": 0.01, "learning_rate": 4.990230900115806e-05, "loss": 2.4357, "step": 302 }, { "epoch": 0.01, "learning_rate": 4.990166204090083e-05, "loss": 2.6087, "step": 304 }, { "epoch": 0.01, "learning_rate": 4.99010150806436e-05, "loss": 2.6681, "step": 306 }, { "epoch": 0.01, "learning_rate": 4.990036812038637e-05, "loss": 2.1479, "step": 308 }, { "epoch": 0.01, "learning_rate": 4.9899721160129136e-05, "loss": 2.3646, "step": 310 }, { "epoch": 0.01, "learning_rate": 4.9899074199871905e-05, "loss": 2.0756, "step": 312 }, { "epoch": 0.01, "learning_rate": 4.989842723961467e-05, "loss": 2.2927, "step": 314 }, { "epoch": 0.01, "learning_rate": 4.989778027935744e-05, "loss": 2.4599, "step": 316 }, { "epoch": 0.01, "learning_rate": 4.9897133319100205e-05, "loss": 2.3704, "step": 318 }, { "epoch": 0.01, "learning_rate": 4.989648635884298e-05, "loss": 2.321, "step": 320 }, { "epoch": 0.01, "learning_rate": 4.989583939858575e-05, "loss": 1.8712, "step": 322 }, { "epoch": 0.01, "learning_rate": 4.989519243832851e-05, "loss": 1.9911, "step": 324 }, { "epoch": 0.01, "learning_rate": 4.989454547807129e-05, "loss": 2.1339, "step": 326 }, { "epoch": 0.01, "learning_rate": 4.989389851781405e-05, "loss": 2.2953, "step": 328 }, { "epoch": 0.01, "learning_rate": 4.989325155755682e-05, "loss": 2.4024, "step": 330 }, { "epoch": 0.01, "learning_rate": 4.989260459729959e-05, "loss": 1.8214, "step": 332 }, { "epoch": 0.01, "learning_rate": 4.989195763704236e-05, "loss": 2.0689, "step": 334 }, { "epoch": 0.01, "learning_rate": 4.989131067678513e-05, "loss": 2.2834, "step": 336 }, { "epoch": 0.01, "learning_rate": 4.9890663716527896e-05, "loss": 2.0693, "step": 338 }, { "epoch": 0.01, "learning_rate": 4.9890016756270665e-05, "loss": 2.1429, "step": 340 }, { "epoch": 0.01, "learning_rate": 4.9889369796013435e-05, "loss": 2.2443, "step": 342 }, { "epoch": 0.01, "learning_rate": 4.9888722835756204e-05, "loss": 1.9779, "step": 344 }, { "epoch": 0.01, "learning_rate": 4.9888075875498966e-05, "loss": 1.8979, "step": 346 }, { "epoch": 0.01, "learning_rate": 4.988742891524174e-05, "loss": 2.3826, "step": 348 }, { "epoch": 0.01, "learning_rate": 4.9886781954984504e-05, "loss": 2.3665, "step": 350 }, { "epoch": 0.01, "learning_rate": 4.988613499472727e-05, "loss": 2.1294, "step": 352 }, { "epoch": 0.01, "learning_rate": 4.988548803447005e-05, "loss": 2.3596, "step": 354 }, { "epoch": 0.01, "learning_rate": 4.988484107421281e-05, "loss": 2.3387, "step": 356 }, { "epoch": 0.01, "learning_rate": 4.988419411395559e-05, "loss": 1.9482, "step": 358 }, { "epoch": 0.01, "learning_rate": 4.988354715369835e-05, "loss": 1.993, "step": 360 }, { "epoch": 0.01, "learning_rate": 4.988290019344112e-05, "loss": 2.025, "step": 362 }, { "epoch": 0.01, "learning_rate": 4.988225323318389e-05, "loss": 1.7222, "step": 364 }, { "epoch": 0.01, "learning_rate": 4.988160627292666e-05, "loss": 1.7964, "step": 366 }, { "epoch": 0.01, "learning_rate": 4.9880959312669426e-05, "loss": 1.9092, "step": 368 }, { "epoch": 0.01, "learning_rate": 4.9880312352412195e-05, "loss": 1.9691, "step": 370 }, { "epoch": 0.01, "learning_rate": 4.9879665392154964e-05, "loss": 1.8555, "step": 372 }, { "epoch": 0.01, "learning_rate": 4.9879018431897726e-05, "loss": 1.8891, "step": 374 }, { "epoch": 0.01, "learning_rate": 4.98783714716405e-05, "loss": 2.1726, "step": 376 }, { "epoch": 0.01, "learning_rate": 4.9877724511383264e-05, "loss": 1.7558, "step": 378 }, { "epoch": 0.01, "learning_rate": 4.987707755112604e-05, "loss": 1.8576, "step": 380 }, { "epoch": 0.01, "learning_rate": 4.98764305908688e-05, "loss": 1.6923, "step": 382 }, { "epoch": 0.01, "learning_rate": 4.987578363061157e-05, "loss": 1.6228, "step": 384 }, { "epoch": 0.01, "learning_rate": 4.987513667035435e-05, "loss": 1.7857, "step": 386 }, { "epoch": 0.01, "learning_rate": 4.987448971009711e-05, "loss": 1.6437, "step": 388 }, { "epoch": 0.01, "learning_rate": 4.987384274983988e-05, "loss": 1.8031, "step": 390 }, { "epoch": 0.01, "learning_rate": 4.987319578958265e-05, "loss": 1.8493, "step": 392 }, { "epoch": 0.01, "learning_rate": 4.987254882932542e-05, "loss": 1.9473, "step": 394 }, { "epoch": 0.01, "learning_rate": 4.9871901869068186e-05, "loss": 1.9283, "step": 396 }, { "epoch": 0.01, "learning_rate": 4.9871254908810955e-05, "loss": 1.6596, "step": 398 }, { "epoch": 0.01, "learning_rate": 4.9870607948553724e-05, "loss": 1.6597, "step": 400 }, { "epoch": 0.01, "learning_rate": 4.9869960988296493e-05, "loss": 1.8636, "step": 402 }, { "epoch": 0.01, "learning_rate": 4.986931402803926e-05, "loss": 1.871, "step": 404 }, { "epoch": 0.01, "learning_rate": 4.9868667067782025e-05, "loss": 1.7696, "step": 406 }, { "epoch": 0.01, "learning_rate": 4.98680201075248e-05, "loss": 1.8002, "step": 408 }, { "epoch": 0.01, "learning_rate": 4.986737314726756e-05, "loss": 1.613, "step": 410 }, { "epoch": 0.01, "learning_rate": 4.986672618701033e-05, "loss": 1.9089, "step": 412 }, { "epoch": 0.01, "learning_rate": 4.98660792267531e-05, "loss": 1.7179, "step": 414 }, { "epoch": 0.01, "learning_rate": 4.986543226649587e-05, "loss": 1.7105, "step": 416 }, { "epoch": 0.01, "learning_rate": 4.986478530623864e-05, "loss": 1.7527, "step": 418 }, { "epoch": 0.01, "learning_rate": 4.986413834598141e-05, "loss": 1.7135, "step": 420 }, { "epoch": 0.01, "learning_rate": 4.986349138572418e-05, "loss": 1.7523, "step": 422 }, { "epoch": 0.01, "learning_rate": 4.986284442546695e-05, "loss": 1.7421, "step": 424 }, { "epoch": 0.01, "learning_rate": 4.9862197465209716e-05, "loss": 1.794, "step": 426 }, { "epoch": 0.01, "learning_rate": 4.986155050495248e-05, "loss": 1.6678, "step": 428 }, { "epoch": 0.01, "learning_rate": 4.9860903544695254e-05, "loss": 1.8281, "step": 430 }, { "epoch": 0.01, "learning_rate": 4.986025658443802e-05, "loss": 1.7302, "step": 432 }, { "epoch": 0.01, "learning_rate": 4.9859609624180785e-05, "loss": 2.0101, "step": 434 }, { "epoch": 0.01, "learning_rate": 4.985896266392356e-05, "loss": 1.4022, "step": 436 }, { "epoch": 0.01, "learning_rate": 4.9858315703666323e-05, "loss": 1.6733, "step": 438 }, { "epoch": 0.01, "learning_rate": 4.98576687434091e-05, "loss": 1.3846, "step": 440 }, { "epoch": 0.01, "learning_rate": 4.985702178315186e-05, "loss": 1.7236, "step": 442 }, { "epoch": 0.01, "learning_rate": 4.985637482289463e-05, "loss": 1.4488, "step": 444 }, { "epoch": 0.01, "learning_rate": 4.98557278626374e-05, "loss": 1.61, "step": 446 }, { "epoch": 0.01, "learning_rate": 4.985508090238017e-05, "loss": 1.6083, "step": 448 }, { "epoch": 0.01, "learning_rate": 4.985443394212294e-05, "loss": 1.4279, "step": 450 }, { "epoch": 0.01, "learning_rate": 4.985378698186571e-05, "loss": 1.5, "step": 452 }, { "epoch": 0.01, "learning_rate": 4.9853140021608476e-05, "loss": 1.6171, "step": 454 }, { "epoch": 0.01, "learning_rate": 4.985249306135124e-05, "loss": 1.513, "step": 456 }, { "epoch": 0.01, "learning_rate": 4.9851846101094014e-05, "loss": 1.6798, "step": 458 }, { "epoch": 0.01, "learning_rate": 4.9851199140836777e-05, "loss": 1.3956, "step": 460 }, { "epoch": 0.01, "learning_rate": 4.985055218057955e-05, "loss": 1.7786, "step": 462 }, { "epoch": 0.01, "learning_rate": 4.984990522032232e-05, "loss": 1.4506, "step": 464 }, { "epoch": 0.01, "learning_rate": 4.9849258260065084e-05, "loss": 1.454, "step": 466 }, { "epoch": 0.01, "learning_rate": 4.984861129980786e-05, "loss": 1.5723, "step": 468 }, { "epoch": 0.01, "learning_rate": 4.984796433955062e-05, "loss": 1.5365, "step": 470 }, { "epoch": 0.01, "learning_rate": 4.984731737929339e-05, "loss": 1.4179, "step": 472 }, { "epoch": 0.01, "learning_rate": 4.984667041903616e-05, "loss": 1.39, "step": 474 }, { "epoch": 0.01, "learning_rate": 4.984602345877893e-05, "loss": 1.4688, "step": 476 }, { "epoch": 0.01, "learning_rate": 4.98453764985217e-05, "loss": 1.705, "step": 478 }, { "epoch": 0.01, "learning_rate": 4.984472953826447e-05, "loss": 1.2332, "step": 480 }, { "epoch": 0.01, "learning_rate": 4.9844082578007237e-05, "loss": 1.41, "step": 482 }, { "epoch": 0.01, "learning_rate": 4.9843435617750006e-05, "loss": 1.4656, "step": 484 }, { "epoch": 0.01, "learning_rate": 4.9842788657492775e-05, "loss": 1.3603, "step": 486 }, { "epoch": 0.01, "learning_rate": 4.984214169723554e-05, "loss": 1.2847, "step": 488 }, { "epoch": 0.01, "learning_rate": 4.984149473697831e-05, "loss": 1.4079, "step": 490 }, { "epoch": 0.01, "learning_rate": 4.9840847776721075e-05, "loss": 1.3523, "step": 492 }, { "epoch": 0.01, "learning_rate": 4.9840200816463844e-05, "loss": 1.4302, "step": 494 }, { "epoch": 0.01, "learning_rate": 4.983955385620661e-05, "loss": 1.5102, "step": 496 }, { "epoch": 0.01, "learning_rate": 4.983890689594938e-05, "loss": 1.2528, "step": 498 }, { "epoch": 0.01, "learning_rate": 4.983825993569216e-05, "loss": 1.2923, "step": 500 }, { "epoch": 0.01, "learning_rate": 4.983761297543492e-05, "loss": 1.1437, "step": 502 }, { "epoch": 0.01, "learning_rate": 4.983696601517769e-05, "loss": 1.628, "step": 504 }, { "epoch": 0.01, "learning_rate": 4.983631905492046e-05, "loss": 1.4683, "step": 506 }, { "epoch": 0.01, "learning_rate": 4.983567209466323e-05, "loss": 1.6244, "step": 508 }, { "epoch": 0.01, "learning_rate": 4.9835025134406e-05, "loss": 1.3527, "step": 510 }, { "epoch": 0.01, "learning_rate": 4.9834378174148766e-05, "loss": 1.1217, "step": 512 }, { "epoch": 0.01, "learning_rate": 4.9833731213891535e-05, "loss": 1.3489, "step": 514 }, { "epoch": 0.01, "learning_rate": 4.98330842536343e-05, "loss": 1.1912, "step": 516 }, { "epoch": 0.01, "learning_rate": 4.983243729337707e-05, "loss": 1.0923, "step": 518 }, { "epoch": 0.01, "learning_rate": 4.9831790333119836e-05, "loss": 1.3409, "step": 520 }, { "epoch": 0.01, "learning_rate": 4.983114337286261e-05, "loss": 1.4352, "step": 522 }, { "epoch": 0.01, "learning_rate": 4.9830496412605374e-05, "loss": 1.4428, "step": 524 }, { "epoch": 0.01, "learning_rate": 4.982984945234814e-05, "loss": 1.1087, "step": 526 }, { "epoch": 0.01, "learning_rate": 4.982920249209091e-05, "loss": 1.4864, "step": 528 }, { "epoch": 0.01, "learning_rate": 4.982855553183368e-05, "loss": 1.3159, "step": 530 }, { "epoch": 0.01, "learning_rate": 4.982790857157645e-05, "loss": 1.2209, "step": 532 }, { "epoch": 0.01, "learning_rate": 4.982726161131922e-05, "loss": 1.2949, "step": 534 }, { "epoch": 0.01, "learning_rate": 4.982661465106199e-05, "loss": 1.0704, "step": 536 }, { "epoch": 0.01, "learning_rate": 4.982596769080476e-05, "loss": 1.242, "step": 538 }, { "epoch": 0.01, "learning_rate": 4.9825320730547526e-05, "loss": 1.2367, "step": 540 }, { "epoch": 0.01, "learning_rate": 4.9824673770290296e-05, "loss": 1.394, "step": 542 }, { "epoch": 0.01, "learning_rate": 4.9824026810033065e-05, "loss": 1.184, "step": 544 }, { "epoch": 0.01, "learning_rate": 4.9823379849775834e-05, "loss": 1.1463, "step": 546 }, { "epoch": 0.01, "learning_rate": 4.9822732889518596e-05, "loss": 1.1476, "step": 548 }, { "epoch": 0.01, "learning_rate": 4.982208592926137e-05, "loss": 1.1113, "step": 550 }, { "epoch": 0.01, "learning_rate": 4.9821438969004134e-05, "loss": 1.3712, "step": 552 }, { "epoch": 0.01, "learning_rate": 4.98207920087469e-05, "loss": 1.3147, "step": 554 }, { "epoch": 0.01, "learning_rate": 4.982014504848967e-05, "loss": 1.0626, "step": 556 }, { "epoch": 0.01, "learning_rate": 4.981949808823244e-05, "loss": 1.4687, "step": 558 }, { "epoch": 0.01, "learning_rate": 4.981885112797521e-05, "loss": 1.1493, "step": 560 }, { "epoch": 0.01, "learning_rate": 4.981820416771798e-05, "loss": 1.3464, "step": 562 }, { "epoch": 0.01, "learning_rate": 4.981755720746075e-05, "loss": 1.0922, "step": 564 }, { "epoch": 0.01, "learning_rate": 4.981691024720352e-05, "loss": 1.2273, "step": 566 }, { "epoch": 0.01, "learning_rate": 4.981626328694629e-05, "loss": 1.0882, "step": 568 }, { "epoch": 0.01, "learning_rate": 4.981561632668905e-05, "loss": 1.0609, "step": 570 }, { "epoch": 0.01, "learning_rate": 4.9814969366431825e-05, "loss": 0.9874, "step": 572 }, { "epoch": 0.01, "learning_rate": 4.981432240617459e-05, "loss": 1.1738, "step": 574 }, { "epoch": 0.01, "learning_rate": 4.9813675445917356e-05, "loss": 1.0735, "step": 576 }, { "epoch": 0.01, "learning_rate": 4.981302848566013e-05, "loss": 1.1204, "step": 578 }, { "epoch": 0.01, "learning_rate": 4.9812381525402895e-05, "loss": 1.0066, "step": 580 }, { "epoch": 0.01, "learning_rate": 4.981173456514567e-05, "loss": 1.624, "step": 582 }, { "epoch": 0.01, "learning_rate": 4.981108760488843e-05, "loss": 1.235, "step": 584 }, { "epoch": 0.01, "learning_rate": 4.98104406446312e-05, "loss": 1.1288, "step": 586 }, { "epoch": 0.01, "learning_rate": 4.980979368437397e-05, "loss": 1.2266, "step": 588 }, { "epoch": 0.01, "learning_rate": 4.980914672411674e-05, "loss": 1.1235, "step": 590 }, { "epoch": 0.01, "learning_rate": 4.980849976385951e-05, "loss": 1.1192, "step": 592 }, { "epoch": 0.01, "learning_rate": 4.980785280360228e-05, "loss": 1.0076, "step": 594 }, { "epoch": 0.01, "learning_rate": 4.980720584334505e-05, "loss": 1.2006, "step": 596 }, { "epoch": 0.01, "learning_rate": 4.980655888308781e-05, "loss": 0.9436, "step": 598 }, { "epoch": 0.01, "learning_rate": 4.9805911922830585e-05, "loss": 1.131, "step": 600 }, { "epoch": 0.01, "learning_rate": 4.980526496257335e-05, "loss": 1.1674, "step": 602 }, { "epoch": 0.01, "learning_rate": 4.9804618002316124e-05, "loss": 1.1324, "step": 604 }, { "epoch": 0.01, "learning_rate": 4.9803971042058886e-05, "loss": 1.4155, "step": 606 }, { "epoch": 0.01, "learning_rate": 4.9803324081801655e-05, "loss": 1.1722, "step": 608 }, { "epoch": 0.01, "learning_rate": 4.980267712154443e-05, "loss": 1.0976, "step": 610 }, { "epoch": 0.01, "learning_rate": 4.980203016128719e-05, "loss": 1.2858, "step": 612 }, { "epoch": 0.01, "learning_rate": 4.980138320102996e-05, "loss": 0.9913, "step": 614 }, { "epoch": 0.01, "learning_rate": 4.980073624077273e-05, "loss": 0.9411, "step": 616 }, { "epoch": 0.01, "learning_rate": 4.98000892805155e-05, "loss": 1.0893, "step": 618 }, { "epoch": 0.01, "learning_rate": 4.979944232025827e-05, "loss": 1.1777, "step": 620 }, { "epoch": 0.01, "learning_rate": 4.979879536000104e-05, "loss": 0.9051, "step": 622 }, { "epoch": 0.01, "learning_rate": 4.979814839974381e-05, "loss": 1.0321, "step": 624 }, { "epoch": 0.01, "learning_rate": 4.979750143948658e-05, "loss": 1.0917, "step": 626 }, { "epoch": 0.01, "learning_rate": 4.9796854479229346e-05, "loss": 1.1503, "step": 628 }, { "epoch": 0.01, "learning_rate": 4.979620751897211e-05, "loss": 0.9256, "step": 630 }, { "epoch": 0.01, "learning_rate": 4.9795560558714884e-05, "loss": 0.9947, "step": 632 }, { "epoch": 0.01, "learning_rate": 4.9794913598457646e-05, "loss": 1.2583, "step": 634 }, { "epoch": 0.01, "learning_rate": 4.9794266638200415e-05, "loss": 1.0301, "step": 636 }, { "epoch": 0.01, "learning_rate": 4.9793619677943184e-05, "loss": 0.9304, "step": 638 }, { "epoch": 0.01, "learning_rate": 4.9792972717685954e-05, "loss": 0.9063, "step": 640 }, { "epoch": 0.01, "learning_rate": 4.979232575742873e-05, "loss": 1.1197, "step": 642 }, { "epoch": 0.01, "learning_rate": 4.979167879717149e-05, "loss": 0.939, "step": 644 }, { "epoch": 0.01, "learning_rate": 4.979103183691426e-05, "loss": 0.8917, "step": 646 }, { "epoch": 0.01, "learning_rate": 4.979038487665703e-05, "loss": 0.9756, "step": 648 }, { "epoch": 0.01, "learning_rate": 4.97897379163998e-05, "loss": 1.1025, "step": 650 }, { "epoch": 0.01, "learning_rate": 4.978909095614256e-05, "loss": 1.0281, "step": 652 }, { "epoch": 0.01, "learning_rate": 4.978844399588534e-05, "loss": 0.8539, "step": 654 }, { "epoch": 0.01, "learning_rate": 4.9787797035628106e-05, "loss": 0.9905, "step": 656 }, { "epoch": 0.01, "learning_rate": 4.978715007537087e-05, "loss": 0.9046, "step": 658 }, { "epoch": 0.01, "learning_rate": 4.9786503115113644e-05, "loss": 1.0028, "step": 660 }, { "epoch": 0.01, "learning_rate": 4.978585615485641e-05, "loss": 0.8714, "step": 662 }, { "epoch": 0.01, "learning_rate": 4.978520919459918e-05, "loss": 1.0507, "step": 664 }, { "epoch": 0.01, "learning_rate": 4.9784562234341945e-05, "loss": 0.8658, "step": 666 }, { "epoch": 0.01, "learning_rate": 4.9783915274084714e-05, "loss": 0.9604, "step": 668 }, { "epoch": 0.01, "learning_rate": 4.978326831382748e-05, "loss": 1.0652, "step": 670 }, { "epoch": 0.01, "learning_rate": 4.978262135357025e-05, "loss": 1.0489, "step": 672 }, { "epoch": 0.01, "learning_rate": 4.978197439331302e-05, "loss": 0.9805, "step": 674 }, { "epoch": 0.01, "learning_rate": 4.978132743305579e-05, "loss": 0.9611, "step": 676 }, { "epoch": 0.01, "learning_rate": 4.978068047279856e-05, "loss": 1.0882, "step": 678 }, { "epoch": 0.01, "learning_rate": 4.978003351254133e-05, "loss": 0.9495, "step": 680 }, { "epoch": 0.01, "learning_rate": 4.97793865522841e-05, "loss": 1.0601, "step": 682 }, { "epoch": 0.01, "learning_rate": 4.977873959202686e-05, "loss": 1.0151, "step": 684 }, { "epoch": 0.01, "learning_rate": 4.9778092631769636e-05, "loss": 0.7641, "step": 686 }, { "epoch": 0.01, "learning_rate": 4.9777445671512405e-05, "loss": 0.8845, "step": 688 }, { "epoch": 0.01, "learning_rate": 4.977679871125517e-05, "loss": 0.8677, "step": 690 }, { "epoch": 0.01, "learning_rate": 4.977615175099794e-05, "loss": 1.1119, "step": 692 }, { "epoch": 0.01, "learning_rate": 4.9775504790740705e-05, "loss": 0.7907, "step": 694 }, { "epoch": 0.01, "learning_rate": 4.9774857830483474e-05, "loss": 1.0665, "step": 696 }, { "epoch": 0.01, "learning_rate": 4.9774210870226243e-05, "loss": 0.9662, "step": 698 }, { "epoch": 0.01, "learning_rate": 4.977356390996901e-05, "loss": 0.7631, "step": 700 }, { "epoch": 0.01, "learning_rate": 4.977291694971178e-05, "loss": 0.9225, "step": 702 }, { "epoch": 0.01, "learning_rate": 4.977226998945455e-05, "loss": 0.7057, "step": 704 }, { "epoch": 0.01, "learning_rate": 4.977162302919732e-05, "loss": 0.8489, "step": 706 }, { "epoch": 0.01, "learning_rate": 4.977097606894009e-05, "loss": 0.7504, "step": 708 }, { "epoch": 0.01, "learning_rate": 4.977032910868286e-05, "loss": 1.0052, "step": 710 }, { "epoch": 0.01, "learning_rate": 4.976968214842562e-05, "loss": 0.8506, "step": 712 }, { "epoch": 0.01, "learning_rate": 4.9769035188168396e-05, "loss": 0.6797, "step": 714 }, { "epoch": 0.01, "learning_rate": 4.976838822791116e-05, "loss": 0.8483, "step": 716 }, { "epoch": 0.01, "learning_rate": 4.976774126765393e-05, "loss": 0.9958, "step": 718 }, { "epoch": 0.01, "learning_rate": 4.97670943073967e-05, "loss": 0.8406, "step": 720 }, { "epoch": 0.01, "learning_rate": 4.9766447347139466e-05, "loss": 0.8617, "step": 722 }, { "epoch": 0.01, "learning_rate": 4.976580038688224e-05, "loss": 0.7285, "step": 724 }, { "epoch": 0.01, "learning_rate": 4.9765153426625004e-05, "loss": 0.847, "step": 726 }, { "epoch": 0.01, "learning_rate": 4.976450646636777e-05, "loss": 0.9041, "step": 728 }, { "epoch": 0.01, "learning_rate": 4.976385950611054e-05, "loss": 0.6769, "step": 730 }, { "epoch": 0.01, "learning_rate": 4.976321254585331e-05, "loss": 0.9203, "step": 732 }, { "epoch": 0.01, "learning_rate": 4.976256558559608e-05, "loss": 0.6796, "step": 734 }, { "epoch": 0.01, "learning_rate": 4.976191862533885e-05, "loss": 0.7871, "step": 736 }, { "epoch": 0.01, "learning_rate": 4.976127166508162e-05, "loss": 0.7575, "step": 738 }, { "epoch": 0.01, "learning_rate": 4.976062470482438e-05, "loss": 0.7714, "step": 740 }, { "epoch": 0.01, "learning_rate": 4.9759977744567157e-05, "loss": 0.6998, "step": 742 }, { "epoch": 0.01, "learning_rate": 4.975933078430992e-05, "loss": 0.7788, "step": 744 }, { "epoch": 0.01, "learning_rate": 4.9758683824052695e-05, "loss": 0.8785, "step": 746 }, { "epoch": 0.01, "learning_rate": 4.975803686379546e-05, "loss": 0.8712, "step": 748 }, { "epoch": 0.01, "learning_rate": 4.9757389903538226e-05, "loss": 0.8119, "step": 750 }, { "epoch": 0.01, "learning_rate": 4.9756742943280995e-05, "loss": 0.7014, "step": 752 }, { "epoch": 0.01, "learning_rate": 4.9756095983023764e-05, "loss": 1.0898, "step": 754 }, { "epoch": 0.01, "learning_rate": 4.975544902276653e-05, "loss": 0.7692, "step": 756 }, { "epoch": 0.01, "learning_rate": 4.97548020625093e-05, "loss": 0.8054, "step": 758 }, { "epoch": 0.01, "learning_rate": 4.975415510225207e-05, "loss": 0.8432, "step": 760 }, { "epoch": 0.01, "learning_rate": 4.975350814199484e-05, "loss": 0.7216, "step": 762 }, { "epoch": 0.01, "learning_rate": 4.975286118173761e-05, "loss": 0.7077, "step": 764 }, { "epoch": 0.01, "learning_rate": 4.975221422148038e-05, "loss": 1.0554, "step": 766 }, { "epoch": 0.01, "learning_rate": 4.975156726122315e-05, "loss": 0.8534, "step": 768 }, { "epoch": 0.01, "learning_rate": 4.975092030096592e-05, "loss": 0.9094, "step": 770 }, { "epoch": 0.01, "learning_rate": 4.975027334070868e-05, "loss": 0.7504, "step": 772 }, { "epoch": 0.02, "learning_rate": 4.9749626380451455e-05, "loss": 0.6936, "step": 774 }, { "epoch": 0.02, "learning_rate": 4.974897942019422e-05, "loss": 0.8697, "step": 776 }, { "epoch": 0.02, "learning_rate": 4.9748332459936986e-05, "loss": 0.7018, "step": 778 }, { "epoch": 0.02, "learning_rate": 4.9747685499679756e-05, "loss": 0.8992, "step": 780 }, { "epoch": 0.02, "learning_rate": 4.9747038539422525e-05, "loss": 0.6812, "step": 782 }, { "epoch": 0.02, "learning_rate": 4.9746391579165294e-05, "loss": 0.8636, "step": 784 }, { "epoch": 0.02, "learning_rate": 4.974574461890806e-05, "loss": 0.8034, "step": 786 }, { "epoch": 0.02, "learning_rate": 4.974509765865083e-05, "loss": 0.6331, "step": 788 }, { "epoch": 0.02, "learning_rate": 4.97444506983936e-05, "loss": 0.9043, "step": 790 }, { "epoch": 0.02, "learning_rate": 4.974380373813637e-05, "loss": 0.7185, "step": 792 }, { "epoch": 0.02, "learning_rate": 4.974315677787913e-05, "loss": 0.7028, "step": 794 }, { "epoch": 0.02, "learning_rate": 4.974250981762191e-05, "loss": 0.7783, "step": 796 }, { "epoch": 0.02, "learning_rate": 4.974186285736467e-05, "loss": 0.834, "step": 798 }, { "epoch": 0.02, "learning_rate": 4.974121589710744e-05, "loss": 0.7198, "step": 800 }, { "epoch": 0.02, "learning_rate": 4.9740568936850215e-05, "loss": 0.7112, "step": 802 }, { "epoch": 0.02, "learning_rate": 4.973992197659298e-05, "loss": 0.7774, "step": 804 }, { "epoch": 0.02, "learning_rate": 4.9739275016335754e-05, "loss": 0.8853, "step": 806 }, { "epoch": 0.02, "learning_rate": 4.9738628056078516e-05, "loss": 0.7231, "step": 808 }, { "epoch": 0.02, "learning_rate": 4.9737981095821285e-05, "loss": 0.7044, "step": 810 }, { "epoch": 0.02, "learning_rate": 4.9737334135564054e-05, "loss": 0.6994, "step": 812 }, { "epoch": 0.02, "learning_rate": 4.973668717530682e-05, "loss": 0.8031, "step": 814 }, { "epoch": 0.02, "learning_rate": 4.973604021504959e-05, "loss": 0.7543, "step": 816 }, { "epoch": 0.02, "learning_rate": 4.973539325479236e-05, "loss": 0.7793, "step": 818 }, { "epoch": 0.02, "learning_rate": 4.973474629453513e-05, "loss": 0.6948, "step": 820 }, { "epoch": 0.02, "learning_rate": 4.973409933427789e-05, "loss": 0.6267, "step": 822 }, { "epoch": 0.02, "learning_rate": 4.973345237402067e-05, "loss": 0.5628, "step": 824 }, { "epoch": 0.02, "learning_rate": 4.973280541376343e-05, "loss": 0.722, "step": 826 }, { "epoch": 0.02, "learning_rate": 4.973215845350621e-05, "loss": 0.6738, "step": 828 }, { "epoch": 0.02, "learning_rate": 4.973151149324897e-05, "loss": 0.8005, "step": 830 }, { "epoch": 0.02, "learning_rate": 4.973086453299174e-05, "loss": 0.7117, "step": 832 }, { "epoch": 0.02, "learning_rate": 4.9730217572734514e-05, "loss": 0.5747, "step": 834 }, { "epoch": 0.02, "learning_rate": 4.9729570612477276e-05, "loss": 0.7542, "step": 836 }, { "epoch": 0.02, "learning_rate": 4.9728923652220045e-05, "loss": 0.8115, "step": 838 }, { "epoch": 0.02, "learning_rate": 4.9728276691962815e-05, "loss": 0.7017, "step": 840 }, { "epoch": 0.02, "learning_rate": 4.9727629731705584e-05, "loss": 0.5835, "step": 842 }, { "epoch": 0.02, "learning_rate": 4.972698277144835e-05, "loss": 0.6623, "step": 844 }, { "epoch": 0.02, "learning_rate": 4.972633581119112e-05, "loss": 0.7123, "step": 846 }, { "epoch": 0.02, "learning_rate": 4.972568885093389e-05, "loss": 0.5549, "step": 848 }, { "epoch": 0.02, "learning_rate": 4.972504189067666e-05, "loss": 0.6999, "step": 850 }, { "epoch": 0.02, "learning_rate": 4.972439493041943e-05, "loss": 0.571, "step": 852 }, { "epoch": 0.02, "learning_rate": 4.972374797016219e-05, "loss": 0.6145, "step": 854 }, { "epoch": 0.02, "learning_rate": 4.972310100990497e-05, "loss": 0.6057, "step": 856 }, { "epoch": 0.02, "learning_rate": 4.972245404964773e-05, "loss": 0.6443, "step": 858 }, { "epoch": 0.02, "learning_rate": 4.97218070893905e-05, "loss": 0.7058, "step": 860 }, { "epoch": 0.02, "learning_rate": 4.972116012913327e-05, "loss": 0.8731, "step": 862 }, { "epoch": 0.02, "learning_rate": 4.972051316887604e-05, "loss": 0.6529, "step": 864 }, { "epoch": 0.02, "learning_rate": 4.971986620861881e-05, "loss": 0.6056, "step": 866 }, { "epoch": 0.02, "learning_rate": 4.9719219248361575e-05, "loss": 0.5553, "step": 868 }, { "epoch": 0.02, "learning_rate": 4.9718572288104344e-05, "loss": 0.6491, "step": 870 }, { "epoch": 0.02, "learning_rate": 4.971792532784711e-05, "loss": 0.7516, "step": 872 }, { "epoch": 0.02, "learning_rate": 4.971727836758988e-05, "loss": 0.6776, "step": 874 }, { "epoch": 0.02, "learning_rate": 4.9716631407332645e-05, "loss": 0.5225, "step": 876 }, { "epoch": 0.02, "learning_rate": 4.971598444707542e-05, "loss": 0.6561, "step": 878 }, { "epoch": 0.02, "learning_rate": 4.971533748681819e-05, "loss": 0.6347, "step": 880 }, { "epoch": 0.02, "learning_rate": 4.971469052656095e-05, "loss": 0.6, "step": 882 }, { "epoch": 0.02, "learning_rate": 4.971404356630373e-05, "loss": 0.6632, "step": 884 }, { "epoch": 0.02, "learning_rate": 4.971339660604649e-05, "loss": 0.6133, "step": 886 }, { "epoch": 0.02, "learning_rate": 4.9712749645789266e-05, "loss": 0.8037, "step": 888 }, { "epoch": 0.02, "learning_rate": 4.971210268553203e-05, "loss": 0.5696, "step": 890 }, { "epoch": 0.02, "learning_rate": 4.97114557252748e-05, "loss": 0.5888, "step": 892 }, { "epoch": 0.02, "learning_rate": 4.9710808765017566e-05, "loss": 0.6084, "step": 894 }, { "epoch": 0.02, "learning_rate": 4.9710161804760335e-05, "loss": 0.7891, "step": 896 }, { "epoch": 0.02, "learning_rate": 4.9709514844503104e-05, "loss": 0.8151, "step": 898 }, { "epoch": 0.02, "learning_rate": 4.9708867884245874e-05, "loss": 0.6607, "step": 900 }, { "epoch": 0.02, "learning_rate": 4.970822092398864e-05, "loss": 0.6793, "step": 902 }, { "epoch": 0.02, "learning_rate": 4.970757396373141e-05, "loss": 0.5348, "step": 904 }, { "epoch": 0.02, "learning_rate": 4.970692700347418e-05, "loss": 0.7303, "step": 906 }, { "epoch": 0.02, "learning_rate": 4.970628004321694e-05, "loss": 0.7285, "step": 908 }, { "epoch": 0.02, "learning_rate": 4.970563308295972e-05, "loss": 0.7019, "step": 910 }, { "epoch": 0.02, "learning_rate": 4.970498612270249e-05, "loss": 0.5888, "step": 912 }, { "epoch": 0.02, "learning_rate": 4.970433916244525e-05, "loss": 0.5973, "step": 914 }, { "epoch": 0.02, "learning_rate": 4.9703692202188026e-05, "loss": 0.6444, "step": 916 }, { "epoch": 0.02, "learning_rate": 4.970304524193079e-05, "loss": 0.6158, "step": 918 }, { "epoch": 0.02, "learning_rate": 4.970239828167356e-05, "loss": 0.4536, "step": 920 }, { "epoch": 0.02, "learning_rate": 4.970175132141633e-05, "loss": 0.7304, "step": 922 }, { "epoch": 0.02, "learning_rate": 4.9701104361159096e-05, "loss": 0.5191, "step": 924 }, { "epoch": 0.02, "learning_rate": 4.9700457400901865e-05, "loss": 0.5434, "step": 926 }, { "epoch": 0.02, "learning_rate": 4.9699810440644634e-05, "loss": 0.5497, "step": 928 }, { "epoch": 0.02, "learning_rate": 4.96991634803874e-05, "loss": 0.608, "step": 930 }, { "epoch": 0.02, "learning_rate": 4.969851652013017e-05, "loss": 0.6695, "step": 932 }, { "epoch": 0.02, "learning_rate": 4.969786955987294e-05, "loss": 0.5534, "step": 934 }, { "epoch": 0.02, "learning_rate": 4.9697222599615703e-05, "loss": 0.4371, "step": 936 }, { "epoch": 0.02, "learning_rate": 4.969657563935848e-05, "loss": 0.5756, "step": 938 }, { "epoch": 0.02, "learning_rate": 4.969592867910124e-05, "loss": 0.4919, "step": 940 }, { "epoch": 0.02, "learning_rate": 4.969528171884401e-05, "loss": 0.6685, "step": 942 }, { "epoch": 0.02, "learning_rate": 4.9694634758586787e-05, "loss": 0.5089, "step": 944 }, { "epoch": 0.02, "learning_rate": 4.969398779832955e-05, "loss": 0.5833, "step": 946 }, { "epoch": 0.02, "learning_rate": 4.9693340838072325e-05, "loss": 0.6404, "step": 948 }, { "epoch": 0.02, "learning_rate": 4.969269387781509e-05, "loss": 0.5909, "step": 950 }, { "epoch": 0.02, "learning_rate": 4.9692046917557856e-05, "loss": 0.5073, "step": 952 }, { "epoch": 0.02, "learning_rate": 4.9691399957300625e-05, "loss": 0.6165, "step": 954 }, { "epoch": 0.02, "learning_rate": 4.9690752997043394e-05, "loss": 0.5664, "step": 956 }, { "epoch": 0.02, "learning_rate": 4.9690106036786163e-05, "loss": 0.5624, "step": 958 }, { "epoch": 0.02, "learning_rate": 4.968945907652893e-05, "loss": 0.5847, "step": 960 }, { "epoch": 0.02, "learning_rate": 4.96888121162717e-05, "loss": 0.4451, "step": 962 }, { "epoch": 0.02, "learning_rate": 4.9688165156014464e-05, "loss": 0.5245, "step": 964 }, { "epoch": 0.02, "learning_rate": 4.968751819575724e-05, "loss": 0.5547, "step": 966 }, { "epoch": 0.02, "learning_rate": 4.96868712355e-05, "loss": 0.4746, "step": 968 }, { "epoch": 0.02, "learning_rate": 4.968622427524278e-05, "loss": 0.5271, "step": 970 }, { "epoch": 0.02, "learning_rate": 4.968557731498554e-05, "loss": 0.5886, "step": 972 }, { "epoch": 0.02, "learning_rate": 4.968493035472831e-05, "loss": 0.6671, "step": 974 }, { "epoch": 0.02, "learning_rate": 4.968428339447108e-05, "loss": 0.628, "step": 976 }, { "epoch": 0.02, "learning_rate": 4.968363643421385e-05, "loss": 0.7741, "step": 978 }, { "epoch": 0.02, "learning_rate": 4.9682989473956617e-05, "loss": 0.6094, "step": 980 }, { "epoch": 0.02, "learning_rate": 4.9682342513699386e-05, "loss": 0.5116, "step": 982 }, { "epoch": 0.02, "learning_rate": 4.9681695553442155e-05, "loss": 0.6362, "step": 984 }, { "epoch": 0.02, "learning_rate": 4.9681048593184924e-05, "loss": 0.552, "step": 986 }, { "epoch": 0.02, "learning_rate": 4.968040163292769e-05, "loss": 0.4485, "step": 988 }, { "epoch": 0.02, "learning_rate": 4.967975467267046e-05, "loss": 0.6242, "step": 990 }, { "epoch": 0.02, "learning_rate": 4.967910771241323e-05, "loss": 0.4603, "step": 992 }, { "epoch": 0.02, "learning_rate": 4.9678460752156e-05, "loss": 0.5442, "step": 994 }, { "epoch": 0.02, "learning_rate": 4.967781379189876e-05, "loss": 0.4921, "step": 996 }, { "epoch": 0.02, "learning_rate": 4.967716683164154e-05, "loss": 0.4821, "step": 998 }, { "epoch": 0.02, "learning_rate": 4.96765198713843e-05, "loss": 0.4329, "step": 1000 }, { "epoch": 0.02, "learning_rate": 4.967587291112707e-05, "loss": 0.4271, "step": 1002 }, { "epoch": 0.02, "learning_rate": 4.967522595086984e-05, "loss": 0.5235, "step": 1004 }, { "epoch": 0.02, "learning_rate": 4.967457899061261e-05, "loss": 0.6941, "step": 1006 }, { "epoch": 0.02, "learning_rate": 4.967393203035538e-05, "loss": 0.5899, "step": 1008 }, { "epoch": 0.02, "learning_rate": 4.9673285070098146e-05, "loss": 0.4467, "step": 1010 }, { "epoch": 0.02, "learning_rate": 4.9672638109840915e-05, "loss": 0.5929, "step": 1012 }, { "epoch": 0.02, "learning_rate": 4.9671991149583684e-05, "loss": 0.4744, "step": 1014 }, { "epoch": 0.02, "learning_rate": 4.967134418932645e-05, "loss": 0.4396, "step": 1016 }, { "epoch": 0.02, "learning_rate": 4.9670697229069216e-05, "loss": 0.5701, "step": 1018 }, { "epoch": 0.02, "learning_rate": 4.967005026881199e-05, "loss": 0.4756, "step": 1020 }, { "epoch": 0.02, "learning_rate": 4.966940330855476e-05, "loss": 0.4459, "step": 1022 }, { "epoch": 0.02, "learning_rate": 4.966875634829752e-05, "loss": 0.5305, "step": 1024 }, { "epoch": 0.02, "learning_rate": 4.96681093880403e-05, "loss": 0.4727, "step": 1026 }, { "epoch": 0.02, "learning_rate": 4.966746242778306e-05, "loss": 0.522, "step": 1028 }, { "epoch": 0.02, "learning_rate": 4.966681546752584e-05, "loss": 0.4388, "step": 1030 }, { "epoch": 0.02, "learning_rate": 4.96661685072686e-05, "loss": 0.5361, "step": 1032 }, { "epoch": 0.02, "learning_rate": 4.966552154701137e-05, "loss": 0.5232, "step": 1034 }, { "epoch": 0.02, "learning_rate": 4.966487458675414e-05, "loss": 0.5081, "step": 1036 }, { "epoch": 0.02, "learning_rate": 4.9664227626496906e-05, "loss": 0.7518, "step": 1038 }, { "epoch": 0.02, "learning_rate": 4.9663580666239676e-05, "loss": 0.5073, "step": 1040 }, { "epoch": 0.02, "learning_rate": 4.9662933705982445e-05, "loss": 0.6093, "step": 1042 }, { "epoch": 0.02, "learning_rate": 4.9662286745725214e-05, "loss": 0.6018, "step": 1044 }, { "epoch": 0.02, "learning_rate": 4.966163978546798e-05, "loss": 0.4355, "step": 1046 }, { "epoch": 0.02, "learning_rate": 4.966099282521075e-05, "loss": 0.4858, "step": 1048 }, { "epoch": 0.02, "learning_rate": 4.9660345864953514e-05, "loss": 0.5694, "step": 1050 }, { "epoch": 0.02, "learning_rate": 4.965969890469629e-05, "loss": 0.4954, "step": 1052 }, { "epoch": 0.02, "learning_rate": 4.965905194443905e-05, "loss": 0.5453, "step": 1054 }, { "epoch": 0.02, "learning_rate": 4.965840498418182e-05, "loss": 0.4887, "step": 1056 }, { "epoch": 0.02, "learning_rate": 4.96577580239246e-05, "loss": 0.5347, "step": 1058 }, { "epoch": 0.02, "learning_rate": 4.965711106366736e-05, "loss": 0.3958, "step": 1060 }, { "epoch": 0.02, "learning_rate": 4.965646410341013e-05, "loss": 0.4504, "step": 1062 }, { "epoch": 0.02, "learning_rate": 4.96558171431529e-05, "loss": 0.4699, "step": 1064 }, { "epoch": 0.02, "learning_rate": 4.965517018289567e-05, "loss": 0.4664, "step": 1066 }, { "epoch": 0.02, "learning_rate": 4.9654523222638436e-05, "loss": 0.5887, "step": 1068 }, { "epoch": 0.02, "learning_rate": 4.9653876262381205e-05, "loss": 0.5597, "step": 1070 }, { "epoch": 0.02, "learning_rate": 4.9653229302123974e-05, "loss": 0.4538, "step": 1072 }, { "epoch": 0.02, "learning_rate": 4.965258234186674e-05, "loss": 0.4959, "step": 1074 }, { "epoch": 0.02, "learning_rate": 4.965193538160951e-05, "loss": 0.4517, "step": 1076 }, { "epoch": 0.02, "learning_rate": 4.9651288421352275e-05, "loss": 0.4309, "step": 1078 }, { "epoch": 0.02, "learning_rate": 4.965064146109505e-05, "loss": 0.5779, "step": 1080 }, { "epoch": 0.02, "learning_rate": 4.964999450083781e-05, "loss": 0.4482, "step": 1082 }, { "epoch": 0.02, "learning_rate": 4.964934754058058e-05, "loss": 0.46, "step": 1084 }, { "epoch": 0.02, "learning_rate": 4.964870058032335e-05, "loss": 0.4548, "step": 1086 }, { "epoch": 0.02, "learning_rate": 4.964805362006612e-05, "loss": 0.3843, "step": 1088 }, { "epoch": 0.02, "learning_rate": 4.9647406659808896e-05, "loss": 0.3666, "step": 1090 }, { "epoch": 0.02, "learning_rate": 4.964675969955166e-05, "loss": 0.5016, "step": 1092 }, { "epoch": 0.02, "learning_rate": 4.964611273929443e-05, "loss": 0.4142, "step": 1094 }, { "epoch": 0.02, "learning_rate": 4.9645465779037196e-05, "loss": 0.4366, "step": 1096 }, { "epoch": 0.02, "learning_rate": 4.9644818818779965e-05, "loss": 0.4656, "step": 1098 }, { "epoch": 0.02, "learning_rate": 4.964417185852273e-05, "loss": 0.4336, "step": 1100 }, { "epoch": 0.02, "learning_rate": 4.9643524898265504e-05, "loss": 0.5804, "step": 1102 }, { "epoch": 0.02, "learning_rate": 4.964287793800827e-05, "loss": 0.6062, "step": 1104 }, { "epoch": 0.02, "learning_rate": 4.9642230977751035e-05, "loss": 0.3814, "step": 1106 }, { "epoch": 0.02, "learning_rate": 4.964158401749381e-05, "loss": 0.4013, "step": 1108 }, { "epoch": 0.02, "learning_rate": 4.964093705723657e-05, "loss": 0.5414, "step": 1110 }, { "epoch": 0.02, "learning_rate": 4.964029009697935e-05, "loss": 0.4353, "step": 1112 }, { "epoch": 0.02, "learning_rate": 4.963964313672211e-05, "loss": 0.4456, "step": 1114 }, { "epoch": 0.02, "learning_rate": 4.963899617646488e-05, "loss": 0.4383, "step": 1116 }, { "epoch": 0.02, "learning_rate": 4.963834921620765e-05, "loss": 0.5132, "step": 1118 }, { "epoch": 0.02, "learning_rate": 4.963770225595042e-05, "loss": 0.3433, "step": 1120 }, { "epoch": 0.02, "learning_rate": 4.963705529569319e-05, "loss": 0.3474, "step": 1122 }, { "epoch": 0.02, "learning_rate": 4.963640833543596e-05, "loss": 0.3582, "step": 1124 }, { "epoch": 0.02, "learning_rate": 4.9635761375178726e-05, "loss": 0.3026, "step": 1126 }, { "epoch": 0.02, "learning_rate": 4.9635114414921495e-05, "loss": 0.3546, "step": 1128 }, { "epoch": 0.02, "learning_rate": 4.9634467454664264e-05, "loss": 0.346, "step": 1130 }, { "epoch": 0.02, "learning_rate": 4.9633820494407026e-05, "loss": 0.4242, "step": 1132 }, { "epoch": 0.02, "learning_rate": 4.96331735341498e-05, "loss": 0.6713, "step": 1134 }, { "epoch": 0.02, "learning_rate": 4.963252657389257e-05, "loss": 0.4483, "step": 1136 }, { "epoch": 0.02, "learning_rate": 4.9631879613635334e-05, "loss": 0.4197, "step": 1138 }, { "epoch": 0.02, "learning_rate": 4.963123265337811e-05, "loss": 0.3141, "step": 1140 }, { "epoch": 0.02, "learning_rate": 4.963058569312087e-05, "loss": 0.4606, "step": 1142 }, { "epoch": 0.02, "learning_rate": 4.962993873286364e-05, "loss": 0.4751, "step": 1144 }, { "epoch": 0.02, "learning_rate": 4.962929177260641e-05, "loss": 0.5025, "step": 1146 }, { "epoch": 0.02, "learning_rate": 4.962864481234918e-05, "loss": 0.3066, "step": 1148 }, { "epoch": 0.02, "learning_rate": 4.962799785209195e-05, "loss": 0.3967, "step": 1150 }, { "epoch": 0.02, "learning_rate": 4.962735089183472e-05, "loss": 0.3945, "step": 1152 }, { "epoch": 0.02, "learning_rate": 4.9626703931577486e-05, "loss": 0.3273, "step": 1154 }, { "epoch": 0.02, "learning_rate": 4.9626056971320255e-05, "loss": 0.3783, "step": 1156 }, { "epoch": 0.02, "learning_rate": 4.9625410011063024e-05, "loss": 0.4007, "step": 1158 }, { "epoch": 0.02, "learning_rate": 4.962476305080579e-05, "loss": 0.4949, "step": 1160 }, { "epoch": 0.02, "learning_rate": 4.962411609054856e-05, "loss": 0.4349, "step": 1162 }, { "epoch": 0.02, "learning_rate": 4.9623469130291325e-05, "loss": 0.4244, "step": 1164 }, { "epoch": 0.02, "learning_rate": 4.9622822170034094e-05, "loss": 0.3508, "step": 1166 }, { "epoch": 0.02, "learning_rate": 4.962217520977687e-05, "loss": 0.5813, "step": 1168 }, { "epoch": 0.02, "learning_rate": 4.962152824951963e-05, "loss": 0.4225, "step": 1170 }, { "epoch": 0.02, "learning_rate": 4.962088128926241e-05, "loss": 0.4451, "step": 1172 }, { "epoch": 0.02, "learning_rate": 4.962023432900517e-05, "loss": 0.4744, "step": 1174 }, { "epoch": 0.02, "learning_rate": 4.961958736874794e-05, "loss": 0.4156, "step": 1176 }, { "epoch": 0.02, "learning_rate": 4.961894040849071e-05, "loss": 0.5055, "step": 1178 }, { "epoch": 0.02, "learning_rate": 4.961829344823348e-05, "loss": 0.366, "step": 1180 }, { "epoch": 0.02, "learning_rate": 4.961764648797625e-05, "loss": 0.37, "step": 1182 }, { "epoch": 0.02, "learning_rate": 4.9616999527719016e-05, "loss": 0.2947, "step": 1184 }, { "epoch": 0.02, "learning_rate": 4.9616352567461785e-05, "loss": 0.3946, "step": 1186 }, { "epoch": 0.02, "learning_rate": 4.9615705607204554e-05, "loss": 0.4116, "step": 1188 }, { "epoch": 0.02, "learning_rate": 4.961505864694732e-05, "loss": 0.4889, "step": 1190 }, { "epoch": 0.02, "learning_rate": 4.9614411686690085e-05, "loss": 0.3918, "step": 1192 }, { "epoch": 0.02, "learning_rate": 4.961376472643286e-05, "loss": 0.3559, "step": 1194 }, { "epoch": 0.02, "learning_rate": 4.9613117766175623e-05, "loss": 0.4197, "step": 1196 }, { "epoch": 0.02, "learning_rate": 4.961247080591839e-05, "loss": 0.371, "step": 1198 }, { "epoch": 0.02, "learning_rate": 4.961182384566117e-05, "loss": 0.3228, "step": 1200 }, { "epoch": 0.02, "learning_rate": 4.961117688540393e-05, "loss": 0.5729, "step": 1202 }, { "epoch": 0.02, "learning_rate": 4.96105299251467e-05, "loss": 0.3301, "step": 1204 }, { "epoch": 0.02, "learning_rate": 4.960988296488947e-05, "loss": 0.3658, "step": 1206 }, { "epoch": 0.02, "learning_rate": 4.960923600463224e-05, "loss": 0.3167, "step": 1208 }, { "epoch": 0.02, "learning_rate": 4.960858904437501e-05, "loss": 0.3199, "step": 1210 }, { "epoch": 0.02, "learning_rate": 4.9607942084117776e-05, "loss": 0.3494, "step": 1212 }, { "epoch": 0.02, "learning_rate": 4.9607295123860545e-05, "loss": 0.387, "step": 1214 }, { "epoch": 0.02, "learning_rate": 4.9606648163603314e-05, "loss": 0.459, "step": 1216 }, { "epoch": 0.02, "learning_rate": 4.9606001203346083e-05, "loss": 0.3653, "step": 1218 }, { "epoch": 0.02, "learning_rate": 4.9605354243088846e-05, "loss": 0.3314, "step": 1220 }, { "epoch": 0.02, "learning_rate": 4.960470728283162e-05, "loss": 0.3217, "step": 1222 }, { "epoch": 0.02, "learning_rate": 4.9604060322574384e-05, "loss": 0.3239, "step": 1224 }, { "epoch": 0.02, "learning_rate": 4.960341336231715e-05, "loss": 0.3069, "step": 1226 }, { "epoch": 0.02, "learning_rate": 4.960276640205992e-05, "loss": 0.284, "step": 1228 }, { "epoch": 0.02, "learning_rate": 4.960211944180269e-05, "loss": 0.3788, "step": 1230 }, { "epoch": 0.02, "learning_rate": 4.960147248154546e-05, "loss": 0.4107, "step": 1232 }, { "epoch": 0.02, "learning_rate": 4.960082552128823e-05, "loss": 0.3752, "step": 1234 }, { "epoch": 0.02, "learning_rate": 4.9600178561031e-05, "loss": 0.3517, "step": 1236 }, { "epoch": 0.02, "learning_rate": 4.959953160077377e-05, "loss": 0.3658, "step": 1238 }, { "epoch": 0.02, "learning_rate": 4.9598884640516537e-05, "loss": 0.3118, "step": 1240 }, { "epoch": 0.02, "learning_rate": 4.95982376802593e-05, "loss": 0.4758, "step": 1242 }, { "epoch": 0.02, "learning_rate": 4.9597590720002075e-05, "loss": 0.3192, "step": 1244 }, { "epoch": 0.02, "learning_rate": 4.9596943759744844e-05, "loss": 0.3145, "step": 1246 }, { "epoch": 0.02, "learning_rate": 4.9596296799487606e-05, "loss": 0.3179, "step": 1248 }, { "epoch": 0.02, "learning_rate": 4.959564983923038e-05, "loss": 0.3923, "step": 1250 }, { "epoch": 0.02, "learning_rate": 4.9595002878973144e-05, "loss": 0.363, "step": 1252 }, { "epoch": 0.02, "learning_rate": 4.959435591871592e-05, "loss": 0.4115, "step": 1254 }, { "epoch": 0.02, "learning_rate": 4.959370895845868e-05, "loss": 0.336, "step": 1256 }, { "epoch": 0.02, "learning_rate": 4.959306199820145e-05, "loss": 0.5084, "step": 1258 }, { "epoch": 0.02, "learning_rate": 4.959241503794422e-05, "loss": 0.3782, "step": 1260 }, { "epoch": 0.02, "learning_rate": 4.959176807768699e-05, "loss": 0.3089, "step": 1262 }, { "epoch": 0.02, "learning_rate": 4.959112111742976e-05, "loss": 0.3181, "step": 1264 }, { "epoch": 0.02, "learning_rate": 4.959047415717253e-05, "loss": 0.4271, "step": 1266 }, { "epoch": 0.02, "learning_rate": 4.95898271969153e-05, "loss": 0.3027, "step": 1268 }, { "epoch": 0.02, "learning_rate": 4.9589180236658066e-05, "loss": 0.4159, "step": 1270 }, { "epoch": 0.02, "learning_rate": 4.9588533276400835e-05, "loss": 0.4163, "step": 1272 }, { "epoch": 0.02, "learning_rate": 4.95878863161436e-05, "loss": 0.3023, "step": 1274 }, { "epoch": 0.02, "learning_rate": 4.958723935588637e-05, "loss": 0.3419, "step": 1276 }, { "epoch": 0.02, "learning_rate": 4.9586592395629136e-05, "loss": 0.4032, "step": 1278 }, { "epoch": 0.02, "learning_rate": 4.9585945435371905e-05, "loss": 0.4222, "step": 1280 }, { "epoch": 0.02, "learning_rate": 4.958529847511468e-05, "loss": 0.358, "step": 1282 }, { "epoch": 0.02, "learning_rate": 4.958465151485744e-05, "loss": 0.3473, "step": 1284 }, { "epoch": 0.02, "learning_rate": 4.958400455460021e-05, "loss": 0.2763, "step": 1286 }, { "epoch": 0.02, "learning_rate": 4.958335759434298e-05, "loss": 0.4347, "step": 1288 }, { "epoch": 0.03, "learning_rate": 4.958271063408575e-05, "loss": 0.306, "step": 1290 }, { "epoch": 0.03, "learning_rate": 4.958206367382852e-05, "loss": 0.4645, "step": 1292 }, { "epoch": 0.03, "learning_rate": 4.958141671357129e-05, "loss": 0.3226, "step": 1294 }, { "epoch": 0.03, "learning_rate": 4.958076975331406e-05, "loss": 0.3053, "step": 1296 }, { "epoch": 0.03, "learning_rate": 4.9580122793056826e-05, "loss": 0.3069, "step": 1298 }, { "epoch": 0.03, "learning_rate": 4.9579475832799596e-05, "loss": 0.3449, "step": 1300 }, { "epoch": 0.03, "learning_rate": 4.957882887254236e-05, "loss": 0.327, "step": 1302 }, { "epoch": 0.03, "learning_rate": 4.9578181912285134e-05, "loss": 0.3048, "step": 1304 }, { "epoch": 0.03, "learning_rate": 4.9577534952027896e-05, "loss": 0.3216, "step": 1306 }, { "epoch": 0.03, "learning_rate": 4.9576887991770665e-05, "loss": 0.2934, "step": 1308 }, { "epoch": 0.03, "learning_rate": 4.9576241031513434e-05, "loss": 0.2799, "step": 1310 }, { "epoch": 0.03, "learning_rate": 4.95755940712562e-05, "loss": 0.2813, "step": 1312 }, { "epoch": 0.03, "learning_rate": 4.957494711099898e-05, "loss": 0.305, "step": 1314 }, { "epoch": 0.03, "learning_rate": 4.957430015074174e-05, "loss": 0.3381, "step": 1316 }, { "epoch": 0.03, "learning_rate": 4.957365319048451e-05, "loss": 0.3605, "step": 1318 }, { "epoch": 0.03, "learning_rate": 4.957300623022728e-05, "loss": 0.3346, "step": 1320 }, { "epoch": 0.03, "learning_rate": 4.957235926997005e-05, "loss": 0.4216, "step": 1322 }, { "epoch": 0.03, "learning_rate": 4.957171230971282e-05, "loss": 0.2774, "step": 1324 }, { "epoch": 0.03, "learning_rate": 4.957106534945559e-05, "loss": 0.234, "step": 1326 }, { "epoch": 0.03, "learning_rate": 4.9570418389198356e-05, "loss": 0.3489, "step": 1328 }, { "epoch": 0.03, "learning_rate": 4.9569771428941125e-05, "loss": 0.3578, "step": 1330 }, { "epoch": 0.03, "learning_rate": 4.9569124468683894e-05, "loss": 0.2453, "step": 1332 }, { "epoch": 0.03, "learning_rate": 4.9568477508426656e-05, "loss": 0.2868, "step": 1334 }, { "epoch": 0.03, "learning_rate": 4.956783054816943e-05, "loss": 0.287, "step": 1336 }, { "epoch": 0.03, "learning_rate": 4.9567183587912195e-05, "loss": 0.2874, "step": 1338 }, { "epoch": 0.03, "learning_rate": 4.9566536627654964e-05, "loss": 0.297, "step": 1340 }, { "epoch": 0.03, "learning_rate": 4.956588966739773e-05, "loss": 0.3631, "step": 1342 }, { "epoch": 0.03, "learning_rate": 4.95652427071405e-05, "loss": 0.2831, "step": 1344 }, { "epoch": 0.03, "learning_rate": 4.956459574688327e-05, "loss": 0.3552, "step": 1346 }, { "epoch": 0.03, "learning_rate": 4.956394878662604e-05, "loss": 0.307, "step": 1348 }, { "epoch": 0.03, "learning_rate": 4.956330182636881e-05, "loss": 0.3126, "step": 1350 }, { "epoch": 0.03, "learning_rate": 4.956265486611158e-05, "loss": 0.261, "step": 1352 }, { "epoch": 0.03, "learning_rate": 4.956200790585435e-05, "loss": 0.3308, "step": 1354 }, { "epoch": 0.03, "learning_rate": 4.956136094559711e-05, "loss": 0.2826, "step": 1356 }, { "epoch": 0.03, "learning_rate": 4.9560713985339885e-05, "loss": 0.2499, "step": 1358 }, { "epoch": 0.03, "learning_rate": 4.9560067025082655e-05, "loss": 0.3853, "step": 1360 }, { "epoch": 0.03, "learning_rate": 4.955942006482542e-05, "loss": 0.2762, "step": 1362 }, { "epoch": 0.03, "learning_rate": 4.955877310456819e-05, "loss": 0.3046, "step": 1364 }, { "epoch": 0.03, "learning_rate": 4.9558126144310955e-05, "loss": 0.3306, "step": 1366 }, { "epoch": 0.03, "learning_rate": 4.9557479184053724e-05, "loss": 0.2839, "step": 1368 }, { "epoch": 0.03, "learning_rate": 4.955683222379649e-05, "loss": 0.2689, "step": 1370 }, { "epoch": 0.03, "learning_rate": 4.955618526353926e-05, "loss": 0.2933, "step": 1372 }, { "epoch": 0.03, "learning_rate": 4.955553830328203e-05, "loss": 0.3021, "step": 1374 }, { "epoch": 0.03, "learning_rate": 4.95548913430248e-05, "loss": 0.2561, "step": 1376 }, { "epoch": 0.03, "learning_rate": 4.955424438276757e-05, "loss": 0.2754, "step": 1378 }, { "epoch": 0.03, "learning_rate": 4.955359742251034e-05, "loss": 0.3431, "step": 1380 }, { "epoch": 0.03, "learning_rate": 4.955295046225311e-05, "loss": 0.3669, "step": 1382 }, { "epoch": 0.03, "learning_rate": 4.955230350199587e-05, "loss": 0.3397, "step": 1384 }, { "epoch": 0.03, "learning_rate": 4.9551656541738646e-05, "loss": 0.4115, "step": 1386 }, { "epoch": 0.03, "learning_rate": 4.955100958148141e-05, "loss": 0.303, "step": 1388 }, { "epoch": 0.03, "learning_rate": 4.955036262122418e-05, "loss": 0.2901, "step": 1390 }, { "epoch": 0.03, "learning_rate": 4.954971566096695e-05, "loss": 0.3026, "step": 1392 }, { "epoch": 0.03, "learning_rate": 4.9549068700709715e-05, "loss": 0.3418, "step": 1394 }, { "epoch": 0.03, "learning_rate": 4.954842174045249e-05, "loss": 0.2511, "step": 1396 }, { "epoch": 0.03, "learning_rate": 4.9547774780195254e-05, "loss": 0.3689, "step": 1398 }, { "epoch": 0.03, "learning_rate": 4.954712781993802e-05, "loss": 0.2696, "step": 1400 }, { "epoch": 0.03, "learning_rate": 4.954648085968079e-05, "loss": 0.5088, "step": 1402 }, { "epoch": 0.03, "learning_rate": 4.954583389942356e-05, "loss": 0.3129, "step": 1404 }, { "epoch": 0.03, "learning_rate": 4.954518693916633e-05, "loss": 0.3163, "step": 1406 }, { "epoch": 0.03, "learning_rate": 4.95445399789091e-05, "loss": 0.3324, "step": 1408 }, { "epoch": 0.03, "learning_rate": 4.954389301865187e-05, "loss": 0.2908, "step": 1410 }, { "epoch": 0.03, "learning_rate": 4.954324605839464e-05, "loss": 0.3106, "step": 1412 }, { "epoch": 0.03, "learning_rate": 4.9542599098137406e-05, "loss": 0.2309, "step": 1414 }, { "epoch": 0.03, "learning_rate": 4.954195213788017e-05, "loss": 0.2532, "step": 1416 }, { "epoch": 0.03, "learning_rate": 4.9541305177622944e-05, "loss": 0.3102, "step": 1418 }, { "epoch": 0.03, "learning_rate": 4.954065821736571e-05, "loss": 0.2713, "step": 1420 }, { "epoch": 0.03, "learning_rate": 4.9540011257108476e-05, "loss": 0.3086, "step": 1422 }, { "epoch": 0.03, "learning_rate": 4.953936429685125e-05, "loss": 0.231, "step": 1424 }, { "epoch": 0.03, "learning_rate": 4.9538717336594014e-05, "loss": 0.2995, "step": 1426 }, { "epoch": 0.03, "learning_rate": 4.953807037633678e-05, "loss": 0.3251, "step": 1428 }, { "epoch": 0.03, "learning_rate": 4.953742341607955e-05, "loss": 0.4228, "step": 1430 }, { "epoch": 0.03, "learning_rate": 4.953677645582232e-05, "loss": 0.3055, "step": 1432 }, { "epoch": 0.03, "learning_rate": 4.953612949556509e-05, "loss": 0.2231, "step": 1434 }, { "epoch": 0.03, "learning_rate": 4.953548253530786e-05, "loss": 0.3029, "step": 1436 }, { "epoch": 0.03, "learning_rate": 4.953483557505063e-05, "loss": 0.2771, "step": 1438 }, { "epoch": 0.03, "learning_rate": 4.95341886147934e-05, "loss": 0.2261, "step": 1440 }, { "epoch": 0.03, "learning_rate": 4.953354165453617e-05, "loss": 0.3375, "step": 1442 }, { "epoch": 0.03, "learning_rate": 4.953289469427893e-05, "loss": 0.2497, "step": 1444 }, { "epoch": 0.03, "learning_rate": 4.9532247734021705e-05, "loss": 0.3172, "step": 1446 }, { "epoch": 0.03, "learning_rate": 4.953160077376447e-05, "loss": 0.3232, "step": 1448 }, { "epoch": 0.03, "learning_rate": 4.9530953813507236e-05, "loss": 0.2585, "step": 1450 }, { "epoch": 0.03, "learning_rate": 4.9530306853250005e-05, "loss": 0.2887, "step": 1452 }, { "epoch": 0.03, "learning_rate": 4.9529659892992774e-05, "loss": 0.3684, "step": 1454 }, { "epoch": 0.03, "learning_rate": 4.9529012932735543e-05, "loss": 0.2703, "step": 1456 }, { "epoch": 0.03, "learning_rate": 4.952836597247831e-05, "loss": 0.2935, "step": 1458 }, { "epoch": 0.03, "learning_rate": 4.952771901222108e-05, "loss": 0.2493, "step": 1460 }, { "epoch": 0.03, "learning_rate": 4.952707205196385e-05, "loss": 0.3481, "step": 1462 }, { "epoch": 0.03, "learning_rate": 4.952642509170662e-05, "loss": 0.2368, "step": 1464 }, { "epoch": 0.03, "learning_rate": 4.952577813144938e-05, "loss": 0.3077, "step": 1466 }, { "epoch": 0.03, "learning_rate": 4.952513117119216e-05, "loss": 0.3025, "step": 1468 }, { "epoch": 0.03, "learning_rate": 4.952448421093493e-05, "loss": 0.2435, "step": 1470 }, { "epoch": 0.03, "learning_rate": 4.9523837250677696e-05, "loss": 0.3098, "step": 1472 }, { "epoch": 0.03, "learning_rate": 4.9523190290420465e-05, "loss": 0.3048, "step": 1474 }, { "epoch": 0.03, "learning_rate": 4.952254333016323e-05, "loss": 0.3147, "step": 1476 }, { "epoch": 0.03, "learning_rate": 4.9521896369906e-05, "loss": 0.2727, "step": 1478 }, { "epoch": 0.03, "learning_rate": 4.9521249409648766e-05, "loss": 0.2082, "step": 1480 }, { "epoch": 0.03, "learning_rate": 4.9520602449391535e-05, "loss": 0.2465, "step": 1482 }, { "epoch": 0.03, "learning_rate": 4.9519955489134304e-05, "loss": 0.2791, "step": 1484 }, { "epoch": 0.03, "learning_rate": 4.951930852887707e-05, "loss": 0.5383, "step": 1486 }, { "epoch": 0.03, "learning_rate": 4.951866156861984e-05, "loss": 0.3433, "step": 1488 }, { "epoch": 0.03, "learning_rate": 4.951801460836261e-05, "loss": 0.2657, "step": 1490 }, { "epoch": 0.03, "learning_rate": 4.951736764810538e-05, "loss": 0.2465, "step": 1492 }, { "epoch": 0.03, "learning_rate": 4.951672068784815e-05, "loss": 0.3251, "step": 1494 }, { "epoch": 0.03, "learning_rate": 4.951607372759092e-05, "loss": 0.412, "step": 1496 }, { "epoch": 0.03, "learning_rate": 4.951542676733368e-05, "loss": 0.2513, "step": 1498 }, { "epoch": 0.03, "learning_rate": 4.9514779807076457e-05, "loss": 0.2158, "step": 1500 }, { "epoch": 0.03, "learning_rate": 4.9514132846819226e-05, "loss": 0.2828, "step": 1502 }, { "epoch": 0.03, "learning_rate": 4.951348588656199e-05, "loss": 0.2331, "step": 1504 }, { "epoch": 0.03, "learning_rate": 4.9512838926304764e-05, "loss": 0.3539, "step": 1506 }, { "epoch": 0.03, "learning_rate": 4.9512191966047526e-05, "loss": 0.2368, "step": 1508 }, { "epoch": 0.03, "learning_rate": 4.9511545005790295e-05, "loss": 0.2717, "step": 1510 }, { "epoch": 0.03, "learning_rate": 4.9510898045533064e-05, "loss": 0.268, "step": 1512 }, { "epoch": 0.03, "learning_rate": 4.951025108527583e-05, "loss": 0.3302, "step": 1514 }, { "epoch": 0.03, "learning_rate": 4.95096041250186e-05, "loss": 0.2986, "step": 1516 }, { "epoch": 0.03, "learning_rate": 4.950895716476137e-05, "loss": 0.2495, "step": 1518 }, { "epoch": 0.03, "learning_rate": 4.950831020450414e-05, "loss": 0.2511, "step": 1520 }, { "epoch": 0.03, "learning_rate": 4.950766324424691e-05, "loss": 0.362, "step": 1522 }, { "epoch": 0.03, "learning_rate": 4.950701628398968e-05, "loss": 0.2375, "step": 1524 }, { "epoch": 0.03, "learning_rate": 4.950636932373244e-05, "loss": 0.4186, "step": 1526 }, { "epoch": 0.03, "learning_rate": 4.950572236347522e-05, "loss": 0.2743, "step": 1528 }, { "epoch": 0.03, "learning_rate": 4.950507540321798e-05, "loss": 0.2981, "step": 1530 }, { "epoch": 0.03, "learning_rate": 4.950442844296075e-05, "loss": 0.2913, "step": 1532 }, { "epoch": 0.03, "learning_rate": 4.950378148270352e-05, "loss": 0.2338, "step": 1534 }, { "epoch": 0.03, "learning_rate": 4.9503134522446286e-05, "loss": 0.279, "step": 1536 }, { "epoch": 0.03, "learning_rate": 4.950248756218906e-05, "loss": 0.1863, "step": 1538 }, { "epoch": 0.03, "learning_rate": 4.9501840601931825e-05, "loss": 0.2826, "step": 1540 }, { "epoch": 0.03, "learning_rate": 4.9501193641674594e-05, "loss": 0.4699, "step": 1542 }, { "epoch": 0.03, "learning_rate": 4.950054668141736e-05, "loss": 0.2829, "step": 1544 }, { "epoch": 0.03, "learning_rate": 4.949989972116013e-05, "loss": 0.1928, "step": 1546 }, { "epoch": 0.03, "learning_rate": 4.94992527609029e-05, "loss": 0.215, "step": 1548 }, { "epoch": 0.03, "learning_rate": 4.949860580064567e-05, "loss": 0.1605, "step": 1550 }, { "epoch": 0.03, "learning_rate": 4.949795884038844e-05, "loss": 0.2976, "step": 1552 }, { "epoch": 0.03, "learning_rate": 4.949731188013121e-05, "loss": 0.1834, "step": 1554 }, { "epoch": 0.03, "learning_rate": 4.949666491987398e-05, "loss": 0.2261, "step": 1556 }, { "epoch": 0.03, "learning_rate": 4.949601795961674e-05, "loss": 0.2154, "step": 1558 }, { "epoch": 0.03, "learning_rate": 4.9495370999359516e-05, "loss": 0.3219, "step": 1560 }, { "epoch": 0.03, "learning_rate": 4.949472403910228e-05, "loss": 0.1917, "step": 1562 }, { "epoch": 0.03, "learning_rate": 4.949407707884505e-05, "loss": 0.2487, "step": 1564 }, { "epoch": 0.03, "learning_rate": 4.9493430118587816e-05, "loss": 0.2919, "step": 1566 }, { "epoch": 0.03, "learning_rate": 4.9492783158330585e-05, "loss": 0.2775, "step": 1568 }, { "epoch": 0.03, "learning_rate": 4.9492136198073354e-05, "loss": 0.2158, "step": 1570 }, { "epoch": 0.03, "learning_rate": 4.949148923781612e-05, "loss": 0.2187, "step": 1572 }, { "epoch": 0.03, "learning_rate": 4.949084227755889e-05, "loss": 0.2559, "step": 1574 }, { "epoch": 0.03, "learning_rate": 4.949019531730166e-05, "loss": 0.2027, "step": 1576 }, { "epoch": 0.03, "learning_rate": 4.948954835704443e-05, "loss": 0.3036, "step": 1578 }, { "epoch": 0.03, "learning_rate": 4.948890139678719e-05, "loss": 0.2634, "step": 1580 }, { "epoch": 0.03, "learning_rate": 4.948825443652997e-05, "loss": 0.2143, "step": 1582 }, { "epoch": 0.03, "learning_rate": 4.948760747627274e-05, "loss": 0.2391, "step": 1584 }, { "epoch": 0.03, "learning_rate": 4.94869605160155e-05, "loss": 0.2258, "step": 1586 }, { "epoch": 0.03, "learning_rate": 4.9486313555758276e-05, "loss": 0.2933, "step": 1588 }, { "epoch": 0.03, "learning_rate": 4.948566659550104e-05, "loss": 0.2223, "step": 1590 }, { "epoch": 0.03, "learning_rate": 4.948501963524381e-05, "loss": 0.22, "step": 1592 }, { "epoch": 0.03, "learning_rate": 4.9484372674986576e-05, "loss": 0.2126, "step": 1594 }, { "epoch": 0.03, "learning_rate": 4.9483725714729345e-05, "loss": 0.2796, "step": 1596 }, { "epoch": 0.03, "learning_rate": 4.9483078754472115e-05, "loss": 0.2381, "step": 1598 }, { "epoch": 0.03, "learning_rate": 4.9482431794214884e-05, "loss": 0.2439, "step": 1600 }, { "epoch": 0.03, "learning_rate": 4.948178483395765e-05, "loss": 0.2096, "step": 1602 }, { "epoch": 0.03, "learning_rate": 4.948113787370042e-05, "loss": 0.2754, "step": 1604 }, { "epoch": 0.03, "learning_rate": 4.948049091344319e-05, "loss": 0.2469, "step": 1606 }, { "epoch": 0.03, "learning_rate": 4.947984395318595e-05, "loss": 0.221, "step": 1608 }, { "epoch": 0.03, "learning_rate": 4.947919699292873e-05, "loss": 0.2134, "step": 1610 }, { "epoch": 0.03, "learning_rate": 4.947855003267149e-05, "loss": 0.236, "step": 1612 }, { "epoch": 0.03, "learning_rate": 4.947790307241427e-05, "loss": 0.2648, "step": 1614 }, { "epoch": 0.03, "learning_rate": 4.9477256112157036e-05, "loss": 0.3356, "step": 1616 }, { "epoch": 0.03, "learning_rate": 4.94766091518998e-05, "loss": 0.1764, "step": 1618 }, { "epoch": 0.03, "learning_rate": 4.9475962191642574e-05, "loss": 0.2061, "step": 1620 }, { "epoch": 0.03, "learning_rate": 4.947531523138534e-05, "loss": 0.3385, "step": 1622 }, { "epoch": 0.03, "learning_rate": 4.9474668271128106e-05, "loss": 0.2522, "step": 1624 }, { "epoch": 0.03, "learning_rate": 4.9474021310870875e-05, "loss": 0.1772, "step": 1626 }, { "epoch": 0.03, "learning_rate": 4.9473374350613644e-05, "loss": 0.2485, "step": 1628 }, { "epoch": 0.03, "learning_rate": 4.947272739035641e-05, "loss": 0.223, "step": 1630 }, { "epoch": 0.03, "learning_rate": 4.947208043009918e-05, "loss": 0.2943, "step": 1632 }, { "epoch": 0.03, "learning_rate": 4.947143346984195e-05, "loss": 0.199, "step": 1634 }, { "epoch": 0.03, "learning_rate": 4.947078650958472e-05, "loss": 0.2578, "step": 1636 }, { "epoch": 0.03, "learning_rate": 4.947013954932749e-05, "loss": 0.2284, "step": 1638 }, { "epoch": 0.03, "learning_rate": 4.946949258907025e-05, "loss": 0.2595, "step": 1640 }, { "epoch": 0.03, "learning_rate": 4.946884562881303e-05, "loss": 0.3172, "step": 1642 }, { "epoch": 0.03, "learning_rate": 4.946819866855579e-05, "loss": 0.2223, "step": 1644 }, { "epoch": 0.03, "learning_rate": 4.946755170829856e-05, "loss": 0.2123, "step": 1646 }, { "epoch": 0.03, "learning_rate": 4.9466904748041335e-05, "loss": 0.2919, "step": 1648 }, { "epoch": 0.03, "learning_rate": 4.94662577877841e-05, "loss": 0.2172, "step": 1650 }, { "epoch": 0.03, "learning_rate": 4.9465610827526866e-05, "loss": 0.2389, "step": 1652 }, { "epoch": 0.03, "learning_rate": 4.9464963867269635e-05, "loss": 0.2848, "step": 1654 }, { "epoch": 0.03, "learning_rate": 4.9464316907012404e-05, "loss": 0.2284, "step": 1656 }, { "epoch": 0.03, "learning_rate": 4.9463669946755174e-05, "loss": 0.2357, "step": 1658 }, { "epoch": 0.03, "learning_rate": 4.946302298649794e-05, "loss": 0.2071, "step": 1660 }, { "epoch": 0.03, "learning_rate": 4.946237602624071e-05, "loss": 0.172, "step": 1662 }, { "epoch": 0.03, "learning_rate": 4.946172906598348e-05, "loss": 0.2001, "step": 1664 }, { "epoch": 0.03, "learning_rate": 4.946108210572625e-05, "loss": 0.1818, "step": 1666 }, { "epoch": 0.03, "learning_rate": 4.946043514546901e-05, "loss": 0.2027, "step": 1668 }, { "epoch": 0.03, "learning_rate": 4.945978818521179e-05, "loss": 0.2032, "step": 1670 }, { "epoch": 0.03, "learning_rate": 4.945914122495455e-05, "loss": 0.1976, "step": 1672 }, { "epoch": 0.03, "learning_rate": 4.945849426469732e-05, "loss": 0.2388, "step": 1674 }, { "epoch": 0.03, "learning_rate": 4.945784730444009e-05, "loss": 0.2881, "step": 1676 }, { "epoch": 0.03, "learning_rate": 4.945720034418286e-05, "loss": 0.2246, "step": 1678 }, { "epoch": 0.03, "learning_rate": 4.9456553383925633e-05, "loss": 0.2281, "step": 1680 }, { "epoch": 0.03, "learning_rate": 4.9455906423668396e-05, "loss": 0.2513, "step": 1682 }, { "epoch": 0.03, "learning_rate": 4.9455259463411165e-05, "loss": 0.2688, "step": 1684 }, { "epoch": 0.03, "learning_rate": 4.9454612503153934e-05, "loss": 0.1837, "step": 1686 }, { "epoch": 0.03, "learning_rate": 4.94539655428967e-05, "loss": 0.3009, "step": 1688 }, { "epoch": 0.03, "learning_rate": 4.9453318582639465e-05, "loss": 0.2029, "step": 1690 }, { "epoch": 0.03, "learning_rate": 4.945267162238224e-05, "loss": 0.2447, "step": 1692 }, { "epoch": 0.03, "learning_rate": 4.945202466212501e-05, "loss": 0.2187, "step": 1694 }, { "epoch": 0.03, "learning_rate": 4.945137770186778e-05, "loss": 0.1736, "step": 1696 }, { "epoch": 0.03, "learning_rate": 4.945073074161055e-05, "loss": 0.2921, "step": 1698 }, { "epoch": 0.03, "learning_rate": 4.945008378135331e-05, "loss": 0.2163, "step": 1700 }, { "epoch": 0.03, "learning_rate": 4.944943682109609e-05, "loss": 0.201, "step": 1702 }, { "epoch": 0.03, "learning_rate": 4.944878986083885e-05, "loss": 0.2542, "step": 1704 }, { "epoch": 0.03, "learning_rate": 4.944814290058162e-05, "loss": 0.2097, "step": 1706 }, { "epoch": 0.03, "learning_rate": 4.944749594032439e-05, "loss": 0.3255, "step": 1708 }, { "epoch": 0.03, "learning_rate": 4.9446848980067156e-05, "loss": 0.2379, "step": 1710 }, { "epoch": 0.03, "learning_rate": 4.9446202019809925e-05, "loss": 0.1853, "step": 1712 }, { "epoch": 0.03, "learning_rate": 4.9445555059552694e-05, "loss": 0.2496, "step": 1714 }, { "epoch": 0.03, "learning_rate": 4.9444908099295463e-05, "loss": 0.2206, "step": 1716 }, { "epoch": 0.03, "learning_rate": 4.944426113903823e-05, "loss": 0.1888, "step": 1718 }, { "epoch": 0.03, "learning_rate": 4.9443614178781e-05, "loss": 0.2476, "step": 1720 }, { "epoch": 0.03, "learning_rate": 4.9442967218523764e-05, "loss": 0.1964, "step": 1722 }, { "epoch": 0.03, "learning_rate": 4.944232025826654e-05, "loss": 0.2083, "step": 1724 }, { "epoch": 0.03, "learning_rate": 4.944167329800931e-05, "loss": 0.1778, "step": 1726 }, { "epoch": 0.03, "learning_rate": 4.944102633775207e-05, "loss": 0.202, "step": 1728 }, { "epoch": 0.03, "learning_rate": 4.944037937749485e-05, "loss": 0.202, "step": 1730 }, { "epoch": 0.03, "learning_rate": 4.943973241723761e-05, "loss": 0.1684, "step": 1732 }, { "epoch": 0.03, "learning_rate": 4.943908545698038e-05, "loss": 0.1816, "step": 1734 }, { "epoch": 0.03, "learning_rate": 4.943843849672315e-05, "loss": 0.1961, "step": 1736 }, { "epoch": 0.03, "learning_rate": 4.9437791536465917e-05, "loss": 0.1888, "step": 1738 }, { "epoch": 0.03, "learning_rate": 4.9437144576208686e-05, "loss": 0.1832, "step": 1740 }, { "epoch": 0.03, "learning_rate": 4.9436497615951455e-05, "loss": 0.187, "step": 1742 }, { "epoch": 0.03, "learning_rate": 4.9435850655694224e-05, "loss": 0.2329, "step": 1744 }, { "epoch": 0.03, "learning_rate": 4.943520369543699e-05, "loss": 0.1453, "step": 1746 }, { "epoch": 0.03, "learning_rate": 4.943455673517976e-05, "loss": 0.2033, "step": 1748 }, { "epoch": 0.03, "learning_rate": 4.9433909774922524e-05, "loss": 0.3305, "step": 1750 }, { "epoch": 0.03, "learning_rate": 4.94332628146653e-05, "loss": 0.2495, "step": 1752 }, { "epoch": 0.03, "learning_rate": 4.943261585440806e-05, "loss": 0.2624, "step": 1754 }, { "epoch": 0.03, "learning_rate": 4.943196889415084e-05, "loss": 0.1985, "step": 1756 }, { "epoch": 0.03, "learning_rate": 4.94313219338936e-05, "loss": 0.1417, "step": 1758 }, { "epoch": 0.03, "learning_rate": 4.943067497363637e-05, "loss": 0.151, "step": 1760 }, { "epoch": 0.03, "learning_rate": 4.9430028013379146e-05, "loss": 0.1497, "step": 1762 }, { "epoch": 0.03, "learning_rate": 4.942938105312191e-05, "loss": 0.1625, "step": 1764 }, { "epoch": 0.03, "learning_rate": 4.942873409286468e-05, "loss": 0.2974, "step": 1766 }, { "epoch": 0.03, "learning_rate": 4.9428087132607446e-05, "loss": 0.2219, "step": 1768 }, { "epoch": 0.03, "learning_rate": 4.9427440172350215e-05, "loss": 0.2101, "step": 1770 }, { "epoch": 0.03, "learning_rate": 4.9426793212092984e-05, "loss": 0.3064, "step": 1772 }, { "epoch": 0.03, "learning_rate": 4.942614625183575e-05, "loss": 0.2657, "step": 1774 }, { "epoch": 0.03, "learning_rate": 4.942549929157852e-05, "loss": 0.1491, "step": 1776 }, { "epoch": 0.03, "learning_rate": 4.942485233132129e-05, "loss": 0.1938, "step": 1778 }, { "epoch": 0.03, "learning_rate": 4.942420537106406e-05, "loss": 0.2489, "step": 1780 }, { "epoch": 0.03, "learning_rate": 4.942355841080682e-05, "loss": 0.1957, "step": 1782 }, { "epoch": 0.03, "learning_rate": 4.94229114505496e-05, "loss": 0.219, "step": 1784 }, { "epoch": 0.03, "learning_rate": 4.942226449029236e-05, "loss": 0.266, "step": 1786 }, { "epoch": 0.03, "learning_rate": 4.942161753003513e-05, "loss": 0.2453, "step": 1788 }, { "epoch": 0.03, "learning_rate": 4.94209705697779e-05, "loss": 0.2512, "step": 1790 }, { "epoch": 0.03, "learning_rate": 4.942032360952067e-05, "loss": 0.1792, "step": 1792 }, { "epoch": 0.03, "learning_rate": 4.941967664926344e-05, "loss": 0.3063, "step": 1794 }, { "epoch": 0.03, "learning_rate": 4.9419029689006206e-05, "loss": 0.1493, "step": 1796 }, { "epoch": 0.03, "learning_rate": 4.9418382728748976e-05, "loss": 0.1609, "step": 1798 }, { "epoch": 0.03, "learning_rate": 4.9417735768491745e-05, "loss": 0.1824, "step": 1800 }, { "epoch": 0.03, "learning_rate": 4.9417088808234514e-05, "loss": 0.1914, "step": 1802 }, { "epoch": 0.04, "learning_rate": 4.941644184797728e-05, "loss": 0.2427, "step": 1804 }, { "epoch": 0.04, "learning_rate": 4.941579488772005e-05, "loss": 0.2165, "step": 1806 }, { "epoch": 0.04, "learning_rate": 4.941514792746282e-05, "loss": 0.1255, "step": 1808 }, { "epoch": 0.04, "learning_rate": 4.941450096720558e-05, "loss": 0.1412, "step": 1810 }, { "epoch": 0.04, "learning_rate": 4.941385400694836e-05, "loss": 0.2056, "step": 1812 }, { "epoch": 0.04, "learning_rate": 4.941320704669112e-05, "loss": 0.165, "step": 1814 }, { "epoch": 0.04, "learning_rate": 4.941256008643389e-05, "loss": 0.2249, "step": 1816 }, { "epoch": 0.04, "learning_rate": 4.941191312617666e-05, "loss": 0.19, "step": 1818 }, { "epoch": 0.04, "learning_rate": 4.941126616591943e-05, "loss": 0.23, "step": 1820 }, { "epoch": 0.04, "learning_rate": 4.94106192056622e-05, "loss": 0.1697, "step": 1822 }, { "epoch": 0.04, "learning_rate": 4.940997224540497e-05, "loss": 0.2592, "step": 1824 }, { "epoch": 0.04, "learning_rate": 4.9409325285147736e-05, "loss": 0.1888, "step": 1826 }, { "epoch": 0.04, "learning_rate": 4.9408678324890505e-05, "loss": 0.245, "step": 1828 }, { "epoch": 0.04, "learning_rate": 4.9408031364633274e-05, "loss": 0.1943, "step": 1830 }, { "epoch": 0.04, "learning_rate": 4.9407384404376036e-05, "loss": 0.1748, "step": 1832 }, { "epoch": 0.04, "learning_rate": 4.940673744411881e-05, "loss": 0.1498, "step": 1834 }, { "epoch": 0.04, "learning_rate": 4.9406090483861575e-05, "loss": 0.1517, "step": 1836 }, { "epoch": 0.04, "learning_rate": 4.940544352360435e-05, "loss": 0.2103, "step": 1838 }, { "epoch": 0.04, "learning_rate": 4.940479656334712e-05, "loss": 0.2343, "step": 1840 }, { "epoch": 0.04, "learning_rate": 4.940414960308988e-05, "loss": 0.2017, "step": 1842 }, { "epoch": 0.04, "learning_rate": 4.940350264283266e-05, "loss": 0.166, "step": 1844 }, { "epoch": 0.04, "learning_rate": 4.940285568257542e-05, "loss": 0.1885, "step": 1846 }, { "epoch": 0.04, "learning_rate": 4.940220872231819e-05, "loss": 0.1638, "step": 1848 }, { "epoch": 0.04, "learning_rate": 4.940156176206096e-05, "loss": 0.1901, "step": 1850 }, { "epoch": 0.04, "learning_rate": 4.940091480180373e-05, "loss": 0.2485, "step": 1852 }, { "epoch": 0.04, "learning_rate": 4.9400267841546496e-05, "loss": 0.1399, "step": 1854 }, { "epoch": 0.04, "learning_rate": 4.9399620881289265e-05, "loss": 0.236, "step": 1856 }, { "epoch": 0.04, "learning_rate": 4.9398973921032035e-05, "loss": 0.1871, "step": 1858 }, { "epoch": 0.04, "learning_rate": 4.9398326960774804e-05, "loss": 0.1773, "step": 1860 }, { "epoch": 0.04, "learning_rate": 4.939768000051757e-05, "loss": 0.1355, "step": 1862 }, { "epoch": 0.04, "learning_rate": 4.9397033040260335e-05, "loss": 0.24, "step": 1864 }, { "epoch": 0.04, "learning_rate": 4.939638608000311e-05, "loss": 0.1971, "step": 1866 }, { "epoch": 0.04, "learning_rate": 4.939573911974587e-05, "loss": 0.1847, "step": 1868 }, { "epoch": 0.04, "learning_rate": 4.939509215948864e-05, "loss": 0.1791, "step": 1870 }, { "epoch": 0.04, "learning_rate": 4.939444519923142e-05, "loss": 0.1768, "step": 1872 }, { "epoch": 0.04, "learning_rate": 4.939379823897418e-05, "loss": 0.158, "step": 1874 }, { "epoch": 0.04, "learning_rate": 4.939315127871695e-05, "loss": 0.1376, "step": 1876 }, { "epoch": 0.04, "learning_rate": 4.939250431845972e-05, "loss": 0.1456, "step": 1878 }, { "epoch": 0.04, "learning_rate": 4.939185735820249e-05, "loss": 0.3203, "step": 1880 }, { "epoch": 0.04, "learning_rate": 4.939121039794526e-05, "loss": 0.2034, "step": 1882 }, { "epoch": 0.04, "learning_rate": 4.9390563437688026e-05, "loss": 0.2047, "step": 1884 }, { "epoch": 0.04, "learning_rate": 4.9389916477430795e-05, "loss": 0.1917, "step": 1886 }, { "epoch": 0.04, "learning_rate": 4.9389269517173564e-05, "loss": 0.1504, "step": 1888 }, { "epoch": 0.04, "learning_rate": 4.938862255691633e-05, "loss": 0.1426, "step": 1890 }, { "epoch": 0.04, "learning_rate": 4.9387975596659095e-05, "loss": 0.207, "step": 1892 }, { "epoch": 0.04, "learning_rate": 4.938732863640187e-05, "loss": 0.2127, "step": 1894 }, { "epoch": 0.04, "learning_rate": 4.9386681676144634e-05, "loss": 0.1764, "step": 1896 }, { "epoch": 0.04, "learning_rate": 4.938603471588741e-05, "loss": 0.2229, "step": 1898 }, { "epoch": 0.04, "learning_rate": 4.938538775563017e-05, "loss": 0.1659, "step": 1900 }, { "epoch": 0.04, "learning_rate": 4.938474079537294e-05, "loss": 0.1361, "step": 1902 }, { "epoch": 0.04, "learning_rate": 4.938409383511572e-05, "loss": 0.2117, "step": 1904 }, { "epoch": 0.04, "learning_rate": 4.938344687485848e-05, "loss": 0.1518, "step": 1906 }, { "epoch": 0.04, "learning_rate": 4.938279991460125e-05, "loss": 0.1356, "step": 1908 }, { "epoch": 0.04, "learning_rate": 4.938215295434402e-05, "loss": 0.1636, "step": 1910 }, { "epoch": 0.04, "learning_rate": 4.9381505994086786e-05, "loss": 0.1617, "step": 1912 }, { "epoch": 0.04, "learning_rate": 4.938085903382955e-05, "loss": 0.1349, "step": 1914 }, { "epoch": 0.04, "learning_rate": 4.9380212073572324e-05, "loss": 0.1466, "step": 1916 }, { "epoch": 0.04, "learning_rate": 4.9379565113315094e-05, "loss": 0.3438, "step": 1918 }, { "epoch": 0.04, "learning_rate": 4.937891815305786e-05, "loss": 0.1803, "step": 1920 }, { "epoch": 0.04, "learning_rate": 4.937827119280063e-05, "loss": 0.1928, "step": 1922 }, { "epoch": 0.04, "learning_rate": 4.9377624232543394e-05, "loss": 0.179, "step": 1924 }, { "epoch": 0.04, "learning_rate": 4.937697727228617e-05, "loss": 0.1632, "step": 1926 }, { "epoch": 0.04, "learning_rate": 4.937633031202893e-05, "loss": 0.199, "step": 1928 }, { "epoch": 0.04, "learning_rate": 4.93756833517717e-05, "loss": 0.2222, "step": 1930 }, { "epoch": 0.04, "learning_rate": 4.937503639151447e-05, "loss": 0.1711, "step": 1932 }, { "epoch": 0.04, "learning_rate": 4.937438943125724e-05, "loss": 0.2349, "step": 1934 }, { "epoch": 0.04, "learning_rate": 4.937374247100001e-05, "loss": 0.2287, "step": 1936 }, { "epoch": 0.04, "learning_rate": 4.937309551074278e-05, "loss": 0.3889, "step": 1938 }, { "epoch": 0.04, "learning_rate": 4.937244855048555e-05, "loss": 0.2307, "step": 1940 }, { "epoch": 0.04, "learning_rate": 4.9371801590228316e-05, "loss": 0.1373, "step": 1942 }, { "epoch": 0.04, "learning_rate": 4.9371154629971085e-05, "loss": 0.1692, "step": 1944 }, { "epoch": 0.04, "learning_rate": 4.937050766971385e-05, "loss": 0.2243, "step": 1946 }, { "epoch": 0.04, "learning_rate": 4.936986070945662e-05, "loss": 0.1521, "step": 1948 }, { "epoch": 0.04, "learning_rate": 4.936921374919939e-05, "loss": 0.1904, "step": 1950 }, { "epoch": 0.04, "learning_rate": 4.9368566788942154e-05, "loss": 0.1365, "step": 1952 }, { "epoch": 0.04, "learning_rate": 4.936791982868493e-05, "loss": 0.1204, "step": 1954 }, { "epoch": 0.04, "learning_rate": 4.936727286842769e-05, "loss": 0.1929, "step": 1956 }, { "epoch": 0.04, "learning_rate": 4.936662590817046e-05, "loss": 0.1965, "step": 1958 }, { "epoch": 0.04, "learning_rate": 4.936597894791323e-05, "loss": 0.1757, "step": 1960 }, { "epoch": 0.04, "learning_rate": 4.9365331987656e-05, "loss": 0.1787, "step": 1962 }, { "epoch": 0.04, "learning_rate": 4.936468502739877e-05, "loss": 0.181, "step": 1964 }, { "epoch": 0.04, "learning_rate": 4.936403806714154e-05, "loss": 0.1667, "step": 1966 }, { "epoch": 0.04, "learning_rate": 4.936339110688431e-05, "loss": 0.2098, "step": 1968 }, { "epoch": 0.04, "learning_rate": 4.9362744146627076e-05, "loss": 0.1452, "step": 1970 }, { "epoch": 0.04, "learning_rate": 4.9362097186369845e-05, "loss": 0.2304, "step": 1972 }, { "epoch": 0.04, "learning_rate": 4.936145022611261e-05, "loss": 0.1727, "step": 1974 }, { "epoch": 0.04, "learning_rate": 4.9360803265855383e-05, "loss": 0.2007, "step": 1976 }, { "epoch": 0.04, "learning_rate": 4.9360156305598146e-05, "loss": 0.2018, "step": 1978 }, { "epoch": 0.04, "learning_rate": 4.935950934534092e-05, "loss": 0.1917, "step": 1980 }, { "epoch": 0.04, "learning_rate": 4.935886238508369e-05, "loss": 0.1819, "step": 1982 }, { "epoch": 0.04, "learning_rate": 4.935821542482645e-05, "loss": 0.1454, "step": 1984 }, { "epoch": 0.04, "learning_rate": 4.935756846456923e-05, "loss": 0.1628, "step": 1986 }, { "epoch": 0.04, "learning_rate": 4.935692150431199e-05, "loss": 0.0887, "step": 1988 }, { "epoch": 0.04, "learning_rate": 4.935627454405476e-05, "loss": 0.1688, "step": 1990 }, { "epoch": 0.04, "learning_rate": 4.935562758379753e-05, "loss": 0.1903, "step": 1992 }, { "epoch": 0.04, "learning_rate": 4.93549806235403e-05, "loss": 0.1149, "step": 1994 }, { "epoch": 0.04, "learning_rate": 4.935433366328307e-05, "loss": 0.2258, "step": 1996 }, { "epoch": 0.04, "learning_rate": 4.9353686703025837e-05, "loss": 0.1922, "step": 1998 }, { "epoch": 0.04, "learning_rate": 4.9353039742768606e-05, "loss": 0.2299, "step": 2000 }, { "epoch": 0.04, "learning_rate": 4.9352392782511375e-05, "loss": 0.1577, "step": 2002 }, { "epoch": 0.04, "learning_rate": 4.9351745822254144e-05, "loss": 0.1143, "step": 2004 }, { "epoch": 0.04, "learning_rate": 4.9351098861996906e-05, "loss": 0.1408, "step": 2006 }, { "epoch": 0.04, "learning_rate": 4.935045190173968e-05, "loss": 0.2427, "step": 2008 }, { "epoch": 0.04, "learning_rate": 4.9349804941482444e-05, "loss": 0.1603, "step": 2010 }, { "epoch": 0.04, "learning_rate": 4.934915798122521e-05, "loss": 0.153, "step": 2012 }, { "epoch": 0.04, "learning_rate": 4.934851102096798e-05, "loss": 0.1983, "step": 2014 }, { "epoch": 0.04, "learning_rate": 4.934786406071075e-05, "loss": 0.1757, "step": 2016 }, { "epoch": 0.04, "learning_rate": 4.934721710045352e-05, "loss": 0.2022, "step": 2018 }, { "epoch": 0.04, "learning_rate": 4.934657014019629e-05, "loss": 0.1817, "step": 2020 }, { "epoch": 0.04, "learning_rate": 4.934592317993906e-05, "loss": 0.2191, "step": 2022 }, { "epoch": 0.04, "learning_rate": 4.934527621968183e-05, "loss": 0.1258, "step": 2024 }, { "epoch": 0.04, "learning_rate": 4.93446292594246e-05, "loss": 0.1923, "step": 2026 }, { "epoch": 0.04, "learning_rate": 4.9343982299167366e-05, "loss": 0.1962, "step": 2028 }, { "epoch": 0.04, "learning_rate": 4.9343335338910135e-05, "loss": 0.1884, "step": 2030 }, { "epoch": 0.04, "learning_rate": 4.9342688378652904e-05, "loss": 0.2049, "step": 2032 }, { "epoch": 0.04, "learning_rate": 4.9342041418395667e-05, "loss": 0.1848, "step": 2034 }, { "epoch": 0.04, "learning_rate": 4.934139445813844e-05, "loss": 0.1169, "step": 2036 }, { "epoch": 0.04, "learning_rate": 4.9340747497881205e-05, "loss": 0.1709, "step": 2038 }, { "epoch": 0.04, "learning_rate": 4.934010053762398e-05, "loss": 0.1191, "step": 2040 }, { "epoch": 0.04, "learning_rate": 4.933945357736674e-05, "loss": 0.1506, "step": 2042 }, { "epoch": 0.04, "learning_rate": 4.933880661710951e-05, "loss": 0.1573, "step": 2044 }, { "epoch": 0.04, "learning_rate": 4.933815965685228e-05, "loss": 0.1831, "step": 2046 }, { "epoch": 0.04, "learning_rate": 4.933751269659505e-05, "loss": 0.1491, "step": 2048 }, { "epoch": 0.04, "learning_rate": 4.933686573633782e-05, "loss": 0.179, "step": 2050 }, { "epoch": 0.04, "learning_rate": 4.933621877608059e-05, "loss": 0.1459, "step": 2052 }, { "epoch": 0.04, "learning_rate": 4.933557181582336e-05, "loss": 0.2174, "step": 2054 }, { "epoch": 0.04, "learning_rate": 4.933492485556612e-05, "loss": 0.213, "step": 2056 }, { "epoch": 0.04, "learning_rate": 4.9334277895308896e-05, "loss": 0.1167, "step": 2058 }, { "epoch": 0.04, "learning_rate": 4.933363093505166e-05, "loss": 0.1871, "step": 2060 }, { "epoch": 0.04, "learning_rate": 4.9332983974794434e-05, "loss": 0.128, "step": 2062 }, { "epoch": 0.04, "learning_rate": 4.93323370145372e-05, "loss": 0.2054, "step": 2064 }, { "epoch": 0.04, "learning_rate": 4.9331690054279965e-05, "loss": 0.1233, "step": 2066 }, { "epoch": 0.04, "learning_rate": 4.933104309402274e-05, "loss": 0.1439, "step": 2068 }, { "epoch": 0.04, "learning_rate": 4.93303961337655e-05, "loss": 0.1945, "step": 2070 }, { "epoch": 0.04, "learning_rate": 4.932974917350827e-05, "loss": 0.1177, "step": 2072 }, { "epoch": 0.04, "learning_rate": 4.932910221325104e-05, "loss": 0.1804, "step": 2074 }, { "epoch": 0.04, "learning_rate": 4.932845525299381e-05, "loss": 0.1124, "step": 2076 }, { "epoch": 0.04, "learning_rate": 4.932780829273658e-05, "loss": 0.1909, "step": 2078 }, { "epoch": 0.04, "learning_rate": 4.932716133247935e-05, "loss": 0.1509, "step": 2080 }, { "epoch": 0.04, "learning_rate": 4.932651437222212e-05, "loss": 0.1447, "step": 2082 }, { "epoch": 0.04, "learning_rate": 4.932586741196489e-05, "loss": 0.0718, "step": 2084 }, { "epoch": 0.04, "learning_rate": 4.9325220451707656e-05, "loss": 0.1658, "step": 2086 }, { "epoch": 0.04, "learning_rate": 4.932457349145042e-05, "loss": 0.3124, "step": 2088 }, { "epoch": 0.04, "learning_rate": 4.9323926531193194e-05, "loss": 0.1725, "step": 2090 }, { "epoch": 0.04, "learning_rate": 4.9323279570935956e-05, "loss": 0.1677, "step": 2092 }, { "epoch": 0.04, "learning_rate": 4.9322632610678726e-05, "loss": 0.1566, "step": 2094 }, { "epoch": 0.04, "learning_rate": 4.93219856504215e-05, "loss": 0.1037, "step": 2096 }, { "epoch": 0.04, "learning_rate": 4.9321338690164264e-05, "loss": 0.1737, "step": 2098 }, { "epoch": 0.04, "learning_rate": 4.932069172990703e-05, "loss": 0.2565, "step": 2100 }, { "epoch": 0.04, "learning_rate": 4.93200447696498e-05, "loss": 0.1766, "step": 2102 }, { "epoch": 0.04, "learning_rate": 4.931939780939257e-05, "loss": 0.1509, "step": 2104 }, { "epoch": 0.04, "learning_rate": 4.931875084913534e-05, "loss": 0.1127, "step": 2106 }, { "epoch": 0.04, "learning_rate": 4.931810388887811e-05, "loss": 0.1486, "step": 2108 }, { "epoch": 0.04, "learning_rate": 4.931745692862088e-05, "loss": 0.2038, "step": 2110 }, { "epoch": 0.04, "learning_rate": 4.931680996836365e-05, "loss": 0.1345, "step": 2112 }, { "epoch": 0.04, "learning_rate": 4.9316163008106416e-05, "loss": 0.1825, "step": 2114 }, { "epoch": 0.04, "learning_rate": 4.931551604784918e-05, "loss": 0.1048, "step": 2116 }, { "epoch": 0.04, "learning_rate": 4.9314869087591955e-05, "loss": 0.1581, "step": 2118 }, { "epoch": 0.04, "learning_rate": 4.931422212733472e-05, "loss": 0.1385, "step": 2120 }, { "epoch": 0.04, "learning_rate": 4.931357516707749e-05, "loss": 0.248, "step": 2122 }, { "epoch": 0.04, "learning_rate": 4.9312928206820255e-05, "loss": 0.1732, "step": 2124 }, { "epoch": 0.04, "learning_rate": 4.9312281246563024e-05, "loss": 0.1668, "step": 2126 }, { "epoch": 0.04, "learning_rate": 4.93116342863058e-05, "loss": 0.1367, "step": 2128 }, { "epoch": 0.04, "learning_rate": 4.931098732604856e-05, "loss": 0.1662, "step": 2130 }, { "epoch": 0.04, "learning_rate": 4.931034036579133e-05, "loss": 0.1424, "step": 2132 }, { "epoch": 0.04, "learning_rate": 4.93096934055341e-05, "loss": 0.1305, "step": 2134 }, { "epoch": 0.04, "learning_rate": 4.930904644527687e-05, "loss": 0.1517, "step": 2136 }, { "epoch": 0.04, "learning_rate": 4.930839948501963e-05, "loss": 0.1211, "step": 2138 }, { "epoch": 0.04, "learning_rate": 4.930775252476241e-05, "loss": 0.1385, "step": 2140 }, { "epoch": 0.04, "learning_rate": 4.930710556450518e-05, "loss": 0.1167, "step": 2142 }, { "epoch": 0.04, "learning_rate": 4.9306458604247946e-05, "loss": 0.1773, "step": 2144 }, { "epoch": 0.04, "learning_rate": 4.9305811643990715e-05, "loss": 0.2197, "step": 2146 }, { "epoch": 0.04, "learning_rate": 4.930516468373348e-05, "loss": 0.125, "step": 2148 }, { "epoch": 0.04, "learning_rate": 4.930451772347625e-05, "loss": 0.1565, "step": 2150 }, { "epoch": 0.04, "learning_rate": 4.9303870763219015e-05, "loss": 0.2155, "step": 2152 }, { "epoch": 0.04, "learning_rate": 4.9303223802961784e-05, "loss": 0.1444, "step": 2154 }, { "epoch": 0.04, "learning_rate": 4.9302576842704554e-05, "loss": 0.1876, "step": 2156 }, { "epoch": 0.04, "learning_rate": 4.930192988244732e-05, "loss": 0.1695, "step": 2158 }, { "epoch": 0.04, "learning_rate": 4.930128292219009e-05, "loss": 0.1688, "step": 2160 }, { "epoch": 0.04, "learning_rate": 4.930063596193286e-05, "loss": 0.2212, "step": 2162 }, { "epoch": 0.04, "learning_rate": 4.929998900167563e-05, "loss": 0.1773, "step": 2164 }, { "epoch": 0.04, "learning_rate": 4.92993420414184e-05, "loss": 0.1125, "step": 2166 }, { "epoch": 0.04, "learning_rate": 4.929869508116117e-05, "loss": 0.2585, "step": 2168 }, { "epoch": 0.04, "learning_rate": 4.929804812090393e-05, "loss": 0.1611, "step": 2170 }, { "epoch": 0.04, "learning_rate": 4.9297401160646706e-05, "loss": 0.2703, "step": 2172 }, { "epoch": 0.04, "learning_rate": 4.9296754200389475e-05, "loss": 0.15, "step": 2174 }, { "epoch": 0.04, "learning_rate": 4.929610724013224e-05, "loss": 0.1498, "step": 2176 }, { "epoch": 0.04, "learning_rate": 4.9295460279875014e-05, "loss": 0.1868, "step": 2178 }, { "epoch": 0.04, "learning_rate": 4.9294813319617776e-05, "loss": 0.1462, "step": 2180 }, { "epoch": 0.04, "learning_rate": 4.929416635936055e-05, "loss": 0.1705, "step": 2182 }, { "epoch": 0.04, "learning_rate": 4.9293519399103314e-05, "loss": 0.1293, "step": 2184 }, { "epoch": 0.04, "learning_rate": 4.929287243884608e-05, "loss": 0.1465, "step": 2186 }, { "epoch": 0.04, "learning_rate": 4.929222547858885e-05, "loss": 0.1922, "step": 2188 }, { "epoch": 0.04, "learning_rate": 4.929157851833162e-05, "loss": 0.1433, "step": 2190 }, { "epoch": 0.04, "learning_rate": 4.929093155807439e-05, "loss": 0.1339, "step": 2192 }, { "epoch": 0.04, "learning_rate": 4.929028459781716e-05, "loss": 0.1926, "step": 2194 }, { "epoch": 0.04, "learning_rate": 4.928963763755993e-05, "loss": 0.1518, "step": 2196 }, { "epoch": 0.04, "learning_rate": 4.928899067730269e-05, "loss": 0.1394, "step": 2198 }, { "epoch": 0.04, "learning_rate": 4.928834371704547e-05, "loss": 0.1337, "step": 2200 }, { "epoch": 0.04, "learning_rate": 4.928769675678823e-05, "loss": 0.1458, "step": 2202 }, { "epoch": 0.04, "learning_rate": 4.9287049796531005e-05, "loss": 0.136, "step": 2204 }, { "epoch": 0.04, "learning_rate": 4.9286402836273774e-05, "loss": 0.1165, "step": 2206 }, { "epoch": 0.04, "learning_rate": 4.9285755876016536e-05, "loss": 0.1794, "step": 2208 }, { "epoch": 0.04, "learning_rate": 4.928510891575931e-05, "loss": 0.2793, "step": 2210 }, { "epoch": 0.04, "learning_rate": 4.9284461955502074e-05, "loss": 0.1854, "step": 2212 }, { "epoch": 0.04, "learning_rate": 4.9283814995244843e-05, "loss": 0.1966, "step": 2214 }, { "epoch": 0.04, "learning_rate": 4.928316803498761e-05, "loss": 0.1438, "step": 2216 }, { "epoch": 0.04, "learning_rate": 4.928252107473038e-05, "loss": 0.235, "step": 2218 }, { "epoch": 0.04, "learning_rate": 4.928187411447315e-05, "loss": 0.2435, "step": 2220 }, { "epoch": 0.04, "learning_rate": 4.928122715421592e-05, "loss": 0.1485, "step": 2222 }, { "epoch": 0.04, "learning_rate": 4.928058019395869e-05, "loss": 0.1008, "step": 2224 }, { "epoch": 0.04, "learning_rate": 4.927993323370146e-05, "loss": 0.1761, "step": 2226 }, { "epoch": 0.04, "learning_rate": 4.927928627344423e-05, "loss": 0.1445, "step": 2228 }, { "epoch": 0.04, "learning_rate": 4.927863931318699e-05, "loss": 0.1491, "step": 2230 }, { "epoch": 0.04, "learning_rate": 4.9277992352929765e-05, "loss": 0.2536, "step": 2232 }, { "epoch": 0.04, "learning_rate": 4.927734539267253e-05, "loss": 0.1245, "step": 2234 }, { "epoch": 0.04, "learning_rate": 4.92766984324153e-05, "loss": 0.1753, "step": 2236 }, { "epoch": 0.04, "learning_rate": 4.9276051472158066e-05, "loss": 0.2305, "step": 2238 }, { "epoch": 0.04, "learning_rate": 4.9275404511900835e-05, "loss": 0.1991, "step": 2240 }, { "epoch": 0.04, "learning_rate": 4.9274757551643604e-05, "loss": 0.1088, "step": 2242 }, { "epoch": 0.04, "learning_rate": 4.927411059138637e-05, "loss": 0.1115, "step": 2244 }, { "epoch": 0.04, "learning_rate": 4.927346363112914e-05, "loss": 0.1497, "step": 2246 }, { "epoch": 0.04, "learning_rate": 4.927281667087191e-05, "loss": 0.2015, "step": 2248 }, { "epoch": 0.04, "learning_rate": 4.927216971061468e-05, "loss": 0.1259, "step": 2250 }, { "epoch": 0.04, "learning_rate": 4.927152275035745e-05, "loss": 0.1335, "step": 2252 }, { "epoch": 0.04, "learning_rate": 4.927087579010022e-05, "loss": 0.0952, "step": 2254 }, { "epoch": 0.04, "learning_rate": 4.927022882984299e-05, "loss": 0.1412, "step": 2256 }, { "epoch": 0.04, "learning_rate": 4.926958186958575e-05, "loss": 0.1896, "step": 2258 }, { "epoch": 0.04, "learning_rate": 4.9268934909328526e-05, "loss": 0.128, "step": 2260 }, { "epoch": 0.04, "learning_rate": 4.926828794907129e-05, "loss": 0.1499, "step": 2262 }, { "epoch": 0.04, "learning_rate": 4.9267640988814064e-05, "loss": 0.1459, "step": 2264 }, { "epoch": 0.04, "learning_rate": 4.9266994028556826e-05, "loss": 0.138, "step": 2266 }, { "epoch": 0.04, "learning_rate": 4.9266347068299595e-05, "loss": 0.17, "step": 2268 }, { "epoch": 0.04, "learning_rate": 4.9265700108042364e-05, "loss": 0.1526, "step": 2270 }, { "epoch": 0.04, "learning_rate": 4.926505314778513e-05, "loss": 0.0725, "step": 2272 }, { "epoch": 0.04, "learning_rate": 4.92644061875279e-05, "loss": 0.1465, "step": 2274 }, { "epoch": 0.04, "learning_rate": 4.926375922727067e-05, "loss": 0.2039, "step": 2276 }, { "epoch": 0.04, "learning_rate": 4.926311226701344e-05, "loss": 0.1164, "step": 2278 }, { "epoch": 0.04, "learning_rate": 4.92624653067562e-05, "loss": 0.2613, "step": 2280 }, { "epoch": 0.04, "learning_rate": 4.926181834649898e-05, "loss": 0.186, "step": 2282 }, { "epoch": 0.04, "learning_rate": 4.926117138624175e-05, "loss": 0.1168, "step": 2284 }, { "epoch": 0.04, "learning_rate": 4.926052442598452e-05, "loss": 0.1734, "step": 2286 }, { "epoch": 0.04, "learning_rate": 4.9259877465727286e-05, "loss": 0.1599, "step": 2288 }, { "epoch": 0.04, "learning_rate": 4.925923050547005e-05, "loss": 0.1703, "step": 2290 }, { "epoch": 0.04, "learning_rate": 4.9258583545212824e-05, "loss": 0.1578, "step": 2292 }, { "epoch": 0.04, "learning_rate": 4.9257936584955587e-05, "loss": 0.1506, "step": 2294 }, { "epoch": 0.04, "learning_rate": 4.9257289624698356e-05, "loss": 0.4167, "step": 2296 }, { "epoch": 0.04, "learning_rate": 4.9256642664441125e-05, "loss": 0.1473, "step": 2298 }, { "epoch": 0.04, "learning_rate": 4.9255995704183894e-05, "loss": 0.1231, "step": 2300 }, { "epoch": 0.04, "learning_rate": 4.925534874392666e-05, "loss": 0.0825, "step": 2302 }, { "epoch": 0.04, "learning_rate": 4.925470178366943e-05, "loss": 0.151, "step": 2304 }, { "epoch": 0.04, "learning_rate": 4.92540548234122e-05, "loss": 0.121, "step": 2306 }, { "epoch": 0.04, "learning_rate": 4.925340786315497e-05, "loss": 0.14, "step": 2308 }, { "epoch": 0.04, "learning_rate": 4.925276090289774e-05, "loss": 0.1066, "step": 2310 }, { "epoch": 0.04, "learning_rate": 4.92521139426405e-05, "loss": 0.144, "step": 2312 }, { "epoch": 0.04, "learning_rate": 4.925146698238328e-05, "loss": 0.1769, "step": 2314 }, { "epoch": 0.04, "learning_rate": 4.925082002212604e-05, "loss": 0.1337, "step": 2316 }, { "epoch": 0.04, "learning_rate": 4.925017306186881e-05, "loss": 0.1659, "step": 2318 }, { "epoch": 0.05, "learning_rate": 4.9249526101611585e-05, "loss": 0.0742, "step": 2320 }, { "epoch": 0.05, "learning_rate": 4.924887914135435e-05, "loss": 0.1329, "step": 2322 }, { "epoch": 0.05, "learning_rate": 4.924823218109712e-05, "loss": 0.13, "step": 2324 }, { "epoch": 0.05, "learning_rate": 4.9247585220839885e-05, "loss": 0.1818, "step": 2326 }, { "epoch": 0.05, "learning_rate": 4.9246938260582654e-05, "loss": 0.1031, "step": 2328 }, { "epoch": 0.05, "learning_rate": 4.924629130032542e-05, "loss": 0.1372, "step": 2330 }, { "epoch": 0.05, "learning_rate": 4.924564434006819e-05, "loss": 0.1988, "step": 2332 }, { "epoch": 0.05, "learning_rate": 4.924499737981096e-05, "loss": 0.1225, "step": 2334 }, { "epoch": 0.05, "learning_rate": 4.924435041955373e-05, "loss": 0.2251, "step": 2336 }, { "epoch": 0.05, "learning_rate": 4.92437034592965e-05, "loss": 0.0881, "step": 2338 }, { "epoch": 0.05, "learning_rate": 4.924305649903926e-05, "loss": 0.1109, "step": 2340 }, { "epoch": 0.05, "learning_rate": 4.924240953878204e-05, "loss": 0.2086, "step": 2342 }, { "epoch": 0.05, "learning_rate": 4.92417625785248e-05, "loss": 0.101, "step": 2344 }, { "epoch": 0.05, "learning_rate": 4.9241115618267576e-05, "loss": 0.2403, "step": 2346 }, { "epoch": 0.05, "learning_rate": 4.924046865801034e-05, "loss": 0.1294, "step": 2348 }, { "epoch": 0.05, "learning_rate": 4.923982169775311e-05, "loss": 0.0821, "step": 2350 }, { "epoch": 0.05, "learning_rate": 4.923917473749588e-05, "loss": 0.1513, "step": 2352 }, { "epoch": 0.05, "learning_rate": 4.9238527777238645e-05, "loss": 0.1872, "step": 2354 }, { "epoch": 0.05, "learning_rate": 4.9237880816981415e-05, "loss": 0.0996, "step": 2356 }, { "epoch": 0.05, "learning_rate": 4.9237233856724184e-05, "loss": 0.1063, "step": 2358 }, { "epoch": 0.05, "learning_rate": 4.923658689646695e-05, "loss": 0.2262, "step": 2360 }, { "epoch": 0.05, "learning_rate": 4.923593993620972e-05, "loss": 0.0564, "step": 2362 }, { "epoch": 0.05, "learning_rate": 4.923529297595249e-05, "loss": 0.1684, "step": 2364 }, { "epoch": 0.05, "learning_rate": 4.923464601569526e-05, "loss": 0.1078, "step": 2366 }, { "epoch": 0.05, "learning_rate": 4.923399905543803e-05, "loss": 0.1103, "step": 2368 }, { "epoch": 0.05, "learning_rate": 4.92333520951808e-05, "loss": 0.1655, "step": 2370 }, { "epoch": 0.05, "learning_rate": 4.923270513492356e-05, "loss": 0.1706, "step": 2372 }, { "epoch": 0.05, "learning_rate": 4.9232058174666336e-05, "loss": 0.1958, "step": 2374 }, { "epoch": 0.05, "learning_rate": 4.92314112144091e-05, "loss": 0.1414, "step": 2376 }, { "epoch": 0.05, "learning_rate": 4.923076425415187e-05, "loss": 0.0886, "step": 2378 }, { "epoch": 0.05, "learning_rate": 4.923011729389464e-05, "loss": 0.1527, "step": 2380 }, { "epoch": 0.05, "learning_rate": 4.9229470333637406e-05, "loss": 0.1141, "step": 2382 }, { "epoch": 0.05, "learning_rate": 4.9228823373380175e-05, "loss": 0.0612, "step": 2384 }, { "epoch": 0.05, "learning_rate": 4.9228176413122944e-05, "loss": 0.077, "step": 2386 }, { "epoch": 0.05, "learning_rate": 4.922752945286571e-05, "loss": 0.0888, "step": 2388 }, { "epoch": 0.05, "learning_rate": 4.922688249260848e-05, "loss": 0.1409, "step": 2390 }, { "epoch": 0.05, "learning_rate": 4.922623553235125e-05, "loss": 0.123, "step": 2392 }, { "epoch": 0.05, "learning_rate": 4.9225588572094014e-05, "loss": 0.1599, "step": 2394 }, { "epoch": 0.05, "learning_rate": 4.922494161183679e-05, "loss": 0.1176, "step": 2396 }, { "epoch": 0.05, "learning_rate": 4.922429465157956e-05, "loss": 0.1953, "step": 2398 }, { "epoch": 0.05, "learning_rate": 4.922364769132232e-05, "loss": 0.1517, "step": 2400 }, { "epoch": 0.05, "learning_rate": 4.92230007310651e-05, "loss": 0.1377, "step": 2402 }, { "epoch": 0.05, "learning_rate": 4.922235377080786e-05, "loss": 0.1265, "step": 2404 }, { "epoch": 0.05, "learning_rate": 4.9221706810550635e-05, "loss": 0.0807, "step": 2406 }, { "epoch": 0.05, "learning_rate": 4.92210598502934e-05, "loss": 0.1145, "step": 2408 }, { "epoch": 0.05, "learning_rate": 4.9220412890036166e-05, "loss": 0.136, "step": 2410 }, { "epoch": 0.05, "learning_rate": 4.9219765929778935e-05, "loss": 0.1512, "step": 2412 }, { "epoch": 0.05, "learning_rate": 4.9219118969521704e-05, "loss": 0.2028, "step": 2414 }, { "epoch": 0.05, "learning_rate": 4.9218472009264474e-05, "loss": 0.1157, "step": 2416 }, { "epoch": 0.05, "learning_rate": 4.921782504900724e-05, "loss": 0.1095, "step": 2418 }, { "epoch": 0.05, "learning_rate": 4.921717808875001e-05, "loss": 0.1726, "step": 2420 }, { "epoch": 0.05, "learning_rate": 4.9216531128492774e-05, "loss": 0.131, "step": 2422 }, { "epoch": 0.05, "learning_rate": 4.921588416823555e-05, "loss": 0.2118, "step": 2424 }, { "epoch": 0.05, "learning_rate": 4.921523720797831e-05, "loss": 0.1233, "step": 2426 }, { "epoch": 0.05, "learning_rate": 4.921459024772109e-05, "loss": 0.1515, "step": 2428 }, { "epoch": 0.05, "learning_rate": 4.921394328746386e-05, "loss": 0.1344, "step": 2430 }, { "epoch": 0.05, "learning_rate": 4.921329632720662e-05, "loss": 0.0976, "step": 2432 }, { "epoch": 0.05, "learning_rate": 4.9212649366949395e-05, "loss": 0.2172, "step": 2434 }, { "epoch": 0.05, "learning_rate": 4.921200240669216e-05, "loss": 0.0984, "step": 2436 }, { "epoch": 0.05, "learning_rate": 4.921135544643493e-05, "loss": 0.1176, "step": 2438 }, { "epoch": 0.05, "learning_rate": 4.9210708486177696e-05, "loss": 0.086, "step": 2440 }, { "epoch": 0.05, "learning_rate": 4.9210061525920465e-05, "loss": 0.1792, "step": 2442 }, { "epoch": 0.05, "learning_rate": 4.9209414565663234e-05, "loss": 0.131, "step": 2444 }, { "epoch": 0.05, "learning_rate": 4.9208767605406e-05, "loss": 0.1636, "step": 2446 }, { "epoch": 0.05, "learning_rate": 4.920812064514877e-05, "loss": 0.1724, "step": 2448 }, { "epoch": 0.05, "learning_rate": 4.920747368489154e-05, "loss": 0.1586, "step": 2450 }, { "epoch": 0.05, "learning_rate": 4.920682672463431e-05, "loss": 0.0907, "step": 2452 }, { "epoch": 0.05, "learning_rate": 4.920617976437707e-05, "loss": 0.0972, "step": 2454 }, { "epoch": 0.05, "learning_rate": 4.920553280411985e-05, "loss": 0.0855, "step": 2456 }, { "epoch": 0.05, "learning_rate": 4.920488584386261e-05, "loss": 0.144, "step": 2458 }, { "epoch": 0.05, "learning_rate": 4.920423888360538e-05, "loss": 0.119, "step": 2460 }, { "epoch": 0.05, "learning_rate": 4.9203591923348156e-05, "loss": 0.1358, "step": 2462 }, { "epoch": 0.05, "learning_rate": 4.920294496309092e-05, "loss": 0.1245, "step": 2464 }, { "epoch": 0.05, "learning_rate": 4.920229800283369e-05, "loss": 0.1088, "step": 2466 }, { "epoch": 0.05, "learning_rate": 4.9201651042576456e-05, "loss": 0.0958, "step": 2468 }, { "epoch": 0.05, "learning_rate": 4.9201004082319225e-05, "loss": 0.1096, "step": 2470 }, { "epoch": 0.05, "learning_rate": 4.9200357122061994e-05, "loss": 0.2403, "step": 2472 }, { "epoch": 0.05, "learning_rate": 4.9199710161804763e-05, "loss": 0.1035, "step": 2474 }, { "epoch": 0.05, "learning_rate": 4.919906320154753e-05, "loss": 0.2499, "step": 2476 }, { "epoch": 0.05, "learning_rate": 4.91984162412903e-05, "loss": 0.194, "step": 2478 }, { "epoch": 0.05, "learning_rate": 4.919776928103307e-05, "loss": 0.0768, "step": 2480 }, { "epoch": 0.05, "learning_rate": 4.919712232077583e-05, "loss": 0.1434, "step": 2482 }, { "epoch": 0.05, "learning_rate": 4.919647536051861e-05, "loss": 0.2004, "step": 2484 }, { "epoch": 0.05, "learning_rate": 4.919582840026137e-05, "loss": 0.1203, "step": 2486 }, { "epoch": 0.05, "learning_rate": 4.919518144000415e-05, "loss": 0.1174, "step": 2488 }, { "epoch": 0.05, "learning_rate": 4.919453447974691e-05, "loss": 0.3597, "step": 2490 }, { "epoch": 0.05, "learning_rate": 4.919388751948968e-05, "loss": 0.068, "step": 2492 }, { "epoch": 0.05, "learning_rate": 4.919324055923245e-05, "loss": 0.1237, "step": 2494 }, { "epoch": 0.05, "learning_rate": 4.9192593598975217e-05, "loss": 0.1355, "step": 2496 }, { "epoch": 0.05, "learning_rate": 4.9191946638717986e-05, "loss": 0.1042, "step": 2498 }, { "epoch": 0.05, "learning_rate": 4.9191299678460755e-05, "loss": 0.13, "step": 2500 }, { "epoch": 0.05, "learning_rate": 4.9190652718203524e-05, "loss": 0.075, "step": 2502 }, { "epoch": 0.05, "learning_rate": 4.9190005757946286e-05, "loss": 0.1854, "step": 2504 }, { "epoch": 0.05, "learning_rate": 4.918935879768906e-05, "loss": 0.1308, "step": 2506 }, { "epoch": 0.05, "learning_rate": 4.918871183743183e-05, "loss": 0.1252, "step": 2508 }, { "epoch": 0.05, "learning_rate": 4.91880648771746e-05, "loss": 0.1725, "step": 2510 }, { "epoch": 0.05, "learning_rate": 4.918741791691737e-05, "loss": 0.11, "step": 2512 }, { "epoch": 0.05, "learning_rate": 4.918677095666013e-05, "loss": 0.1002, "step": 2514 }, { "epoch": 0.05, "learning_rate": 4.918612399640291e-05, "loss": 0.128, "step": 2516 }, { "epoch": 0.05, "learning_rate": 4.918547703614567e-05, "loss": 0.0709, "step": 2518 }, { "epoch": 0.05, "learning_rate": 4.918483007588844e-05, "loss": 0.1139, "step": 2520 }, { "epoch": 0.05, "learning_rate": 4.918418311563121e-05, "loss": 0.1176, "step": 2522 }, { "epoch": 0.05, "learning_rate": 4.918353615537398e-05, "loss": 0.1456, "step": 2524 }, { "epoch": 0.05, "learning_rate": 4.9182889195116746e-05, "loss": 0.0979, "step": 2526 }, { "epoch": 0.05, "learning_rate": 4.9182242234859515e-05, "loss": 0.1114, "step": 2528 }, { "epoch": 0.05, "learning_rate": 4.9181595274602284e-05, "loss": 0.1039, "step": 2530 }, { "epoch": 0.05, "learning_rate": 4.918094831434505e-05, "loss": 0.1253, "step": 2532 }, { "epoch": 0.05, "learning_rate": 4.918030135408782e-05, "loss": 0.086, "step": 2534 }, { "epoch": 0.05, "learning_rate": 4.9179654393830585e-05, "loss": 0.0728, "step": 2536 }, { "epoch": 0.05, "learning_rate": 4.917900743357336e-05, "loss": 0.0794, "step": 2538 }, { "epoch": 0.05, "learning_rate": 4.917836047331612e-05, "loss": 0.1635, "step": 2540 }, { "epoch": 0.05, "learning_rate": 4.917771351305889e-05, "loss": 0.0933, "step": 2542 }, { "epoch": 0.05, "learning_rate": 4.917706655280167e-05, "loss": 0.1048, "step": 2544 }, { "epoch": 0.05, "learning_rate": 4.917641959254443e-05, "loss": 0.1505, "step": 2546 }, { "epoch": 0.05, "learning_rate": 4.9175772632287206e-05, "loss": 0.1362, "step": 2548 }, { "epoch": 0.05, "learning_rate": 4.917512567202997e-05, "loss": 0.1012, "step": 2550 }, { "epoch": 0.05, "learning_rate": 4.917447871177274e-05, "loss": 0.0969, "step": 2552 }, { "epoch": 0.05, "learning_rate": 4.9173831751515506e-05, "loss": 0.1173, "step": 2554 }, { "epoch": 0.05, "learning_rate": 4.9173184791258276e-05, "loss": 0.1551, "step": 2556 }, { "epoch": 0.05, "learning_rate": 4.9172537831001045e-05, "loss": 0.1081, "step": 2558 }, { "epoch": 0.05, "learning_rate": 4.9171890870743814e-05, "loss": 0.184, "step": 2560 }, { "epoch": 0.05, "learning_rate": 4.917124391048658e-05, "loss": 0.1258, "step": 2562 }, { "epoch": 0.05, "learning_rate": 4.9170596950229345e-05, "loss": 0.2296, "step": 2564 }, { "epoch": 0.05, "learning_rate": 4.916994998997212e-05, "loss": 0.1749, "step": 2566 }, { "epoch": 0.05, "learning_rate": 4.916930302971488e-05, "loss": 0.1033, "step": 2568 }, { "epoch": 0.05, "learning_rate": 4.916865606945766e-05, "loss": 0.1183, "step": 2570 }, { "epoch": 0.05, "learning_rate": 4.916800910920042e-05, "loss": 0.1101, "step": 2572 }, { "epoch": 0.05, "learning_rate": 4.916736214894319e-05, "loss": 0.1237, "step": 2574 }, { "epoch": 0.05, "learning_rate": 4.9166715188685966e-05, "loss": 0.0865, "step": 2576 }, { "epoch": 0.05, "learning_rate": 4.916606822842873e-05, "loss": 0.1148, "step": 2578 }, { "epoch": 0.05, "learning_rate": 4.91654212681715e-05, "loss": 0.1094, "step": 2580 }, { "epoch": 0.05, "learning_rate": 4.916477430791427e-05, "loss": 0.1367, "step": 2582 }, { "epoch": 0.05, "learning_rate": 4.9164127347657036e-05, "loss": 0.1257, "step": 2584 }, { "epoch": 0.05, "learning_rate": 4.9163480387399805e-05, "loss": 0.0994, "step": 2586 }, { "epoch": 0.05, "learning_rate": 4.9162833427142574e-05, "loss": 0.0874, "step": 2588 }, { "epoch": 0.05, "learning_rate": 4.916218646688534e-05, "loss": 0.1553, "step": 2590 }, { "epoch": 0.05, "learning_rate": 4.916153950662811e-05, "loss": 0.2047, "step": 2592 }, { "epoch": 0.05, "learning_rate": 4.916089254637088e-05, "loss": 0.1083, "step": 2594 }, { "epoch": 0.05, "learning_rate": 4.9160245586113644e-05, "loss": 0.1314, "step": 2596 }, { "epoch": 0.05, "learning_rate": 4.915959862585642e-05, "loss": 0.1137, "step": 2598 }, { "epoch": 0.05, "learning_rate": 4.915895166559918e-05, "loss": 0.0811, "step": 2600 }, { "epoch": 0.05, "learning_rate": 4.915830470534195e-05, "loss": 0.1439, "step": 2602 }, { "epoch": 0.05, "learning_rate": 4.915765774508472e-05, "loss": 0.1489, "step": 2604 }, { "epoch": 0.05, "learning_rate": 4.915701078482749e-05, "loss": 0.1468, "step": 2606 }, { "epoch": 0.05, "learning_rate": 4.915636382457026e-05, "loss": 0.1518, "step": 2608 }, { "epoch": 0.05, "learning_rate": 4.915571686431303e-05, "loss": 0.1017, "step": 2610 }, { "epoch": 0.05, "learning_rate": 4.9155069904055796e-05, "loss": 0.1157, "step": 2612 }, { "epoch": 0.05, "learning_rate": 4.9154422943798565e-05, "loss": 0.1705, "step": 2614 }, { "epoch": 0.05, "learning_rate": 4.9153775983541335e-05, "loss": 0.2924, "step": 2616 }, { "epoch": 0.05, "learning_rate": 4.91531290232841e-05, "loss": 0.1081, "step": 2618 }, { "epoch": 0.05, "learning_rate": 4.915248206302687e-05, "loss": 0.0838, "step": 2620 }, { "epoch": 0.05, "learning_rate": 4.915183510276964e-05, "loss": 0.1481, "step": 2622 }, { "epoch": 0.05, "learning_rate": 4.9151188142512404e-05, "loss": 0.0838, "step": 2624 }, { "epoch": 0.05, "learning_rate": 4.915054118225518e-05, "loss": 0.0888, "step": 2626 }, { "epoch": 0.05, "learning_rate": 4.914989422199794e-05, "loss": 0.0878, "step": 2628 }, { "epoch": 0.05, "learning_rate": 4.914924726174072e-05, "loss": 0.0981, "step": 2630 }, { "epoch": 0.05, "learning_rate": 4.914860030148348e-05, "loss": 0.1031, "step": 2632 }, { "epoch": 0.05, "learning_rate": 4.914795334122625e-05, "loss": 0.1373, "step": 2634 }, { "epoch": 0.05, "learning_rate": 4.914730638096902e-05, "loss": 0.1387, "step": 2636 }, { "epoch": 0.05, "learning_rate": 4.914665942071179e-05, "loss": 0.0949, "step": 2638 }, { "epoch": 0.05, "learning_rate": 4.914601246045456e-05, "loss": 0.0976, "step": 2640 }, { "epoch": 0.05, "learning_rate": 4.9145365500197326e-05, "loss": 0.1047, "step": 2642 }, { "epoch": 0.05, "learning_rate": 4.9144718539940095e-05, "loss": 0.251, "step": 2644 }, { "epoch": 0.05, "learning_rate": 4.914407157968286e-05, "loss": 0.1605, "step": 2646 }, { "epoch": 0.05, "learning_rate": 4.914342461942563e-05, "loss": 0.0901, "step": 2648 }, { "epoch": 0.05, "learning_rate": 4.9142777659168395e-05, "loss": 0.0901, "step": 2650 }, { "epoch": 0.05, "learning_rate": 4.914213069891117e-05, "loss": 0.1228, "step": 2652 }, { "epoch": 0.05, "learning_rate": 4.914148373865394e-05, "loss": 0.1072, "step": 2654 }, { "epoch": 0.05, "learning_rate": 4.91408367783967e-05, "loss": 0.105, "step": 2656 }, { "epoch": 0.05, "learning_rate": 4.914018981813948e-05, "loss": 0.2049, "step": 2658 }, { "epoch": 0.05, "learning_rate": 4.913954285788224e-05, "loss": 0.1365, "step": 2660 }, { "epoch": 0.05, "learning_rate": 4.913889589762501e-05, "loss": 0.1282, "step": 2662 }, { "epoch": 0.05, "learning_rate": 4.913824893736778e-05, "loss": 0.1808, "step": 2664 }, { "epoch": 0.05, "learning_rate": 4.913760197711055e-05, "loss": 0.1007, "step": 2666 }, { "epoch": 0.05, "learning_rate": 4.913695501685332e-05, "loss": 0.102, "step": 2668 }, { "epoch": 0.05, "learning_rate": 4.9136308056596086e-05, "loss": 0.1027, "step": 2670 }, { "epoch": 0.05, "learning_rate": 4.9135661096338855e-05, "loss": 0.1186, "step": 2672 }, { "epoch": 0.05, "learning_rate": 4.9135014136081624e-05, "loss": 0.1838, "step": 2674 }, { "epoch": 0.05, "learning_rate": 4.9134367175824394e-05, "loss": 0.1202, "step": 2676 }, { "epoch": 0.05, "learning_rate": 4.9133720215567156e-05, "loss": 0.1067, "step": 2678 }, { "epoch": 0.05, "learning_rate": 4.913307325530993e-05, "loss": 0.104, "step": 2680 }, { "epoch": 0.05, "learning_rate": 4.9132426295052694e-05, "loss": 0.0818, "step": 2682 }, { "epoch": 0.05, "learning_rate": 4.913177933479546e-05, "loss": 0.1027, "step": 2684 }, { "epoch": 0.05, "learning_rate": 4.913113237453824e-05, "loss": 0.1203, "step": 2686 }, { "epoch": 0.05, "learning_rate": 4.9130485414281e-05, "loss": 0.0493, "step": 2688 }, { "epoch": 0.05, "learning_rate": 4.912983845402378e-05, "loss": 0.156, "step": 2690 }, { "epoch": 0.05, "learning_rate": 4.912919149376654e-05, "loss": 0.1219, "step": 2692 }, { "epoch": 0.05, "learning_rate": 4.912854453350931e-05, "loss": 0.1192, "step": 2694 }, { "epoch": 0.05, "learning_rate": 4.912789757325208e-05, "loss": 0.1471, "step": 2696 }, { "epoch": 0.05, "learning_rate": 4.912725061299485e-05, "loss": 0.1204, "step": 2698 }, { "epoch": 0.05, "learning_rate": 4.9126603652737616e-05, "loss": 0.1108, "step": 2700 }, { "epoch": 0.05, "learning_rate": 4.9125956692480385e-05, "loss": 0.1408, "step": 2702 }, { "epoch": 0.05, "learning_rate": 4.9125309732223154e-05, "loss": 0.1319, "step": 2704 }, { "epoch": 0.05, "learning_rate": 4.9124662771965916e-05, "loss": 0.1571, "step": 2706 }, { "epoch": 0.05, "learning_rate": 4.912401581170869e-05, "loss": 0.0642, "step": 2708 }, { "epoch": 0.05, "learning_rate": 4.9123368851451454e-05, "loss": 0.0755, "step": 2710 }, { "epoch": 0.05, "learning_rate": 4.912272189119423e-05, "loss": 0.1184, "step": 2712 }, { "epoch": 0.05, "learning_rate": 4.912207493093699e-05, "loss": 0.11, "step": 2714 }, { "epoch": 0.05, "learning_rate": 4.912142797067976e-05, "loss": 0.0557, "step": 2716 }, { "epoch": 0.05, "learning_rate": 4.912078101042253e-05, "loss": 0.101, "step": 2718 }, { "epoch": 0.05, "learning_rate": 4.91201340501653e-05, "loss": 0.0854, "step": 2720 }, { "epoch": 0.05, "learning_rate": 4.911948708990807e-05, "loss": 0.1331, "step": 2722 }, { "epoch": 0.05, "learning_rate": 4.911884012965084e-05, "loss": 0.0749, "step": 2724 }, { "epoch": 0.05, "learning_rate": 4.911819316939361e-05, "loss": 0.19, "step": 2726 }, { "epoch": 0.05, "learning_rate": 4.911754620913637e-05, "loss": 0.0651, "step": 2728 }, { "epoch": 0.05, "learning_rate": 4.9116899248879145e-05, "loss": 0.0533, "step": 2730 }, { "epoch": 0.05, "learning_rate": 4.9116252288621914e-05, "loss": 0.1247, "step": 2732 }, { "epoch": 0.05, "learning_rate": 4.9115605328364683e-05, "loss": 0.173, "step": 2734 }, { "epoch": 0.05, "learning_rate": 4.911495836810745e-05, "loss": 0.0837, "step": 2736 }, { "epoch": 0.05, "learning_rate": 4.9114311407850215e-05, "loss": 0.1269, "step": 2738 }, { "epoch": 0.05, "learning_rate": 4.911366444759299e-05, "loss": 0.0661, "step": 2740 }, { "epoch": 0.05, "learning_rate": 4.911301748733575e-05, "loss": 0.1117, "step": 2742 }, { "epoch": 0.05, "learning_rate": 4.911237052707852e-05, "loss": 0.109, "step": 2744 }, { "epoch": 0.05, "learning_rate": 4.911172356682129e-05, "loss": 0.0845, "step": 2746 }, { "epoch": 0.05, "learning_rate": 4.911107660656406e-05, "loss": 0.1148, "step": 2748 }, { "epoch": 0.05, "learning_rate": 4.911042964630683e-05, "loss": 0.0717, "step": 2750 }, { "epoch": 0.05, "learning_rate": 4.91097826860496e-05, "loss": 0.0954, "step": 2752 }, { "epoch": 0.05, "learning_rate": 4.910913572579237e-05, "loss": 0.1292, "step": 2754 }, { "epoch": 0.05, "learning_rate": 4.9108488765535137e-05, "loss": 0.0977, "step": 2756 }, { "epoch": 0.05, "learning_rate": 4.9107841805277906e-05, "loss": 0.1143, "step": 2758 }, { "epoch": 0.05, "learning_rate": 4.910719484502067e-05, "loss": 0.1051, "step": 2760 }, { "epoch": 0.05, "learning_rate": 4.9106547884763444e-05, "loss": 0.1022, "step": 2762 }, { "epoch": 0.05, "learning_rate": 4.910590092450621e-05, "loss": 0.1329, "step": 2764 }, { "epoch": 0.05, "learning_rate": 4.9105253964248975e-05, "loss": 0.0894, "step": 2766 }, { "epoch": 0.05, "learning_rate": 4.910460700399175e-05, "loss": 0.1327, "step": 2768 }, { "epoch": 0.05, "learning_rate": 4.9103960043734513e-05, "loss": 0.2019, "step": 2770 }, { "epoch": 0.05, "learning_rate": 4.910331308347729e-05, "loss": 0.0915, "step": 2772 }, { "epoch": 0.05, "learning_rate": 4.910266612322005e-05, "loss": 0.1476, "step": 2774 }, { "epoch": 0.05, "learning_rate": 4.910201916296282e-05, "loss": 0.1325, "step": 2776 }, { "epoch": 0.05, "learning_rate": 4.910137220270559e-05, "loss": 0.0813, "step": 2778 }, { "epoch": 0.05, "learning_rate": 4.910072524244836e-05, "loss": 0.0578, "step": 2780 }, { "epoch": 0.05, "learning_rate": 4.910007828219113e-05, "loss": 0.0833, "step": 2782 }, { "epoch": 0.05, "learning_rate": 4.90994313219339e-05, "loss": 0.0767, "step": 2784 }, { "epoch": 0.05, "learning_rate": 4.9098784361676666e-05, "loss": 0.1231, "step": 2786 }, { "epoch": 0.05, "learning_rate": 4.909813740141943e-05, "loss": 0.2113, "step": 2788 }, { "epoch": 0.05, "learning_rate": 4.9097490441162204e-05, "loss": 0.0739, "step": 2790 }, { "epoch": 0.05, "learning_rate": 4.9096843480904967e-05, "loss": 0.0734, "step": 2792 }, { "epoch": 0.05, "learning_rate": 4.909619652064774e-05, "loss": 0.1337, "step": 2794 }, { "epoch": 0.05, "learning_rate": 4.9095549560390505e-05, "loss": 0.1443, "step": 2796 }, { "epoch": 0.05, "learning_rate": 4.9094902600133274e-05, "loss": 0.1033, "step": 2798 }, { "epoch": 0.05, "learning_rate": 4.909425563987605e-05, "loss": 0.1779, "step": 2800 }, { "epoch": 0.05, "learning_rate": 4.909360867961881e-05, "loss": 0.1267, "step": 2802 }, { "epoch": 0.05, "learning_rate": 4.909296171936158e-05, "loss": 0.0896, "step": 2804 }, { "epoch": 0.05, "learning_rate": 4.909231475910435e-05, "loss": 0.1526, "step": 2806 }, { "epoch": 0.05, "learning_rate": 4.909166779884712e-05, "loss": 0.0645, "step": 2808 }, { "epoch": 0.05, "learning_rate": 4.909102083858989e-05, "loss": 0.1092, "step": 2810 }, { "epoch": 0.05, "learning_rate": 4.909037387833266e-05, "loss": 0.0811, "step": 2812 }, { "epoch": 0.05, "learning_rate": 4.9089726918075426e-05, "loss": 0.1299, "step": 2814 }, { "epoch": 0.05, "learning_rate": 4.9089079957818196e-05, "loss": 0.0961, "step": 2816 }, { "epoch": 0.05, "learning_rate": 4.9088432997560965e-05, "loss": 0.0995, "step": 2818 }, { "epoch": 0.05, "learning_rate": 4.908778603730373e-05, "loss": 0.0882, "step": 2820 }, { "epoch": 0.05, "learning_rate": 4.90871390770465e-05, "loss": 0.0563, "step": 2822 }, { "epoch": 0.05, "learning_rate": 4.9086492116789265e-05, "loss": 0.1143, "step": 2824 }, { "epoch": 0.05, "learning_rate": 4.9085845156532034e-05, "loss": 0.1094, "step": 2826 }, { "epoch": 0.05, "learning_rate": 4.90851981962748e-05, "loss": 0.1135, "step": 2828 }, { "epoch": 0.05, "learning_rate": 4.908455123601757e-05, "loss": 0.0788, "step": 2830 }, { "epoch": 0.05, "learning_rate": 4.908390427576035e-05, "loss": 0.1192, "step": 2832 }, { "epoch": 0.06, "learning_rate": 4.908325731550311e-05, "loss": 0.069, "step": 2834 }, { "epoch": 0.06, "learning_rate": 4.908261035524588e-05, "loss": 0.1067, "step": 2836 }, { "epoch": 0.06, "learning_rate": 4.908196339498865e-05, "loss": 0.1055, "step": 2838 }, { "epoch": 0.06, "learning_rate": 4.908131643473142e-05, "loss": 0.0835, "step": 2840 }, { "epoch": 0.06, "learning_rate": 4.908066947447419e-05, "loss": 0.102, "step": 2842 }, { "epoch": 0.06, "learning_rate": 4.9080022514216956e-05, "loss": 0.093, "step": 2844 }, { "epoch": 0.06, "learning_rate": 4.9079375553959725e-05, "loss": 0.1047, "step": 2846 }, { "epoch": 0.06, "learning_rate": 4.907872859370249e-05, "loss": 0.1246, "step": 2848 }, { "epoch": 0.06, "learning_rate": 4.907808163344526e-05, "loss": 0.0791, "step": 2850 }, { "epoch": 0.06, "learning_rate": 4.9077434673188026e-05, "loss": 0.1039, "step": 2852 }, { "epoch": 0.06, "learning_rate": 4.90767877129308e-05, "loss": 0.1094, "step": 2854 }, { "epoch": 0.06, "learning_rate": 4.9076140752673564e-05, "loss": 0.0849, "step": 2856 }, { "epoch": 0.06, "learning_rate": 4.907549379241633e-05, "loss": 0.1164, "step": 2858 }, { "epoch": 0.06, "learning_rate": 4.90748468321591e-05, "loss": 0.1267, "step": 2860 }, { "epoch": 0.06, "learning_rate": 4.907419987190187e-05, "loss": 0.0922, "step": 2862 }, { "epoch": 0.06, "learning_rate": 4.907355291164464e-05, "loss": 0.1467, "step": 2864 }, { "epoch": 0.06, "learning_rate": 4.907290595138741e-05, "loss": 0.0831, "step": 2866 }, { "epoch": 0.06, "learning_rate": 4.907225899113018e-05, "loss": 0.1016, "step": 2868 }, { "epoch": 0.06, "learning_rate": 4.907161203087294e-05, "loss": 0.0991, "step": 2870 }, { "epoch": 0.06, "learning_rate": 4.9070965070615716e-05, "loss": 0.069, "step": 2872 }, { "epoch": 0.06, "learning_rate": 4.907031811035848e-05, "loss": 0.1, "step": 2874 }, { "epoch": 0.06, "learning_rate": 4.9069671150101255e-05, "loss": 0.0764, "step": 2876 }, { "epoch": 0.06, "learning_rate": 4.9069024189844024e-05, "loss": 0.0936, "step": 2878 }, { "epoch": 0.06, "learning_rate": 4.9068377229586786e-05, "loss": 0.1098, "step": 2880 }, { "epoch": 0.06, "learning_rate": 4.906773026932956e-05, "loss": 0.1156, "step": 2882 }, { "epoch": 0.06, "learning_rate": 4.9067083309072324e-05, "loss": 0.0746, "step": 2884 }, { "epoch": 0.06, "learning_rate": 4.906643634881509e-05, "loss": 0.0568, "step": 2886 }, { "epoch": 0.06, "learning_rate": 4.906578938855786e-05, "loss": 0.0831, "step": 2888 }, { "epoch": 0.06, "learning_rate": 4.906514242830063e-05, "loss": 0.087, "step": 2890 }, { "epoch": 0.06, "learning_rate": 4.90644954680434e-05, "loss": 0.0881, "step": 2892 }, { "epoch": 0.06, "learning_rate": 4.906384850778617e-05, "loss": 0.1979, "step": 2894 }, { "epoch": 0.06, "learning_rate": 4.906320154752894e-05, "loss": 0.0558, "step": 2896 }, { "epoch": 0.06, "learning_rate": 4.906255458727171e-05, "loss": 0.068, "step": 2898 }, { "epoch": 0.06, "learning_rate": 4.906190762701448e-05, "loss": 0.05, "step": 2900 }, { "epoch": 0.06, "learning_rate": 4.906126066675724e-05, "loss": 0.1043, "step": 2902 }, { "epoch": 0.06, "learning_rate": 4.9060613706500015e-05, "loss": 0.1672, "step": 2904 }, { "epoch": 0.06, "learning_rate": 4.905996674624278e-05, "loss": 0.0799, "step": 2906 }, { "epoch": 0.06, "learning_rate": 4.9059319785985546e-05, "loss": 0.066, "step": 2908 }, { "epoch": 0.06, "learning_rate": 4.905867282572832e-05, "loss": 0.0684, "step": 2910 }, { "epoch": 0.06, "learning_rate": 4.9058025865471085e-05, "loss": 0.0631, "step": 2912 }, { "epoch": 0.06, "learning_rate": 4.905737890521386e-05, "loss": 0.1484, "step": 2914 }, { "epoch": 0.06, "learning_rate": 4.905673194495662e-05, "loss": 0.095, "step": 2916 }, { "epoch": 0.06, "learning_rate": 4.905608498469939e-05, "loss": 0.1006, "step": 2918 }, { "epoch": 0.06, "learning_rate": 4.905543802444216e-05, "loss": 0.084, "step": 2920 }, { "epoch": 0.06, "learning_rate": 4.905479106418493e-05, "loss": 0.0845, "step": 2922 }, { "epoch": 0.06, "learning_rate": 4.90541441039277e-05, "loss": 0.1408, "step": 2924 }, { "epoch": 0.06, "learning_rate": 4.905349714367047e-05, "loss": 0.1087, "step": 2926 }, { "epoch": 0.06, "learning_rate": 4.905285018341324e-05, "loss": 0.0698, "step": 2928 }, { "epoch": 0.06, "learning_rate": 4.9052203223156e-05, "loss": 0.0486, "step": 2930 }, { "epoch": 0.06, "learning_rate": 4.9051556262898775e-05, "loss": 0.1164, "step": 2932 }, { "epoch": 0.06, "learning_rate": 4.905090930264154e-05, "loss": 0.0541, "step": 2934 }, { "epoch": 0.06, "learning_rate": 4.9050262342384314e-05, "loss": 0.1562, "step": 2936 }, { "epoch": 0.06, "learning_rate": 4.9049615382127076e-05, "loss": 0.084, "step": 2938 }, { "epoch": 0.06, "learning_rate": 4.9048968421869845e-05, "loss": 0.1463, "step": 2940 }, { "epoch": 0.06, "learning_rate": 4.904832146161262e-05, "loss": 0.1118, "step": 2942 }, { "epoch": 0.06, "learning_rate": 4.904767450135538e-05, "loss": 0.1123, "step": 2944 }, { "epoch": 0.06, "learning_rate": 4.904702754109815e-05, "loss": 0.1148, "step": 2946 }, { "epoch": 0.06, "learning_rate": 4.904638058084092e-05, "loss": 0.0694, "step": 2948 }, { "epoch": 0.06, "learning_rate": 4.904573362058369e-05, "loss": 0.0674, "step": 2950 }, { "epoch": 0.06, "learning_rate": 4.904508666032645e-05, "loss": 0.0651, "step": 2952 }, { "epoch": 0.06, "learning_rate": 4.904443970006923e-05, "loss": 0.1132, "step": 2954 }, { "epoch": 0.06, "learning_rate": 4.9043792739812e-05, "loss": 0.0572, "step": 2956 }, { "epoch": 0.06, "learning_rate": 4.904314577955477e-05, "loss": 0.0868, "step": 2958 }, { "epoch": 0.06, "learning_rate": 4.9042498819297536e-05, "loss": 0.0703, "step": 2960 }, { "epoch": 0.06, "learning_rate": 4.90418518590403e-05, "loss": 0.0636, "step": 2962 }, { "epoch": 0.06, "learning_rate": 4.9041204898783074e-05, "loss": 0.1147, "step": 2964 }, { "epoch": 0.06, "learning_rate": 4.9040557938525836e-05, "loss": 0.1203, "step": 2966 }, { "epoch": 0.06, "learning_rate": 4.9039910978268605e-05, "loss": 0.1047, "step": 2968 }, { "epoch": 0.06, "learning_rate": 4.9039264018011374e-05, "loss": 0.08, "step": 2970 }, { "epoch": 0.06, "learning_rate": 4.9038617057754143e-05, "loss": 0.0958, "step": 2972 }, { "epoch": 0.06, "learning_rate": 4.903797009749691e-05, "loss": 0.0979, "step": 2974 }, { "epoch": 0.06, "learning_rate": 4.903732313723968e-05, "loss": 0.1811, "step": 2976 }, { "epoch": 0.06, "learning_rate": 4.903667617698245e-05, "loss": 0.0775, "step": 2978 }, { "epoch": 0.06, "learning_rate": 4.903602921672522e-05, "loss": 0.0655, "step": 2980 }, { "epoch": 0.06, "learning_rate": 4.903538225646799e-05, "loss": 0.1104, "step": 2982 }, { "epoch": 0.06, "learning_rate": 4.903473529621075e-05, "loss": 0.0887, "step": 2984 }, { "epoch": 0.06, "learning_rate": 4.903408833595353e-05, "loss": 0.1632, "step": 2986 }, { "epoch": 0.06, "learning_rate": 4.9033441375696296e-05, "loss": 0.0635, "step": 2988 }, { "epoch": 0.06, "learning_rate": 4.903279441543906e-05, "loss": 0.1192, "step": 2990 }, { "epoch": 0.06, "learning_rate": 4.9032147455181834e-05, "loss": 0.0727, "step": 2992 }, { "epoch": 0.06, "learning_rate": 4.90315004949246e-05, "loss": 0.1735, "step": 2994 }, { "epoch": 0.06, "learning_rate": 4.903085353466737e-05, "loss": 0.0792, "step": 2996 }, { "epoch": 0.06, "learning_rate": 4.9030206574410135e-05, "loss": 0.0947, "step": 2998 }, { "epoch": 0.06, "learning_rate": 4.9029559614152904e-05, "loss": 0.1145, "step": 3000 }, { "epoch": 0.06, "learning_rate": 4.902891265389567e-05, "loss": 0.0767, "step": 3002 }, { "epoch": 0.06, "learning_rate": 4.902826569363844e-05, "loss": 0.1044, "step": 3004 }, { "epoch": 0.06, "learning_rate": 4.902761873338121e-05, "loss": 0.1015, "step": 3006 }, { "epoch": 0.06, "learning_rate": 4.902697177312398e-05, "loss": 0.0754, "step": 3008 }, { "epoch": 0.06, "learning_rate": 4.902632481286675e-05, "loss": 0.0789, "step": 3010 }, { "epoch": 0.06, "learning_rate": 4.902567785260951e-05, "loss": 0.1546, "step": 3012 }, { "epoch": 0.06, "learning_rate": 4.902503089235229e-05, "loss": 0.0937, "step": 3014 }, { "epoch": 0.06, "learning_rate": 4.902438393209505e-05, "loss": 0.112, "step": 3016 }, { "epoch": 0.06, "learning_rate": 4.9023736971837826e-05, "loss": 0.0768, "step": 3018 }, { "epoch": 0.06, "learning_rate": 4.902309001158059e-05, "loss": 0.0939, "step": 3020 }, { "epoch": 0.06, "learning_rate": 4.902244305132336e-05, "loss": 0.0911, "step": 3022 }, { "epoch": 0.06, "learning_rate": 4.902179609106613e-05, "loss": 0.0989, "step": 3024 }, { "epoch": 0.06, "learning_rate": 4.9021149130808895e-05, "loss": 0.2013, "step": 3026 }, { "epoch": 0.06, "learning_rate": 4.9020502170551664e-05, "loss": 0.0921, "step": 3028 }, { "epoch": 0.06, "learning_rate": 4.901985521029443e-05, "loss": 0.0713, "step": 3030 }, { "epoch": 0.06, "learning_rate": 4.90192082500372e-05, "loss": 0.0809, "step": 3032 }, { "epoch": 0.06, "learning_rate": 4.901856128977997e-05, "loss": 0.126, "step": 3034 }, { "epoch": 0.06, "learning_rate": 4.901791432952274e-05, "loss": 0.0887, "step": 3036 }, { "epoch": 0.06, "learning_rate": 4.901726736926551e-05, "loss": 0.1064, "step": 3038 }, { "epoch": 0.06, "learning_rate": 4.901662040900828e-05, "loss": 0.0849, "step": 3040 }, { "epoch": 0.06, "learning_rate": 4.901597344875105e-05, "loss": 0.0538, "step": 3042 }, { "epoch": 0.06, "learning_rate": 4.901532648849381e-05, "loss": 0.0974, "step": 3044 }, { "epoch": 0.06, "learning_rate": 4.9014679528236586e-05, "loss": 0.0777, "step": 3046 }, { "epoch": 0.06, "learning_rate": 4.901403256797935e-05, "loss": 0.0766, "step": 3048 }, { "epoch": 0.06, "learning_rate": 4.901338560772212e-05, "loss": 0.0711, "step": 3050 }, { "epoch": 0.06, "learning_rate": 4.9012738647464887e-05, "loss": 0.0834, "step": 3052 }, { "epoch": 0.06, "learning_rate": 4.9012091687207656e-05, "loss": 0.1285, "step": 3054 }, { "epoch": 0.06, "learning_rate": 4.901144472695043e-05, "loss": 0.0821, "step": 3056 }, { "epoch": 0.06, "learning_rate": 4.9010797766693194e-05, "loss": 0.1203, "step": 3058 }, { "epoch": 0.06, "learning_rate": 4.901015080643596e-05, "loss": 0.0521, "step": 3060 }, { "epoch": 0.06, "learning_rate": 4.900950384617873e-05, "loss": 0.073, "step": 3062 }, { "epoch": 0.06, "learning_rate": 4.90088568859215e-05, "loss": 0.1063, "step": 3064 }, { "epoch": 0.06, "learning_rate": 4.900820992566427e-05, "loss": 0.0664, "step": 3066 }, { "epoch": 0.06, "learning_rate": 4.900756296540704e-05, "loss": 0.0573, "step": 3068 }, { "epoch": 0.06, "learning_rate": 4.900691600514981e-05, "loss": 0.1072, "step": 3070 }, { "epoch": 0.06, "learning_rate": 4.900626904489257e-05, "loss": 0.0896, "step": 3072 }, { "epoch": 0.06, "learning_rate": 4.9005622084635346e-05, "loss": 0.1249, "step": 3074 }, { "epoch": 0.06, "learning_rate": 4.900497512437811e-05, "loss": 0.1154, "step": 3076 }, { "epoch": 0.06, "learning_rate": 4.9004328164120885e-05, "loss": 0.0869, "step": 3078 }, { "epoch": 0.06, "learning_rate": 4.900368120386365e-05, "loss": 0.0984, "step": 3080 }, { "epoch": 0.06, "learning_rate": 4.9003034243606416e-05, "loss": 0.0855, "step": 3082 }, { "epoch": 0.06, "learning_rate": 4.9002387283349185e-05, "loss": 0.0859, "step": 3084 }, { "epoch": 0.06, "learning_rate": 4.9001740323091954e-05, "loss": 0.0522, "step": 3086 }, { "epoch": 0.06, "learning_rate": 4.900109336283472e-05, "loss": 0.0996, "step": 3088 }, { "epoch": 0.06, "learning_rate": 4.900044640257749e-05, "loss": 0.1025, "step": 3090 }, { "epoch": 0.06, "learning_rate": 4.899979944232026e-05, "loss": 0.0807, "step": 3092 }, { "epoch": 0.06, "learning_rate": 4.8999152482063024e-05, "loss": 0.1388, "step": 3094 }, { "epoch": 0.06, "learning_rate": 4.89985055218058e-05, "loss": 0.0778, "step": 3096 }, { "epoch": 0.06, "learning_rate": 4.899785856154856e-05, "loss": 0.0722, "step": 3098 }, { "epoch": 0.06, "learning_rate": 4.899721160129134e-05, "loss": 0.1152, "step": 3100 }, { "epoch": 0.06, "learning_rate": 4.899656464103411e-05, "loss": 0.0966, "step": 3102 }, { "epoch": 0.06, "learning_rate": 4.899591768077687e-05, "loss": 0.1095, "step": 3104 }, { "epoch": 0.06, "learning_rate": 4.8995270720519645e-05, "loss": 0.1086, "step": 3106 }, { "epoch": 0.06, "learning_rate": 4.899462376026241e-05, "loss": 0.069, "step": 3108 }, { "epoch": 0.06, "learning_rate": 4.8993976800005176e-05, "loss": 0.1226, "step": 3110 }, { "epoch": 0.06, "learning_rate": 4.8993329839747946e-05, "loss": 0.109, "step": 3112 }, { "epoch": 0.06, "learning_rate": 4.8992682879490715e-05, "loss": 0.0881, "step": 3114 }, { "epoch": 0.06, "learning_rate": 4.8992035919233484e-05, "loss": 0.1712, "step": 3116 }, { "epoch": 0.06, "learning_rate": 4.899138895897625e-05, "loss": 0.0681, "step": 3118 }, { "epoch": 0.06, "learning_rate": 4.899074199871902e-05, "loss": 0.07, "step": 3120 }, { "epoch": 0.06, "learning_rate": 4.899009503846179e-05, "loss": 0.2801, "step": 3122 }, { "epoch": 0.06, "learning_rate": 4.898944807820456e-05, "loss": 0.0567, "step": 3124 }, { "epoch": 0.06, "learning_rate": 4.898880111794732e-05, "loss": 0.0678, "step": 3126 }, { "epoch": 0.06, "learning_rate": 4.89881541576901e-05, "loss": 0.0767, "step": 3128 }, { "epoch": 0.06, "learning_rate": 4.898750719743286e-05, "loss": 0.0803, "step": 3130 }, { "epoch": 0.06, "learning_rate": 4.898686023717563e-05, "loss": 0.0815, "step": 3132 }, { "epoch": 0.06, "learning_rate": 4.8986213276918405e-05, "loss": 0.0826, "step": 3134 }, { "epoch": 0.06, "learning_rate": 4.898556631666117e-05, "loss": 0.0811, "step": 3136 }, { "epoch": 0.06, "learning_rate": 4.8984919356403944e-05, "loss": 0.0687, "step": 3138 }, { "epoch": 0.06, "learning_rate": 4.8984272396146706e-05, "loss": 0.0796, "step": 3140 }, { "epoch": 0.06, "learning_rate": 4.8983625435889475e-05, "loss": 0.0543, "step": 3142 }, { "epoch": 0.06, "learning_rate": 4.8982978475632244e-05, "loss": 0.1067, "step": 3144 }, { "epoch": 0.06, "learning_rate": 4.898233151537501e-05, "loss": 0.0589, "step": 3146 }, { "epoch": 0.06, "learning_rate": 4.898168455511778e-05, "loss": 0.2048, "step": 3148 }, { "epoch": 0.06, "learning_rate": 4.898103759486055e-05, "loss": 0.0326, "step": 3150 }, { "epoch": 0.06, "learning_rate": 4.898039063460332e-05, "loss": 0.0785, "step": 3152 }, { "epoch": 0.06, "learning_rate": 4.897974367434608e-05, "loss": 0.0922, "step": 3154 }, { "epoch": 0.06, "learning_rate": 4.897909671408886e-05, "loss": 0.1006, "step": 3156 }, { "epoch": 0.06, "learning_rate": 4.897844975383162e-05, "loss": 0.1087, "step": 3158 }, { "epoch": 0.06, "learning_rate": 4.89778027935744e-05, "loss": 0.0685, "step": 3160 }, { "epoch": 0.06, "learning_rate": 4.897715583331716e-05, "loss": 0.1117, "step": 3162 }, { "epoch": 0.06, "learning_rate": 4.897650887305993e-05, "loss": 0.1424, "step": 3164 }, { "epoch": 0.06, "learning_rate": 4.8975861912802704e-05, "loss": 0.0631, "step": 3166 }, { "epoch": 0.06, "learning_rate": 4.8975214952545466e-05, "loss": 0.1617, "step": 3168 }, { "epoch": 0.06, "learning_rate": 4.8974567992288235e-05, "loss": 0.0719, "step": 3170 }, { "epoch": 0.06, "learning_rate": 4.8973921032031004e-05, "loss": 0.125, "step": 3172 }, { "epoch": 0.06, "learning_rate": 4.8973274071773774e-05, "loss": 0.0641, "step": 3174 }, { "epoch": 0.06, "learning_rate": 4.8972627111516536e-05, "loss": 0.1204, "step": 3176 }, { "epoch": 0.06, "learning_rate": 4.897198015125931e-05, "loss": 0.0821, "step": 3178 }, { "epoch": 0.06, "learning_rate": 4.897133319100208e-05, "loss": 0.1118, "step": 3180 }, { "epoch": 0.06, "learning_rate": 4.897068623074485e-05, "loss": 0.0834, "step": 3182 }, { "epoch": 0.06, "learning_rate": 4.897003927048762e-05, "loss": 0.1028, "step": 3184 }, { "epoch": 0.06, "learning_rate": 4.896939231023038e-05, "loss": 0.0799, "step": 3186 }, { "epoch": 0.06, "learning_rate": 4.896874534997316e-05, "loss": 0.0373, "step": 3188 }, { "epoch": 0.06, "learning_rate": 4.896809838971592e-05, "loss": 0.044, "step": 3190 }, { "epoch": 0.06, "learning_rate": 4.896745142945869e-05, "loss": 0.0959, "step": 3192 }, { "epoch": 0.06, "learning_rate": 4.896680446920146e-05, "loss": 0.0563, "step": 3194 }, { "epoch": 0.06, "learning_rate": 4.896615750894423e-05, "loss": 0.0785, "step": 3196 }, { "epoch": 0.06, "learning_rate": 4.8965510548686996e-05, "loss": 0.0963, "step": 3198 }, { "epoch": 0.06, "learning_rate": 4.8964863588429765e-05, "loss": 0.1163, "step": 3200 }, { "epoch": 0.06, "learning_rate": 4.8964216628172534e-05, "loss": 0.0634, "step": 3202 }, { "epoch": 0.06, "learning_rate": 4.89635696679153e-05, "loss": 0.0647, "step": 3204 }, { "epoch": 0.06, "learning_rate": 4.896292270765807e-05, "loss": 0.1031, "step": 3206 }, { "epoch": 0.06, "learning_rate": 4.8962275747400834e-05, "loss": 0.0784, "step": 3208 }, { "epoch": 0.06, "learning_rate": 4.896162878714361e-05, "loss": 0.0802, "step": 3210 }, { "epoch": 0.06, "learning_rate": 4.896098182688638e-05, "loss": 0.1355, "step": 3212 }, { "epoch": 0.06, "learning_rate": 4.896033486662914e-05, "loss": 0.0782, "step": 3214 }, { "epoch": 0.06, "learning_rate": 4.895968790637192e-05, "loss": 0.0301, "step": 3216 }, { "epoch": 0.06, "learning_rate": 4.895904094611468e-05, "loss": 0.0914, "step": 3218 }, { "epoch": 0.06, "learning_rate": 4.8958393985857456e-05, "loss": 0.0867, "step": 3220 }, { "epoch": 0.06, "learning_rate": 4.895774702560022e-05, "loss": 0.0885, "step": 3222 }, { "epoch": 0.06, "learning_rate": 4.895710006534299e-05, "loss": 0.0583, "step": 3224 }, { "epoch": 0.06, "learning_rate": 4.8956453105085756e-05, "loss": 0.0425, "step": 3226 }, { "epoch": 0.06, "learning_rate": 4.8955806144828525e-05, "loss": 0.0664, "step": 3228 }, { "epoch": 0.06, "learning_rate": 4.8955159184571294e-05, "loss": 0.0648, "step": 3230 }, { "epoch": 0.06, "learning_rate": 4.8954512224314063e-05, "loss": 0.0848, "step": 3232 }, { "epoch": 0.06, "learning_rate": 4.895386526405683e-05, "loss": 0.2053, "step": 3234 }, { "epoch": 0.06, "learning_rate": 4.8953218303799595e-05, "loss": 0.114, "step": 3236 }, { "epoch": 0.06, "learning_rate": 4.895257134354237e-05, "loss": 0.0959, "step": 3238 }, { "epoch": 0.06, "learning_rate": 4.895192438328513e-05, "loss": 0.0947, "step": 3240 }, { "epoch": 0.06, "learning_rate": 4.895127742302791e-05, "loss": 0.0861, "step": 3242 }, { "epoch": 0.06, "learning_rate": 4.895063046277068e-05, "loss": 0.1088, "step": 3244 }, { "epoch": 0.06, "learning_rate": 4.894998350251344e-05, "loss": 0.0634, "step": 3246 }, { "epoch": 0.06, "learning_rate": 4.8949336542256216e-05, "loss": 0.0804, "step": 3248 }, { "epoch": 0.06, "learning_rate": 4.894868958199898e-05, "loss": 0.0805, "step": 3250 }, { "epoch": 0.06, "learning_rate": 4.894804262174175e-05, "loss": 0.0712, "step": 3252 }, { "epoch": 0.06, "learning_rate": 4.894739566148452e-05, "loss": 0.0658, "step": 3254 }, { "epoch": 0.06, "learning_rate": 4.8946748701227286e-05, "loss": 0.058, "step": 3256 }, { "epoch": 0.06, "learning_rate": 4.8946101740970055e-05, "loss": 0.0504, "step": 3258 }, { "epoch": 0.06, "learning_rate": 4.8945454780712824e-05, "loss": 0.052, "step": 3260 }, { "epoch": 0.06, "learning_rate": 4.894480782045559e-05, "loss": 0.077, "step": 3262 }, { "epoch": 0.06, "learning_rate": 4.894416086019836e-05, "loss": 0.063, "step": 3264 }, { "epoch": 0.06, "learning_rate": 4.894351389994113e-05, "loss": 0.1295, "step": 3266 }, { "epoch": 0.06, "learning_rate": 4.8942866939683893e-05, "loss": 0.0968, "step": 3268 }, { "epoch": 0.06, "learning_rate": 4.894221997942667e-05, "loss": 0.0581, "step": 3270 }, { "epoch": 0.06, "learning_rate": 4.894157301916943e-05, "loss": 0.0854, "step": 3272 }, { "epoch": 0.06, "learning_rate": 4.89409260589122e-05, "loss": 0.0407, "step": 3274 }, { "epoch": 0.06, "learning_rate": 4.894027909865497e-05, "loss": 0.1351, "step": 3276 }, { "epoch": 0.06, "learning_rate": 4.893963213839774e-05, "loss": 0.0759, "step": 3278 }, { "epoch": 0.06, "learning_rate": 4.8938985178140515e-05, "loss": 0.1155, "step": 3280 }, { "epoch": 0.06, "learning_rate": 4.893833821788328e-05, "loss": 0.0728, "step": 3282 }, { "epoch": 0.06, "learning_rate": 4.8937691257626046e-05, "loss": 0.073, "step": 3284 }, { "epoch": 0.06, "learning_rate": 4.8937044297368815e-05, "loss": 0.0727, "step": 3286 }, { "epoch": 0.06, "learning_rate": 4.8936397337111584e-05, "loss": 0.0512, "step": 3288 }, { "epoch": 0.06, "learning_rate": 4.893575037685435e-05, "loss": 0.0775, "step": 3290 }, { "epoch": 0.06, "learning_rate": 4.893510341659712e-05, "loss": 0.0825, "step": 3292 }, { "epoch": 0.06, "learning_rate": 4.893445645633989e-05, "loss": 0.0848, "step": 3294 }, { "epoch": 0.06, "learning_rate": 4.8933809496082654e-05, "loss": 0.0614, "step": 3296 }, { "epoch": 0.06, "learning_rate": 4.893316253582543e-05, "loss": 0.076, "step": 3298 }, { "epoch": 0.06, "learning_rate": 4.893251557556819e-05, "loss": 0.0915, "step": 3300 }, { "epoch": 0.06, "learning_rate": 4.893186861531097e-05, "loss": 0.0579, "step": 3302 }, { "epoch": 0.06, "learning_rate": 4.893122165505373e-05, "loss": 0.1383, "step": 3304 }, { "epoch": 0.06, "learning_rate": 4.89305746947965e-05, "loss": 0.0708, "step": 3306 }, { "epoch": 0.06, "learning_rate": 4.892992773453927e-05, "loss": 0.0839, "step": 3308 }, { "epoch": 0.06, "learning_rate": 4.892928077428204e-05, "loss": 0.0813, "step": 3310 }, { "epoch": 0.06, "learning_rate": 4.8928633814024807e-05, "loss": 0.0483, "step": 3312 }, { "epoch": 0.06, "learning_rate": 4.8927986853767576e-05, "loss": 0.0612, "step": 3314 }, { "epoch": 0.06, "learning_rate": 4.8927339893510345e-05, "loss": 0.1333, "step": 3316 }, { "epoch": 0.06, "learning_rate": 4.892669293325311e-05, "loss": 0.0647, "step": 3318 }, { "epoch": 0.06, "learning_rate": 4.892604597299588e-05, "loss": 0.1445, "step": 3320 }, { "epoch": 0.06, "learning_rate": 4.892539901273865e-05, "loss": 0.1023, "step": 3322 }, { "epoch": 0.06, "learning_rate": 4.892475205248142e-05, "loss": 0.0538, "step": 3324 }, { "epoch": 0.06, "learning_rate": 4.892410509222419e-05, "loss": 0.0491, "step": 3326 }, { "epoch": 0.06, "learning_rate": 4.892345813196695e-05, "loss": 0.094, "step": 3328 }, { "epoch": 0.06, "learning_rate": 4.892281117170973e-05, "loss": 0.0604, "step": 3330 }, { "epoch": 0.06, "learning_rate": 4.892216421145249e-05, "loss": 0.1051, "step": 3332 }, { "epoch": 0.06, "learning_rate": 4.892151725119526e-05, "loss": 0.1085, "step": 3334 }, { "epoch": 0.06, "learning_rate": 4.892087029093803e-05, "loss": 0.0792, "step": 3336 }, { "epoch": 0.06, "learning_rate": 4.89202233306808e-05, "loss": 0.1349, "step": 3338 }, { "epoch": 0.06, "learning_rate": 4.891957637042357e-05, "loss": 0.0749, "step": 3340 }, { "epoch": 0.06, "learning_rate": 4.8918929410166336e-05, "loss": 0.0417, "step": 3342 }, { "epoch": 0.06, "learning_rate": 4.8918282449909105e-05, "loss": 0.0848, "step": 3344 }, { "epoch": 0.06, "learning_rate": 4.8917635489651874e-05, "loss": 0.0799, "step": 3346 }, { "epoch": 0.06, "learning_rate": 4.891698852939464e-05, "loss": 0.0695, "step": 3348 }, { "epoch": 0.07, "learning_rate": 4.8916341569137406e-05, "loss": 0.098, "step": 3350 }, { "epoch": 0.07, "learning_rate": 4.891569460888018e-05, "loss": 0.0689, "step": 3352 }, { "epoch": 0.07, "learning_rate": 4.8915047648622944e-05, "loss": 0.0623, "step": 3354 }, { "epoch": 0.07, "learning_rate": 4.891440068836571e-05, "loss": 0.1276, "step": 3356 }, { "epoch": 0.07, "learning_rate": 4.891375372810849e-05, "loss": 0.1027, "step": 3358 }, { "epoch": 0.07, "learning_rate": 4.891310676785125e-05, "loss": 0.1104, "step": 3360 }, { "epoch": 0.07, "learning_rate": 4.891245980759403e-05, "loss": 0.0629, "step": 3362 }, { "epoch": 0.07, "learning_rate": 4.891181284733679e-05, "loss": 0.0667, "step": 3364 }, { "epoch": 0.07, "learning_rate": 4.891116588707956e-05, "loss": 0.0751, "step": 3366 }, { "epoch": 0.07, "learning_rate": 4.891051892682233e-05, "loss": 0.0766, "step": 3368 }, { "epoch": 0.07, "learning_rate": 4.8909871966565096e-05, "loss": 0.0399, "step": 3370 }, { "epoch": 0.07, "learning_rate": 4.8909225006307865e-05, "loss": 0.0972, "step": 3372 }, { "epoch": 0.07, "learning_rate": 4.8908578046050635e-05, "loss": 0.0536, "step": 3374 }, { "epoch": 0.07, "learning_rate": 4.8907931085793404e-05, "loss": 0.1145, "step": 3376 }, { "epoch": 0.07, "learning_rate": 4.8907284125536166e-05, "loss": 0.0637, "step": 3378 }, { "epoch": 0.07, "learning_rate": 4.890663716527894e-05, "loss": 0.1069, "step": 3380 }, { "epoch": 0.07, "learning_rate": 4.8905990205021704e-05, "loss": 0.0704, "step": 3382 }, { "epoch": 0.07, "learning_rate": 4.890534324476448e-05, "loss": 0.1174, "step": 3384 }, { "epoch": 0.07, "learning_rate": 4.890469628450724e-05, "loss": 0.086, "step": 3386 }, { "epoch": 0.07, "learning_rate": 4.890404932425001e-05, "loss": 0.1038, "step": 3388 }, { "epoch": 0.07, "learning_rate": 4.890340236399279e-05, "loss": 0.0902, "step": 3390 }, { "epoch": 0.07, "learning_rate": 4.890275540373555e-05, "loss": 0.0949, "step": 3392 }, { "epoch": 0.07, "learning_rate": 4.890210844347832e-05, "loss": 0.091, "step": 3394 }, { "epoch": 0.07, "learning_rate": 4.890146148322109e-05, "loss": 0.1156, "step": 3396 }, { "epoch": 0.07, "learning_rate": 4.890081452296386e-05, "loss": 0.0771, "step": 3398 }, { "epoch": 0.07, "learning_rate": 4.8900167562706626e-05, "loss": 0.0552, "step": 3400 }, { "epoch": 0.07, "learning_rate": 4.8899520602449395e-05, "loss": 0.0607, "step": 3402 }, { "epoch": 0.07, "learning_rate": 4.8898873642192164e-05, "loss": 0.0866, "step": 3404 }, { "epoch": 0.07, "learning_rate": 4.889822668193493e-05, "loss": 0.0546, "step": 3406 }, { "epoch": 0.07, "learning_rate": 4.88975797216777e-05, "loss": 0.124, "step": 3408 }, { "epoch": 0.07, "learning_rate": 4.8896932761420465e-05, "loss": 0.0563, "step": 3410 }, { "epoch": 0.07, "learning_rate": 4.889628580116324e-05, "loss": 0.0556, "step": 3412 }, { "epoch": 0.07, "learning_rate": 4.8895638840906e-05, "loss": 0.0353, "step": 3414 }, { "epoch": 0.07, "learning_rate": 4.889499188064877e-05, "loss": 0.0611, "step": 3416 }, { "epoch": 0.07, "learning_rate": 4.889434492039154e-05, "loss": 0.0515, "step": 3418 }, { "epoch": 0.07, "learning_rate": 4.889369796013431e-05, "loss": 0.06, "step": 3420 }, { "epoch": 0.07, "learning_rate": 4.8893050999877086e-05, "loss": 0.0649, "step": 3422 }, { "epoch": 0.07, "learning_rate": 4.889240403961985e-05, "loss": 0.0882, "step": 3424 }, { "epoch": 0.07, "learning_rate": 4.889175707936262e-05, "loss": 0.1074, "step": 3426 }, { "epoch": 0.07, "learning_rate": 4.8891110119105386e-05, "loss": 0.089, "step": 3428 }, { "epoch": 0.07, "learning_rate": 4.8890463158848155e-05, "loss": 0.0988, "step": 3430 }, { "epoch": 0.07, "learning_rate": 4.888981619859092e-05, "loss": 0.1415, "step": 3432 }, { "epoch": 0.07, "learning_rate": 4.8889169238333694e-05, "loss": 0.0779, "step": 3434 }, { "epoch": 0.07, "learning_rate": 4.888852227807646e-05, "loss": 0.0547, "step": 3436 }, { "epoch": 0.07, "learning_rate": 4.8887875317819225e-05, "loss": 0.2245, "step": 3438 }, { "epoch": 0.07, "learning_rate": 4.8887228357562e-05, "loss": 0.0598, "step": 3440 }, { "epoch": 0.07, "learning_rate": 4.888658139730476e-05, "loss": 0.0467, "step": 3442 }, { "epoch": 0.07, "learning_rate": 4.888593443704754e-05, "loss": 0.1091, "step": 3444 }, { "epoch": 0.07, "learning_rate": 4.88852874767903e-05, "loss": 0.1271, "step": 3446 }, { "epoch": 0.07, "learning_rate": 4.888464051653307e-05, "loss": 0.0737, "step": 3448 }, { "epoch": 0.07, "learning_rate": 4.888399355627584e-05, "loss": 0.0541, "step": 3450 }, { "epoch": 0.07, "learning_rate": 4.888334659601861e-05, "loss": 0.0659, "step": 3452 }, { "epoch": 0.07, "learning_rate": 4.888269963576138e-05, "loss": 0.1166, "step": 3454 }, { "epoch": 0.07, "learning_rate": 4.888205267550415e-05, "loss": 0.036, "step": 3456 }, { "epoch": 0.07, "learning_rate": 4.8881405715246916e-05, "loss": 0.0464, "step": 3458 }, { "epoch": 0.07, "learning_rate": 4.888075875498968e-05, "loss": 0.051, "step": 3460 }, { "epoch": 0.07, "learning_rate": 4.8880111794732454e-05, "loss": 0.0666, "step": 3462 }, { "epoch": 0.07, "learning_rate": 4.8879464834475216e-05, "loss": 0.1168, "step": 3464 }, { "epoch": 0.07, "learning_rate": 4.887881787421799e-05, "loss": 0.0612, "step": 3466 }, { "epoch": 0.07, "learning_rate": 4.887817091396076e-05, "loss": 0.0752, "step": 3468 }, { "epoch": 0.07, "learning_rate": 4.8877523953703524e-05, "loss": 0.0869, "step": 3470 }, { "epoch": 0.07, "learning_rate": 4.88768769934463e-05, "loss": 0.0522, "step": 3472 }, { "epoch": 0.07, "learning_rate": 4.887623003318906e-05, "loss": 0.0681, "step": 3474 }, { "epoch": 0.07, "learning_rate": 4.887558307293183e-05, "loss": 0.0767, "step": 3476 }, { "epoch": 0.07, "learning_rate": 4.88749361126746e-05, "loss": 0.0863, "step": 3478 }, { "epoch": 0.07, "learning_rate": 4.887428915241737e-05, "loss": 0.0589, "step": 3480 }, { "epoch": 0.07, "learning_rate": 4.887364219216014e-05, "loss": 0.0627, "step": 3482 }, { "epoch": 0.07, "learning_rate": 4.887299523190291e-05, "loss": 0.0547, "step": 3484 }, { "epoch": 0.07, "learning_rate": 4.8872348271645676e-05, "loss": 0.078, "step": 3486 }, { "epoch": 0.07, "learning_rate": 4.8871701311388445e-05, "loss": 0.0404, "step": 3488 }, { "epoch": 0.07, "learning_rate": 4.8871054351131214e-05, "loss": 0.086, "step": 3490 }, { "epoch": 0.07, "learning_rate": 4.887040739087398e-05, "loss": 0.071, "step": 3492 }, { "epoch": 0.07, "learning_rate": 4.886976043061675e-05, "loss": 0.0659, "step": 3494 }, { "epoch": 0.07, "learning_rate": 4.8869113470359515e-05, "loss": 0.1024, "step": 3496 }, { "epoch": 0.07, "learning_rate": 4.8868466510102284e-05, "loss": 0.0862, "step": 3498 }, { "epoch": 0.07, "learning_rate": 4.886781954984505e-05, "loss": 0.0686, "step": 3500 }, { "epoch": 0.07, "learning_rate": 4.886717258958782e-05, "loss": 0.066, "step": 3502 }, { "epoch": 0.07, "learning_rate": 4.88665256293306e-05, "loss": 0.0626, "step": 3504 }, { "epoch": 0.07, "learning_rate": 4.886587866907336e-05, "loss": 0.0969, "step": 3506 }, { "epoch": 0.07, "learning_rate": 4.886523170881613e-05, "loss": 0.0708, "step": 3508 }, { "epoch": 0.07, "learning_rate": 4.88645847485589e-05, "loss": 0.0837, "step": 3510 }, { "epoch": 0.07, "learning_rate": 4.886393778830167e-05, "loss": 0.04, "step": 3512 }, { "epoch": 0.07, "learning_rate": 4.8863290828044437e-05, "loss": 0.0646, "step": 3514 }, { "epoch": 0.07, "learning_rate": 4.8862643867787206e-05, "loss": 0.1399, "step": 3516 }, { "epoch": 0.07, "learning_rate": 4.8861996907529975e-05, "loss": 0.0661, "step": 3518 }, { "epoch": 0.07, "learning_rate": 4.886134994727274e-05, "loss": 0.0743, "step": 3520 }, { "epoch": 0.07, "learning_rate": 4.886070298701551e-05, "loss": 0.0355, "step": 3522 }, { "epoch": 0.07, "learning_rate": 4.8860056026758275e-05, "loss": 0.089, "step": 3524 }, { "epoch": 0.07, "learning_rate": 4.885940906650105e-05, "loss": 0.0631, "step": 3526 }, { "epoch": 0.07, "learning_rate": 4.8858762106243813e-05, "loss": 0.0947, "step": 3528 }, { "epoch": 0.07, "learning_rate": 4.885811514598658e-05, "loss": 0.0418, "step": 3530 }, { "epoch": 0.07, "learning_rate": 4.885746818572935e-05, "loss": 0.0615, "step": 3532 }, { "epoch": 0.07, "learning_rate": 4.885682122547212e-05, "loss": 0.0982, "step": 3534 }, { "epoch": 0.07, "learning_rate": 4.885617426521489e-05, "loss": 0.0821, "step": 3536 }, { "epoch": 0.07, "learning_rate": 4.885552730495766e-05, "loss": 0.0724, "step": 3538 }, { "epoch": 0.07, "learning_rate": 4.885488034470043e-05, "loss": 0.048, "step": 3540 }, { "epoch": 0.07, "learning_rate": 4.88542333844432e-05, "loss": 0.0813, "step": 3542 }, { "epoch": 0.07, "learning_rate": 4.8853586424185966e-05, "loss": 0.0901, "step": 3544 }, { "epoch": 0.07, "learning_rate": 4.8852939463928735e-05, "loss": 0.0977, "step": 3546 }, { "epoch": 0.07, "learning_rate": 4.8852292503671504e-05, "loss": 0.0705, "step": 3548 }, { "epoch": 0.07, "learning_rate": 4.885164554341427e-05, "loss": 0.1187, "step": 3550 }, { "epoch": 0.07, "learning_rate": 4.8850998583157036e-05, "loss": 0.0598, "step": 3552 }, { "epoch": 0.07, "learning_rate": 4.885035162289981e-05, "loss": 0.0406, "step": 3554 }, { "epoch": 0.07, "learning_rate": 4.8849704662642574e-05, "loss": 0.0461, "step": 3556 }, { "epoch": 0.07, "learning_rate": 4.884905770238534e-05, "loss": 0.0543, "step": 3558 }, { "epoch": 0.07, "learning_rate": 4.884841074212811e-05, "loss": 0.0602, "step": 3560 }, { "epoch": 0.07, "learning_rate": 4.884776378187088e-05, "loss": 0.0657, "step": 3562 }, { "epoch": 0.07, "learning_rate": 4.884711682161365e-05, "loss": 0.0863, "step": 3564 }, { "epoch": 0.07, "learning_rate": 4.884646986135642e-05, "loss": 0.0399, "step": 3566 }, { "epoch": 0.07, "learning_rate": 4.884582290109919e-05, "loss": 0.0628, "step": 3568 }, { "epoch": 0.07, "learning_rate": 4.884517594084196e-05, "loss": 0.1154, "step": 3570 }, { "epoch": 0.07, "learning_rate": 4.8844528980584726e-05, "loss": 0.0652, "step": 3572 }, { "epoch": 0.07, "learning_rate": 4.884388202032749e-05, "loss": 0.0392, "step": 3574 }, { "epoch": 0.07, "learning_rate": 4.8843235060070265e-05, "loss": 0.0537, "step": 3576 }, { "epoch": 0.07, "learning_rate": 4.884258809981303e-05, "loss": 0.0707, "step": 3578 }, { "epoch": 0.07, "learning_rate": 4.8841941139555796e-05, "loss": 0.0915, "step": 3580 }, { "epoch": 0.07, "learning_rate": 4.884129417929857e-05, "loss": 0.0672, "step": 3582 }, { "epoch": 0.07, "learning_rate": 4.8840647219041334e-05, "loss": 0.097, "step": 3584 }, { "epoch": 0.07, "learning_rate": 4.884000025878411e-05, "loss": 0.0644, "step": 3586 }, { "epoch": 0.07, "learning_rate": 4.883935329852687e-05, "loss": 0.0689, "step": 3588 }, { "epoch": 0.07, "learning_rate": 4.883870633826964e-05, "loss": 0.0353, "step": 3590 }, { "epoch": 0.07, "learning_rate": 4.883805937801241e-05, "loss": 0.0279, "step": 3592 }, { "epoch": 0.07, "learning_rate": 4.883741241775518e-05, "loss": 0.0824, "step": 3594 }, { "epoch": 0.07, "learning_rate": 4.883676545749795e-05, "loss": 0.0683, "step": 3596 }, { "epoch": 0.07, "learning_rate": 4.883611849724072e-05, "loss": 0.0534, "step": 3598 }, { "epoch": 0.07, "learning_rate": 4.883547153698349e-05, "loss": 0.062, "step": 3600 }, { "epoch": 0.07, "learning_rate": 4.883482457672625e-05, "loss": 0.0604, "step": 3602 }, { "epoch": 0.07, "learning_rate": 4.8834177616469025e-05, "loss": 0.0632, "step": 3604 }, { "epoch": 0.07, "learning_rate": 4.883353065621179e-05, "loss": 0.1142, "step": 3606 }, { "epoch": 0.07, "learning_rate": 4.883288369595456e-05, "loss": 0.0762, "step": 3608 }, { "epoch": 0.07, "learning_rate": 4.8832236735697326e-05, "loss": 0.078, "step": 3610 }, { "epoch": 0.07, "learning_rate": 4.8831589775440095e-05, "loss": 0.0751, "step": 3612 }, { "epoch": 0.07, "learning_rate": 4.883094281518287e-05, "loss": 0.0517, "step": 3614 }, { "epoch": 0.07, "learning_rate": 4.883029585492563e-05, "loss": 0.0453, "step": 3616 }, { "epoch": 0.07, "learning_rate": 4.88296488946684e-05, "loss": 0.0431, "step": 3618 }, { "epoch": 0.07, "learning_rate": 4.882900193441117e-05, "loss": 0.0501, "step": 3620 }, { "epoch": 0.07, "learning_rate": 4.882835497415394e-05, "loss": 0.1919, "step": 3622 }, { "epoch": 0.07, "learning_rate": 4.882770801389671e-05, "loss": 0.0713, "step": 3624 }, { "epoch": 0.07, "learning_rate": 4.882706105363948e-05, "loss": 0.0764, "step": 3626 }, { "epoch": 0.07, "learning_rate": 4.882641409338225e-05, "loss": 0.1123, "step": 3628 }, { "epoch": 0.07, "learning_rate": 4.8825767133125016e-05, "loss": 0.0777, "step": 3630 }, { "epoch": 0.07, "learning_rate": 4.8825120172867785e-05, "loss": 0.0943, "step": 3632 }, { "epoch": 0.07, "learning_rate": 4.882447321261055e-05, "loss": 0.1349, "step": 3634 }, { "epoch": 0.07, "learning_rate": 4.8823826252353324e-05, "loss": 0.052, "step": 3636 }, { "epoch": 0.07, "learning_rate": 4.8823179292096086e-05, "loss": 0.0648, "step": 3638 }, { "epoch": 0.07, "learning_rate": 4.8822532331838855e-05, "loss": 0.0434, "step": 3640 }, { "epoch": 0.07, "learning_rate": 4.8821885371581624e-05, "loss": 0.0555, "step": 3642 }, { "epoch": 0.07, "learning_rate": 4.882123841132439e-05, "loss": 0.0747, "step": 3644 }, { "epoch": 0.07, "learning_rate": 4.882059145106717e-05, "loss": 0.0322, "step": 3646 }, { "epoch": 0.07, "learning_rate": 4.881994449080993e-05, "loss": 0.0365, "step": 3648 }, { "epoch": 0.07, "learning_rate": 4.88192975305527e-05, "loss": 0.0716, "step": 3650 }, { "epoch": 0.07, "learning_rate": 4.881865057029547e-05, "loss": 0.1023, "step": 3652 }, { "epoch": 0.07, "learning_rate": 4.881800361003824e-05, "loss": 0.0762, "step": 3654 }, { "epoch": 0.07, "learning_rate": 4.8817356649781e-05, "loss": 0.1043, "step": 3656 }, { "epoch": 0.07, "learning_rate": 4.881670968952378e-05, "loss": 0.0489, "step": 3658 }, { "epoch": 0.07, "learning_rate": 4.8816062729266546e-05, "loss": 0.064, "step": 3660 }, { "epoch": 0.07, "learning_rate": 4.881541576900931e-05, "loss": 0.0903, "step": 3662 }, { "epoch": 0.07, "learning_rate": 4.8814768808752084e-05, "loss": 0.0653, "step": 3664 }, { "epoch": 0.07, "learning_rate": 4.8814121848494846e-05, "loss": 0.054, "step": 3666 }, { "epoch": 0.07, "learning_rate": 4.881347488823762e-05, "loss": 0.0805, "step": 3668 }, { "epoch": 0.07, "learning_rate": 4.8812827927980385e-05, "loss": 0.0663, "step": 3670 }, { "epoch": 0.07, "learning_rate": 4.8812180967723154e-05, "loss": 0.0329, "step": 3672 }, { "epoch": 0.07, "learning_rate": 4.881153400746592e-05, "loss": 0.0783, "step": 3674 }, { "epoch": 0.07, "learning_rate": 4.881088704720869e-05, "loss": 0.1239, "step": 3676 }, { "epoch": 0.07, "learning_rate": 4.881024008695146e-05, "loss": 0.0762, "step": 3678 }, { "epoch": 0.07, "learning_rate": 4.880959312669423e-05, "loss": 0.0784, "step": 3680 }, { "epoch": 0.07, "learning_rate": 4.8808946166437e-05, "loss": 0.0806, "step": 3682 }, { "epoch": 0.07, "learning_rate": 4.880829920617977e-05, "loss": 0.0385, "step": 3684 }, { "epoch": 0.07, "learning_rate": 4.880765224592254e-05, "loss": 0.0945, "step": 3686 }, { "epoch": 0.07, "learning_rate": 4.88070052856653e-05, "loss": 0.0502, "step": 3688 }, { "epoch": 0.07, "learning_rate": 4.8806358325408075e-05, "loss": 0.0636, "step": 3690 }, { "epoch": 0.07, "learning_rate": 4.8805711365150844e-05, "loss": 0.0916, "step": 3692 }, { "epoch": 0.07, "learning_rate": 4.880506440489361e-05, "loss": 0.0682, "step": 3694 }, { "epoch": 0.07, "learning_rate": 4.880441744463638e-05, "loss": 0.0825, "step": 3696 }, { "epoch": 0.07, "learning_rate": 4.8803770484379145e-05, "loss": 0.0399, "step": 3698 }, { "epoch": 0.07, "learning_rate": 4.8803123524121914e-05, "loss": 0.0623, "step": 3700 }, { "epoch": 0.07, "learning_rate": 4.880247656386468e-05, "loss": 0.0393, "step": 3702 }, { "epoch": 0.07, "learning_rate": 4.880182960360745e-05, "loss": 0.0671, "step": 3704 }, { "epoch": 0.07, "learning_rate": 4.880118264335022e-05, "loss": 0.0445, "step": 3706 }, { "epoch": 0.07, "learning_rate": 4.880053568309299e-05, "loss": 0.054, "step": 3708 }, { "epoch": 0.07, "learning_rate": 4.879988872283576e-05, "loss": 0.1206, "step": 3710 }, { "epoch": 0.07, "learning_rate": 4.879924176257853e-05, "loss": 0.0416, "step": 3712 }, { "epoch": 0.07, "learning_rate": 4.87985948023213e-05, "loss": 0.0806, "step": 3714 }, { "epoch": 0.07, "learning_rate": 4.879794784206406e-05, "loss": 0.0492, "step": 3716 }, { "epoch": 0.07, "learning_rate": 4.8797300881806836e-05, "loss": 0.047, "step": 3718 }, { "epoch": 0.07, "learning_rate": 4.87966539215496e-05, "loss": 0.0956, "step": 3720 }, { "epoch": 0.07, "learning_rate": 4.879600696129237e-05, "loss": 0.0766, "step": 3722 }, { "epoch": 0.07, "learning_rate": 4.879536000103514e-05, "loss": 0.0465, "step": 3724 }, { "epoch": 0.07, "learning_rate": 4.8794713040777905e-05, "loss": 0.0475, "step": 3726 }, { "epoch": 0.07, "learning_rate": 4.879406608052068e-05, "loss": 0.0593, "step": 3728 }, { "epoch": 0.07, "learning_rate": 4.8793419120263444e-05, "loss": 0.0375, "step": 3730 }, { "epoch": 0.07, "learning_rate": 4.879277216000621e-05, "loss": 0.1017, "step": 3732 }, { "epoch": 0.07, "learning_rate": 4.879212519974898e-05, "loss": 0.0569, "step": 3734 }, { "epoch": 0.07, "learning_rate": 4.879147823949175e-05, "loss": 0.1191, "step": 3736 }, { "epoch": 0.07, "learning_rate": 4.879083127923452e-05, "loss": 0.079, "step": 3738 }, { "epoch": 0.07, "learning_rate": 4.879018431897729e-05, "loss": 0.0531, "step": 3740 }, { "epoch": 0.07, "learning_rate": 4.878953735872006e-05, "loss": 0.0748, "step": 3742 }, { "epoch": 0.07, "learning_rate": 4.878889039846282e-05, "loss": 0.0511, "step": 3744 }, { "epoch": 0.07, "learning_rate": 4.8788243438205596e-05, "loss": 0.0903, "step": 3746 }, { "epoch": 0.07, "learning_rate": 4.878759647794836e-05, "loss": 0.046, "step": 3748 }, { "epoch": 0.07, "learning_rate": 4.8786949517691134e-05, "loss": 0.0734, "step": 3750 }, { "epoch": 0.07, "learning_rate": 4.87863025574339e-05, "loss": 0.0717, "step": 3752 }, { "epoch": 0.07, "learning_rate": 4.8785655597176666e-05, "loss": 0.0623, "step": 3754 }, { "epoch": 0.07, "learning_rate": 4.8785008636919435e-05, "loss": 0.0735, "step": 3756 }, { "epoch": 0.07, "learning_rate": 4.8784361676662204e-05, "loss": 0.0609, "step": 3758 }, { "epoch": 0.07, "learning_rate": 4.878371471640497e-05, "loss": 0.0839, "step": 3760 }, { "epoch": 0.07, "learning_rate": 4.878306775614774e-05, "loss": 0.0511, "step": 3762 }, { "epoch": 0.07, "learning_rate": 4.878242079589051e-05, "loss": 0.0577, "step": 3764 }, { "epoch": 0.07, "learning_rate": 4.878177383563328e-05, "loss": 0.0372, "step": 3766 }, { "epoch": 0.07, "learning_rate": 4.878112687537605e-05, "loss": 0.0787, "step": 3768 }, { "epoch": 0.07, "learning_rate": 4.878047991511882e-05, "loss": 0.067, "step": 3770 }, { "epoch": 0.07, "learning_rate": 4.877983295486159e-05, "loss": 0.063, "step": 3772 }, { "epoch": 0.07, "learning_rate": 4.8779185994604357e-05, "loss": 0.0591, "step": 3774 }, { "epoch": 0.07, "learning_rate": 4.877853903434712e-05, "loss": 0.0714, "step": 3776 }, { "epoch": 0.07, "learning_rate": 4.8777892074089895e-05, "loss": 0.0501, "step": 3778 }, { "epoch": 0.07, "learning_rate": 4.877724511383266e-05, "loss": 0.0803, "step": 3780 }, { "epoch": 0.07, "learning_rate": 4.8776598153575426e-05, "loss": 0.0677, "step": 3782 }, { "epoch": 0.07, "learning_rate": 4.8775951193318195e-05, "loss": 0.0435, "step": 3784 }, { "epoch": 0.07, "learning_rate": 4.8775304233060964e-05, "loss": 0.0432, "step": 3786 }, { "epoch": 0.07, "learning_rate": 4.8774657272803733e-05, "loss": 0.062, "step": 3788 }, { "epoch": 0.07, "learning_rate": 4.87740103125465e-05, "loss": 0.0675, "step": 3790 }, { "epoch": 0.07, "learning_rate": 4.877336335228927e-05, "loss": 0.0667, "step": 3792 }, { "epoch": 0.07, "learning_rate": 4.877271639203204e-05, "loss": 0.0419, "step": 3794 }, { "epoch": 0.07, "learning_rate": 4.877206943177481e-05, "loss": 0.0774, "step": 3796 }, { "epoch": 0.07, "learning_rate": 4.877142247151757e-05, "loss": 0.0455, "step": 3798 }, { "epoch": 0.07, "learning_rate": 4.877077551126035e-05, "loss": 0.1083, "step": 3800 }, { "epoch": 0.07, "learning_rate": 4.877012855100312e-05, "loss": 0.0672, "step": 3802 }, { "epoch": 0.07, "learning_rate": 4.876948159074588e-05, "loss": 0.0863, "step": 3804 }, { "epoch": 0.07, "learning_rate": 4.8768834630488655e-05, "loss": 0.0765, "step": 3806 }, { "epoch": 0.07, "learning_rate": 4.876818767023142e-05, "loss": 0.0557, "step": 3808 }, { "epoch": 0.07, "learning_rate": 4.876754070997419e-05, "loss": 0.1125, "step": 3810 }, { "epoch": 0.07, "learning_rate": 4.8766893749716956e-05, "loss": 0.0825, "step": 3812 }, { "epoch": 0.07, "learning_rate": 4.8766246789459725e-05, "loss": 0.0447, "step": 3814 }, { "epoch": 0.07, "learning_rate": 4.8765599829202494e-05, "loss": 0.0496, "step": 3816 }, { "epoch": 0.07, "learning_rate": 4.876495286894526e-05, "loss": 0.0386, "step": 3818 }, { "epoch": 0.07, "learning_rate": 4.876430590868803e-05, "loss": 0.0753, "step": 3820 }, { "epoch": 0.07, "learning_rate": 4.87636589484308e-05, "loss": 0.0673, "step": 3822 }, { "epoch": 0.07, "learning_rate": 4.876301198817357e-05, "loss": 0.0531, "step": 3824 }, { "epoch": 0.07, "learning_rate": 4.876236502791634e-05, "loss": 0.0396, "step": 3826 }, { "epoch": 0.07, "learning_rate": 4.876171806765911e-05, "loss": 0.0679, "step": 3828 }, { "epoch": 0.07, "learning_rate": 4.876107110740187e-05, "loss": 0.0733, "step": 3830 }, { "epoch": 0.07, "learning_rate": 4.8760424147144646e-05, "loss": 0.0429, "step": 3832 }, { "epoch": 0.07, "learning_rate": 4.875977718688741e-05, "loss": 0.0414, "step": 3834 }, { "epoch": 0.07, "learning_rate": 4.875913022663018e-05, "loss": 0.1077, "step": 3836 }, { "epoch": 0.07, "learning_rate": 4.8758483266372954e-05, "loss": 0.0806, "step": 3838 }, { "epoch": 0.07, "learning_rate": 4.8757836306115716e-05, "loss": 0.0505, "step": 3840 }, { "epoch": 0.07, "learning_rate": 4.8757189345858485e-05, "loss": 0.0752, "step": 3842 }, { "epoch": 0.07, "learning_rate": 4.8756542385601254e-05, "loss": 0.0616, "step": 3844 }, { "epoch": 0.07, "learning_rate": 4.875589542534402e-05, "loss": 0.0477, "step": 3846 }, { "epoch": 0.07, "learning_rate": 4.875524846508679e-05, "loss": 0.0535, "step": 3848 }, { "epoch": 0.07, "learning_rate": 4.875460150482956e-05, "loss": 0.0496, "step": 3850 }, { "epoch": 0.07, "learning_rate": 4.875395454457233e-05, "loss": 0.0524, "step": 3852 }, { "epoch": 0.07, "learning_rate": 4.87533075843151e-05, "loss": 0.0444, "step": 3854 }, { "epoch": 0.07, "learning_rate": 4.875266062405787e-05, "loss": 0.0557, "step": 3856 }, { "epoch": 0.07, "learning_rate": 4.875201366380063e-05, "loss": 0.0558, "step": 3858 }, { "epoch": 0.07, "learning_rate": 4.875136670354341e-05, "loss": 0.0659, "step": 3860 }, { "epoch": 0.07, "learning_rate": 4.875071974328617e-05, "loss": 0.0675, "step": 3862 }, { "epoch": 0.07, "learning_rate": 4.875007278302894e-05, "loss": 0.0692, "step": 3864 }, { "epoch": 0.08, "learning_rate": 4.874942582277171e-05, "loss": 0.0541, "step": 3866 }, { "epoch": 0.08, "learning_rate": 4.8748778862514476e-05, "loss": 0.0617, "step": 3868 }, { "epoch": 0.08, "learning_rate": 4.874813190225725e-05, "loss": 0.0655, "step": 3870 }, { "epoch": 0.08, "learning_rate": 4.8747484942000015e-05, "loss": 0.0388, "step": 3872 }, { "epoch": 0.08, "learning_rate": 4.8746837981742784e-05, "loss": 0.0654, "step": 3874 }, { "epoch": 0.08, "learning_rate": 4.874619102148555e-05, "loss": 0.0672, "step": 3876 }, { "epoch": 0.08, "learning_rate": 4.874554406122832e-05, "loss": 0.0635, "step": 3878 }, { "epoch": 0.08, "learning_rate": 4.8744897100971084e-05, "loss": 0.0363, "step": 3880 }, { "epoch": 0.08, "learning_rate": 4.874425014071386e-05, "loss": 0.0521, "step": 3882 }, { "epoch": 0.08, "learning_rate": 4.874360318045663e-05, "loss": 0.0586, "step": 3884 }, { "epoch": 0.08, "learning_rate": 4.874295622019939e-05, "loss": 0.0578, "step": 3886 }, { "epoch": 0.08, "learning_rate": 4.874230925994217e-05, "loss": 0.111, "step": 3888 }, { "epoch": 0.08, "learning_rate": 4.874166229968493e-05, "loss": 0.0542, "step": 3890 }, { "epoch": 0.08, "learning_rate": 4.8741015339427705e-05, "loss": 0.0344, "step": 3892 }, { "epoch": 0.08, "learning_rate": 4.874036837917047e-05, "loss": 0.085, "step": 3894 }, { "epoch": 0.08, "learning_rate": 4.873972141891324e-05, "loss": 0.0304, "step": 3896 }, { "epoch": 0.08, "learning_rate": 4.8739074458656006e-05, "loss": 0.0349, "step": 3898 }, { "epoch": 0.08, "learning_rate": 4.8738427498398775e-05, "loss": 0.0421, "step": 3900 }, { "epoch": 0.08, "learning_rate": 4.8737780538141544e-05, "loss": 0.0533, "step": 3902 }, { "epoch": 0.08, "learning_rate": 4.873713357788431e-05, "loss": 0.04, "step": 3904 }, { "epoch": 0.08, "learning_rate": 4.873648661762708e-05, "loss": 0.0708, "step": 3906 }, { "epoch": 0.08, "learning_rate": 4.873583965736985e-05, "loss": 0.0549, "step": 3908 }, { "epoch": 0.08, "learning_rate": 4.873519269711262e-05, "loss": 0.0503, "step": 3910 }, { "epoch": 0.08, "learning_rate": 4.873454573685538e-05, "loss": 0.1039, "step": 3912 }, { "epoch": 0.08, "learning_rate": 4.873389877659816e-05, "loss": 0.0659, "step": 3914 }, { "epoch": 0.08, "learning_rate": 4.873325181634093e-05, "loss": 0.056, "step": 3916 }, { "epoch": 0.08, "learning_rate": 4.873260485608369e-05, "loss": 0.071, "step": 3918 }, { "epoch": 0.08, "learning_rate": 4.8731957895826466e-05, "loss": 0.0769, "step": 3920 }, { "epoch": 0.08, "learning_rate": 4.873131093556923e-05, "loss": 0.0592, "step": 3922 }, { "epoch": 0.08, "learning_rate": 4.8730663975312e-05, "loss": 0.0574, "step": 3924 }, { "epoch": 0.08, "learning_rate": 4.8730017015054766e-05, "loss": 0.0305, "step": 3926 }, { "epoch": 0.08, "learning_rate": 4.8729370054797535e-05, "loss": 0.0443, "step": 3928 }, { "epoch": 0.08, "learning_rate": 4.8728723094540305e-05, "loss": 0.0713, "step": 3930 }, { "epoch": 0.08, "learning_rate": 4.8728076134283074e-05, "loss": 0.0893, "step": 3932 }, { "epoch": 0.08, "learning_rate": 4.872742917402584e-05, "loss": 0.0823, "step": 3934 }, { "epoch": 0.08, "learning_rate": 4.872678221376861e-05, "loss": 0.0462, "step": 3936 }, { "epoch": 0.08, "learning_rate": 4.872613525351138e-05, "loss": 0.055, "step": 3938 }, { "epoch": 0.08, "learning_rate": 4.872548829325414e-05, "loss": 0.0898, "step": 3940 }, { "epoch": 0.08, "learning_rate": 4.872484133299692e-05, "loss": 0.092, "step": 3942 }, { "epoch": 0.08, "learning_rate": 4.872419437273968e-05, "loss": 0.0333, "step": 3944 }, { "epoch": 0.08, "learning_rate": 4.872354741248245e-05, "loss": 0.0863, "step": 3946 }, { "epoch": 0.08, "learning_rate": 4.8722900452225226e-05, "loss": 0.0725, "step": 3948 }, { "epoch": 0.08, "learning_rate": 4.872225349196799e-05, "loss": 0.0794, "step": 3950 }, { "epoch": 0.08, "learning_rate": 4.8721606531710764e-05, "loss": 0.0396, "step": 3952 }, { "epoch": 0.08, "learning_rate": 4.872095957145353e-05, "loss": 0.0336, "step": 3954 }, { "epoch": 0.08, "learning_rate": 4.8720312611196296e-05, "loss": 0.0615, "step": 3956 }, { "epoch": 0.08, "learning_rate": 4.8719665650939065e-05, "loss": 0.0381, "step": 3958 }, { "epoch": 0.08, "learning_rate": 4.8719018690681834e-05, "loss": 0.0276, "step": 3960 }, { "epoch": 0.08, "learning_rate": 4.87183717304246e-05, "loss": 0.1082, "step": 3962 }, { "epoch": 0.08, "learning_rate": 4.871772477016737e-05, "loss": 0.0525, "step": 3964 }, { "epoch": 0.08, "learning_rate": 4.871707780991014e-05, "loss": 0.0569, "step": 3966 }, { "epoch": 0.08, "learning_rate": 4.871643084965291e-05, "loss": 0.0536, "step": 3968 }, { "epoch": 0.08, "learning_rate": 4.871578388939568e-05, "loss": 0.081, "step": 3970 }, { "epoch": 0.08, "learning_rate": 4.871513692913844e-05, "loss": 0.0602, "step": 3972 }, { "epoch": 0.08, "learning_rate": 4.871448996888122e-05, "loss": 0.0506, "step": 3974 }, { "epoch": 0.08, "learning_rate": 4.871384300862398e-05, "loss": 0.0802, "step": 3976 }, { "epoch": 0.08, "learning_rate": 4.871319604836675e-05, "loss": 0.0492, "step": 3978 }, { "epoch": 0.08, "learning_rate": 4.871254908810952e-05, "loss": 0.0443, "step": 3980 }, { "epoch": 0.08, "learning_rate": 4.871190212785229e-05, "loss": 0.0756, "step": 3982 }, { "epoch": 0.08, "learning_rate": 4.8711255167595056e-05, "loss": 0.0853, "step": 3984 }, { "epoch": 0.08, "learning_rate": 4.8710608207337825e-05, "loss": 0.0517, "step": 3986 }, { "epoch": 0.08, "learning_rate": 4.8709961247080594e-05, "loss": 0.0544, "step": 3988 }, { "epoch": 0.08, "learning_rate": 4.8709314286823363e-05, "loss": 0.1032, "step": 3990 }, { "epoch": 0.08, "learning_rate": 4.870866732656613e-05, "loss": 0.0498, "step": 3992 }, { "epoch": 0.08, "learning_rate": 4.87080203663089e-05, "loss": 0.0917, "step": 3994 }, { "epoch": 0.08, "learning_rate": 4.870737340605167e-05, "loss": 0.0464, "step": 3996 }, { "epoch": 0.08, "learning_rate": 4.870672644579444e-05, "loss": 0.0239, "step": 3998 }, { "epoch": 0.08, "learning_rate": 4.87060794855372e-05, "loss": 0.0531, "step": 4000 }, { "epoch": 0.08, "learning_rate": 4.870543252527998e-05, "loss": 0.044, "step": 4002 }, { "epoch": 0.08, "learning_rate": 4.870478556502274e-05, "loss": 0.0293, "step": 4004 }, { "epoch": 0.08, "learning_rate": 4.870413860476551e-05, "loss": 0.0476, "step": 4006 }, { "epoch": 0.08, "learning_rate": 4.870349164450828e-05, "loss": 0.0254, "step": 4008 }, { "epoch": 0.08, "learning_rate": 4.870284468425105e-05, "loss": 0.0905, "step": 4010 }, { "epoch": 0.08, "learning_rate": 4.870219772399382e-05, "loss": 0.0417, "step": 4012 }, { "epoch": 0.08, "learning_rate": 4.8701550763736586e-05, "loss": 0.0847, "step": 4014 }, { "epoch": 0.08, "learning_rate": 4.8700903803479355e-05, "loss": 0.0385, "step": 4016 }, { "epoch": 0.08, "learning_rate": 4.8700256843222124e-05, "loss": 0.0494, "step": 4018 }, { "epoch": 0.08, "learning_rate": 4.869960988296489e-05, "loss": 0.0584, "step": 4020 }, { "epoch": 0.08, "learning_rate": 4.8698962922707655e-05, "loss": 0.036, "step": 4022 }, { "epoch": 0.08, "learning_rate": 4.869831596245043e-05, "loss": 0.0761, "step": 4024 }, { "epoch": 0.08, "learning_rate": 4.86976690021932e-05, "loss": 0.0693, "step": 4026 }, { "epoch": 0.08, "learning_rate": 4.869702204193596e-05, "loss": 0.0576, "step": 4028 }, { "epoch": 0.08, "learning_rate": 4.869637508167874e-05, "loss": 0.0433, "step": 4030 }, { "epoch": 0.08, "learning_rate": 4.86957281214215e-05, "loss": 0.0397, "step": 4032 }, { "epoch": 0.08, "learning_rate": 4.8695081161164277e-05, "loss": 0.0643, "step": 4034 }, { "epoch": 0.08, "learning_rate": 4.869443420090704e-05, "loss": 0.0594, "step": 4036 }, { "epoch": 0.08, "learning_rate": 4.869378724064981e-05, "loss": 0.0608, "step": 4038 }, { "epoch": 0.08, "learning_rate": 4.869314028039258e-05, "loss": 0.0449, "step": 4040 }, { "epoch": 0.08, "learning_rate": 4.8692493320135346e-05, "loss": 0.0495, "step": 4042 }, { "epoch": 0.08, "learning_rate": 4.8691846359878115e-05, "loss": 0.0291, "step": 4044 }, { "epoch": 0.08, "learning_rate": 4.8691199399620884e-05, "loss": 0.0696, "step": 4046 }, { "epoch": 0.08, "learning_rate": 4.869055243936365e-05, "loss": 0.0404, "step": 4048 }, { "epoch": 0.08, "learning_rate": 4.868990547910642e-05, "loss": 0.0428, "step": 4050 }, { "epoch": 0.08, "learning_rate": 4.868925851884919e-05, "loss": 0.0697, "step": 4052 }, { "epoch": 0.08, "learning_rate": 4.8688611558591954e-05, "loss": 0.053, "step": 4054 }, { "epoch": 0.08, "learning_rate": 4.868796459833473e-05, "loss": 0.1039, "step": 4056 }, { "epoch": 0.08, "learning_rate": 4.868731763807749e-05, "loss": 0.0672, "step": 4058 }, { "epoch": 0.08, "learning_rate": 4.868667067782026e-05, "loss": 0.0482, "step": 4060 }, { "epoch": 0.08, "learning_rate": 4.868602371756304e-05, "loss": 0.0708, "step": 4062 }, { "epoch": 0.08, "learning_rate": 4.86853767573058e-05, "loss": 0.0402, "step": 4064 }, { "epoch": 0.08, "learning_rate": 4.868472979704857e-05, "loss": 0.051, "step": 4066 }, { "epoch": 0.08, "learning_rate": 4.868408283679134e-05, "loss": 0.0621, "step": 4068 }, { "epoch": 0.08, "learning_rate": 4.8683435876534107e-05, "loss": 0.0246, "step": 4070 }, { "epoch": 0.08, "learning_rate": 4.8682788916276876e-05, "loss": 0.0382, "step": 4072 }, { "epoch": 0.08, "learning_rate": 4.8682141956019645e-05, "loss": 0.0885, "step": 4074 }, { "epoch": 0.08, "learning_rate": 4.8681494995762414e-05, "loss": 0.0307, "step": 4076 }, { "epoch": 0.08, "learning_rate": 4.868084803550518e-05, "loss": 0.0666, "step": 4078 }, { "epoch": 0.08, "learning_rate": 4.868020107524795e-05, "loss": 0.047, "step": 4080 }, { "epoch": 0.08, "learning_rate": 4.8679554114990714e-05, "loss": 0.0631, "step": 4082 }, { "epoch": 0.08, "learning_rate": 4.867890715473349e-05, "loss": 0.4066, "step": 4084 }, { "epoch": 0.08, "learning_rate": 4.867826019447625e-05, "loss": 0.0724, "step": 4086 }, { "epoch": 0.08, "learning_rate": 4.867761323421902e-05, "loss": 0.0527, "step": 4088 }, { "epoch": 0.08, "learning_rate": 4.867696627396179e-05, "loss": 0.0597, "step": 4090 }, { "epoch": 0.08, "learning_rate": 4.867631931370456e-05, "loss": 0.0545, "step": 4092 }, { "epoch": 0.08, "learning_rate": 4.8675672353447336e-05, "loss": 0.0432, "step": 4094 }, { "epoch": 0.08, "learning_rate": 4.86750253931901e-05, "loss": 0.0749, "step": 4096 }, { "epoch": 0.08, "learning_rate": 4.867437843293287e-05, "loss": 0.0685, "step": 4098 }, { "epoch": 0.08, "learning_rate": 4.8673731472675636e-05, "loss": 0.0704, "step": 4100 }, { "epoch": 0.08, "learning_rate": 4.8673084512418405e-05, "loss": 0.0375, "step": 4102 }, { "epoch": 0.08, "learning_rate": 4.8672437552161174e-05, "loss": 0.0858, "step": 4104 }, { "epoch": 0.08, "learning_rate": 4.867179059190394e-05, "loss": 0.0389, "step": 4106 }, { "epoch": 0.08, "learning_rate": 4.867114363164671e-05, "loss": 0.0311, "step": 4108 }, { "epoch": 0.08, "learning_rate": 4.8670496671389475e-05, "loss": 0.0609, "step": 4110 }, { "epoch": 0.08, "learning_rate": 4.866984971113225e-05, "loss": 0.0814, "step": 4112 }, { "epoch": 0.08, "learning_rate": 4.866920275087501e-05, "loss": 0.0917, "step": 4114 }, { "epoch": 0.08, "learning_rate": 4.866855579061779e-05, "loss": 0.0313, "step": 4116 }, { "epoch": 0.08, "learning_rate": 4.866790883036055e-05, "loss": 0.0638, "step": 4118 }, { "epoch": 0.08, "learning_rate": 4.866726187010332e-05, "loss": 0.0556, "step": 4120 }, { "epoch": 0.08, "learning_rate": 4.866661490984609e-05, "loss": 0.0531, "step": 4122 }, { "epoch": 0.08, "learning_rate": 4.866596794958886e-05, "loss": 0.0461, "step": 4124 }, { "epoch": 0.08, "learning_rate": 4.866532098933163e-05, "loss": 0.0501, "step": 4126 }, { "epoch": 0.08, "learning_rate": 4.8664674029074396e-05, "loss": 0.0611, "step": 4128 }, { "epoch": 0.08, "learning_rate": 4.8664027068817166e-05, "loss": 0.102, "step": 4130 }, { "epoch": 0.08, "learning_rate": 4.8663380108559935e-05, "loss": 0.0466, "step": 4132 }, { "epoch": 0.08, "learning_rate": 4.8662733148302704e-05, "loss": 0.0599, "step": 4134 }, { "epoch": 0.08, "learning_rate": 4.8662086188045466e-05, "loss": 0.1337, "step": 4136 }, { "epoch": 0.08, "learning_rate": 4.866143922778824e-05, "loss": 0.0252, "step": 4138 }, { "epoch": 0.08, "learning_rate": 4.866079226753101e-05, "loss": 0.0575, "step": 4140 }, { "epoch": 0.08, "learning_rate": 4.866014530727377e-05, "loss": 0.0574, "step": 4142 }, { "epoch": 0.08, "learning_rate": 4.865949834701655e-05, "loss": 0.046, "step": 4144 }, { "epoch": 0.08, "learning_rate": 4.865885138675931e-05, "loss": 0.0273, "step": 4146 }, { "epoch": 0.08, "learning_rate": 4.865820442650208e-05, "loss": 0.0545, "step": 4148 }, { "epoch": 0.08, "learning_rate": 4.865755746624485e-05, "loss": 0.0728, "step": 4150 }, { "epoch": 0.08, "learning_rate": 4.865691050598762e-05, "loss": 0.0446, "step": 4152 }, { "epoch": 0.08, "learning_rate": 4.865626354573039e-05, "loss": 0.0356, "step": 4154 }, { "epoch": 0.08, "learning_rate": 4.865561658547316e-05, "loss": 0.056, "step": 4156 }, { "epoch": 0.08, "learning_rate": 4.8654969625215926e-05, "loss": 0.0459, "step": 4158 }, { "epoch": 0.08, "learning_rate": 4.8654322664958695e-05, "loss": 0.0886, "step": 4160 }, { "epoch": 0.08, "learning_rate": 4.8653675704701464e-05, "loss": 0.0432, "step": 4162 }, { "epoch": 0.08, "learning_rate": 4.8653028744444226e-05, "loss": 0.0823, "step": 4164 }, { "epoch": 0.08, "learning_rate": 4.8652381784187e-05, "loss": 0.0725, "step": 4166 }, { "epoch": 0.08, "learning_rate": 4.8651734823929765e-05, "loss": 0.0579, "step": 4168 }, { "epoch": 0.08, "learning_rate": 4.8651087863672534e-05, "loss": 0.0609, "step": 4170 }, { "epoch": 0.08, "learning_rate": 4.865044090341531e-05, "loss": 0.0537, "step": 4172 }, { "epoch": 0.08, "learning_rate": 4.864979394315807e-05, "loss": 0.0383, "step": 4174 }, { "epoch": 0.08, "learning_rate": 4.864914698290085e-05, "loss": 0.0358, "step": 4176 }, { "epoch": 0.08, "learning_rate": 4.864850002264361e-05, "loss": 0.0854, "step": 4178 }, { "epoch": 0.08, "learning_rate": 4.864785306238638e-05, "loss": 0.0272, "step": 4180 }, { "epoch": 0.08, "learning_rate": 4.864720610212915e-05, "loss": 0.0536, "step": 4182 }, { "epoch": 0.08, "learning_rate": 4.864655914187192e-05, "loss": 0.0656, "step": 4184 }, { "epoch": 0.08, "learning_rate": 4.8645912181614686e-05, "loss": 0.0696, "step": 4186 }, { "epoch": 0.08, "learning_rate": 4.8645265221357455e-05, "loss": 0.0532, "step": 4188 }, { "epoch": 0.08, "learning_rate": 4.8644618261100224e-05, "loss": 0.0484, "step": 4190 }, { "epoch": 0.08, "learning_rate": 4.8643971300842994e-05, "loss": 0.0437, "step": 4192 }, { "epoch": 0.08, "learning_rate": 4.864332434058576e-05, "loss": 0.0205, "step": 4194 }, { "epoch": 0.08, "learning_rate": 4.8642677380328525e-05, "loss": 0.0618, "step": 4196 }, { "epoch": 0.08, "learning_rate": 4.86420304200713e-05, "loss": 0.026, "step": 4198 }, { "epoch": 0.08, "learning_rate": 4.864138345981406e-05, "loss": 0.0612, "step": 4200 }, { "epoch": 0.08, "learning_rate": 4.864073649955683e-05, "loss": 0.0334, "step": 4202 }, { "epoch": 0.08, "learning_rate": 4.864008953929961e-05, "loss": 0.0633, "step": 4204 }, { "epoch": 0.08, "learning_rate": 4.863944257904237e-05, "loss": 0.0811, "step": 4206 }, { "epoch": 0.08, "learning_rate": 4.863879561878514e-05, "loss": 0.0579, "step": 4208 }, { "epoch": 0.08, "learning_rate": 4.863814865852791e-05, "loss": 0.0976, "step": 4210 }, { "epoch": 0.08, "learning_rate": 4.863750169827068e-05, "loss": 0.0589, "step": 4212 }, { "epoch": 0.08, "learning_rate": 4.863685473801345e-05, "loss": 0.0175, "step": 4214 }, { "epoch": 0.08, "learning_rate": 4.8636207777756216e-05, "loss": 0.0543, "step": 4216 }, { "epoch": 0.08, "learning_rate": 4.8635560817498985e-05, "loss": 0.0707, "step": 4218 }, { "epoch": 0.08, "learning_rate": 4.8634913857241754e-05, "loss": 0.0384, "step": 4220 }, { "epoch": 0.08, "learning_rate": 4.863426689698452e-05, "loss": 0.0794, "step": 4222 }, { "epoch": 0.08, "learning_rate": 4.8633619936727285e-05, "loss": 0.0369, "step": 4224 }, { "epoch": 0.08, "learning_rate": 4.863297297647006e-05, "loss": 0.0668, "step": 4226 }, { "epoch": 0.08, "learning_rate": 4.8632326016212824e-05, "loss": 0.0487, "step": 4228 }, { "epoch": 0.08, "learning_rate": 4.863167905595559e-05, "loss": 0.058, "step": 4230 }, { "epoch": 0.08, "learning_rate": 4.863103209569836e-05, "loss": 0.0673, "step": 4232 }, { "epoch": 0.08, "learning_rate": 4.863038513544113e-05, "loss": 0.0377, "step": 4234 }, { "epoch": 0.08, "learning_rate": 4.86297381751839e-05, "loss": 0.0424, "step": 4236 }, { "epoch": 0.08, "learning_rate": 4.862909121492667e-05, "loss": 0.0489, "step": 4238 }, { "epoch": 0.08, "learning_rate": 4.862844425466944e-05, "loss": 0.0415, "step": 4240 }, { "epoch": 0.08, "learning_rate": 4.862779729441221e-05, "loss": 0.0758, "step": 4242 }, { "epoch": 0.08, "learning_rate": 4.8627150334154976e-05, "loss": 0.0378, "step": 4244 }, { "epoch": 0.08, "learning_rate": 4.862650337389774e-05, "loss": 0.0497, "step": 4246 }, { "epoch": 0.08, "learning_rate": 4.8625856413640514e-05, "loss": 0.0794, "step": 4248 }, { "epoch": 0.08, "learning_rate": 4.8625209453383283e-05, "loss": 0.0446, "step": 4250 }, { "epoch": 0.08, "learning_rate": 4.8624562493126046e-05, "loss": 0.0377, "step": 4252 }, { "epoch": 0.08, "learning_rate": 4.862391553286882e-05, "loss": 0.1527, "step": 4254 }, { "epoch": 0.08, "learning_rate": 4.8623268572611584e-05, "loss": 0.0731, "step": 4256 }, { "epoch": 0.08, "learning_rate": 4.862262161235436e-05, "loss": 0.0756, "step": 4258 }, { "epoch": 0.08, "learning_rate": 4.862197465209712e-05, "loss": 0.0425, "step": 4260 }, { "epoch": 0.08, "learning_rate": 4.862132769183989e-05, "loss": 0.0773, "step": 4262 }, { "epoch": 0.08, "learning_rate": 4.862068073158266e-05, "loss": 0.0353, "step": 4264 }, { "epoch": 0.08, "learning_rate": 4.862003377132543e-05, "loss": 0.0661, "step": 4266 }, { "epoch": 0.08, "learning_rate": 4.86193868110682e-05, "loss": 0.0454, "step": 4268 }, { "epoch": 0.08, "learning_rate": 4.861873985081097e-05, "loss": 0.0759, "step": 4270 }, { "epoch": 0.08, "learning_rate": 4.861809289055374e-05, "loss": 0.0779, "step": 4272 }, { "epoch": 0.08, "learning_rate": 4.8617445930296506e-05, "loss": 0.0919, "step": 4274 }, { "epoch": 0.08, "learning_rate": 4.8616798970039275e-05, "loss": 0.0604, "step": 4276 }, { "epoch": 0.08, "learning_rate": 4.861615200978204e-05, "loss": 0.1958, "step": 4278 }, { "epoch": 0.08, "learning_rate": 4.861550504952481e-05, "loss": 0.0878, "step": 4280 }, { "epoch": 0.08, "learning_rate": 4.861485808926758e-05, "loss": 0.0489, "step": 4282 }, { "epoch": 0.08, "learning_rate": 4.8614211129010344e-05, "loss": 0.0551, "step": 4284 }, { "epoch": 0.08, "learning_rate": 4.861356416875312e-05, "loss": 0.0447, "step": 4286 }, { "epoch": 0.08, "learning_rate": 4.861291720849588e-05, "loss": 0.0351, "step": 4288 }, { "epoch": 0.08, "learning_rate": 4.861227024823865e-05, "loss": 0.0453, "step": 4290 }, { "epoch": 0.08, "learning_rate": 4.861162328798142e-05, "loss": 0.0439, "step": 4292 }, { "epoch": 0.08, "learning_rate": 4.861097632772419e-05, "loss": 0.0564, "step": 4294 }, { "epoch": 0.08, "learning_rate": 4.861032936746696e-05, "loss": 0.0523, "step": 4296 }, { "epoch": 0.08, "learning_rate": 4.860968240720973e-05, "loss": 0.0686, "step": 4298 }, { "epoch": 0.08, "learning_rate": 4.86090354469525e-05, "loss": 0.0603, "step": 4300 }, { "epoch": 0.08, "learning_rate": 4.8608388486695266e-05, "loss": 0.0383, "step": 4302 }, { "epoch": 0.08, "learning_rate": 4.8607741526438035e-05, "loss": 0.0544, "step": 4304 }, { "epoch": 0.08, "learning_rate": 4.86070945661808e-05, "loss": 0.0535, "step": 4306 }, { "epoch": 0.08, "learning_rate": 4.860644760592357e-05, "loss": 0.0957, "step": 4308 }, { "epoch": 0.08, "learning_rate": 4.8605800645666336e-05, "loss": 0.0385, "step": 4310 }, { "epoch": 0.08, "learning_rate": 4.8605153685409105e-05, "loss": 0.018, "step": 4312 }, { "epoch": 0.08, "learning_rate": 4.8604506725151874e-05, "loss": 0.0653, "step": 4314 }, { "epoch": 0.08, "learning_rate": 4.860385976489464e-05, "loss": 0.0332, "step": 4316 }, { "epoch": 0.08, "learning_rate": 4.860321280463742e-05, "loss": 0.0402, "step": 4318 }, { "epoch": 0.08, "learning_rate": 4.860256584438018e-05, "loss": 0.0252, "step": 4320 }, { "epoch": 0.08, "learning_rate": 4.860191888412295e-05, "loss": 0.0358, "step": 4322 }, { "epoch": 0.08, "learning_rate": 4.860127192386572e-05, "loss": 0.0387, "step": 4324 }, { "epoch": 0.08, "learning_rate": 4.860062496360849e-05, "loss": 0.0355, "step": 4326 }, { "epoch": 0.08, "learning_rate": 4.859997800335126e-05, "loss": 0.0388, "step": 4328 }, { "epoch": 0.08, "learning_rate": 4.8599331043094027e-05, "loss": 0.0422, "step": 4330 }, { "epoch": 0.08, "learning_rate": 4.8598684082836796e-05, "loss": 0.0494, "step": 4332 }, { "epoch": 0.08, "learning_rate": 4.8598037122579565e-05, "loss": 0.0456, "step": 4334 }, { "epoch": 0.08, "learning_rate": 4.8597390162322334e-05, "loss": 0.0341, "step": 4336 }, { "epoch": 0.08, "learning_rate": 4.8596743202065096e-05, "loss": 0.0449, "step": 4338 }, { "epoch": 0.08, "learning_rate": 4.859609624180787e-05, "loss": 0.0558, "step": 4340 }, { "epoch": 0.08, "learning_rate": 4.8595449281550634e-05, "loss": 0.0459, "step": 4342 }, { "epoch": 0.08, "learning_rate": 4.85948023212934e-05, "loss": 0.0636, "step": 4344 }, { "epoch": 0.08, "learning_rate": 4.859415536103617e-05, "loss": 0.0533, "step": 4346 }, { "epoch": 0.08, "learning_rate": 4.859350840077894e-05, "loss": 0.0659, "step": 4348 }, { "epoch": 0.08, "learning_rate": 4.859286144052171e-05, "loss": 0.0764, "step": 4350 }, { "epoch": 0.08, "learning_rate": 4.859221448026448e-05, "loss": 0.0516, "step": 4352 }, { "epoch": 0.08, "learning_rate": 4.859156752000725e-05, "loss": 0.0663, "step": 4354 }, { "epoch": 0.08, "learning_rate": 4.859092055975002e-05, "loss": 0.037, "step": 4356 }, { "epoch": 0.08, "learning_rate": 4.859027359949279e-05, "loss": 0.0432, "step": 4358 }, { "epoch": 0.08, "learning_rate": 4.858962663923555e-05, "loss": 0.0491, "step": 4360 }, { "epoch": 0.08, "learning_rate": 4.8588979678978325e-05, "loss": 0.0639, "step": 4362 }, { "epoch": 0.08, "learning_rate": 4.8588332718721094e-05, "loss": 0.0576, "step": 4364 }, { "epoch": 0.08, "learning_rate": 4.8587685758463856e-05, "loss": 0.0997, "step": 4366 }, { "epoch": 0.08, "learning_rate": 4.858703879820663e-05, "loss": 0.0693, "step": 4368 }, { "epoch": 0.08, "learning_rate": 4.8586391837949395e-05, "loss": 0.0406, "step": 4370 }, { "epoch": 0.08, "learning_rate": 4.8585744877692164e-05, "loss": 0.0369, "step": 4372 }, { "epoch": 0.08, "learning_rate": 4.858509791743493e-05, "loss": 0.0776, "step": 4374 }, { "epoch": 0.08, "learning_rate": 4.85844509571777e-05, "loss": 0.1104, "step": 4376 }, { "epoch": 0.08, "learning_rate": 4.858380399692047e-05, "loss": 0.079, "step": 4378 }, { "epoch": 0.09, "learning_rate": 4.858315703666324e-05, "loss": 0.0702, "step": 4380 }, { "epoch": 0.09, "learning_rate": 4.858251007640601e-05, "loss": 0.0423, "step": 4382 }, { "epoch": 0.09, "learning_rate": 4.858186311614878e-05, "loss": 0.0674, "step": 4384 }, { "epoch": 0.09, "learning_rate": 4.858121615589155e-05, "loss": 0.0555, "step": 4386 }, { "epoch": 0.09, "learning_rate": 4.858056919563431e-05, "loss": 0.0275, "step": 4388 }, { "epoch": 0.09, "learning_rate": 4.8579922235377085e-05, "loss": 0.0214, "step": 4390 }, { "epoch": 0.09, "learning_rate": 4.857927527511985e-05, "loss": 0.0407, "step": 4392 }, { "epoch": 0.09, "learning_rate": 4.857862831486262e-05, "loss": 0.0317, "step": 4394 }, { "epoch": 0.09, "learning_rate": 4.857798135460539e-05, "loss": 0.0898, "step": 4396 }, { "epoch": 0.09, "learning_rate": 4.8577334394348155e-05, "loss": 0.0608, "step": 4398 }, { "epoch": 0.09, "learning_rate": 4.857668743409093e-05, "loss": 0.04, "step": 4400 }, { "epoch": 0.09, "learning_rate": 4.857604047383369e-05, "loss": 0.0424, "step": 4402 }, { "epoch": 0.09, "learning_rate": 4.857539351357646e-05, "loss": 0.0636, "step": 4404 }, { "epoch": 0.09, "learning_rate": 4.857474655331923e-05, "loss": 0.0777, "step": 4406 }, { "epoch": 0.09, "learning_rate": 4.8574099593062e-05, "loss": 0.0391, "step": 4408 }, { "epoch": 0.09, "learning_rate": 4.857345263280477e-05, "loss": 0.0912, "step": 4410 }, { "epoch": 0.09, "learning_rate": 4.857280567254754e-05, "loss": 0.0845, "step": 4412 }, { "epoch": 0.09, "learning_rate": 4.857215871229031e-05, "loss": 0.0464, "step": 4414 }, { "epoch": 0.09, "learning_rate": 4.857151175203308e-05, "loss": 0.0505, "step": 4416 }, { "epoch": 0.09, "learning_rate": 4.8570864791775846e-05, "loss": 0.049, "step": 4418 }, { "epoch": 0.09, "learning_rate": 4.857021783151861e-05, "loss": 0.0464, "step": 4420 }, { "epoch": 0.09, "learning_rate": 4.8569570871261384e-05, "loss": 0.0704, "step": 4422 }, { "epoch": 0.09, "learning_rate": 4.8568923911004146e-05, "loss": 0.081, "step": 4424 }, { "epoch": 0.09, "learning_rate": 4.8568276950746915e-05, "loss": 0.0346, "step": 4426 }, { "epoch": 0.09, "learning_rate": 4.856762999048969e-05, "loss": 0.0445, "step": 4428 }, { "epoch": 0.09, "learning_rate": 4.8566983030232454e-05, "loss": 0.0548, "step": 4430 }, { "epoch": 0.09, "learning_rate": 4.856633606997522e-05, "loss": 0.0431, "step": 4432 }, { "epoch": 0.09, "learning_rate": 4.856568910971799e-05, "loss": 0.0378, "step": 4434 }, { "epoch": 0.09, "learning_rate": 4.856504214946076e-05, "loss": 0.0539, "step": 4436 }, { "epoch": 0.09, "learning_rate": 4.856439518920353e-05, "loss": 0.0443, "step": 4438 }, { "epoch": 0.09, "learning_rate": 4.85637482289463e-05, "loss": 0.0767, "step": 4440 }, { "epoch": 0.09, "learning_rate": 4.856310126868907e-05, "loss": 0.0591, "step": 4442 }, { "epoch": 0.09, "learning_rate": 4.856245430843184e-05, "loss": 0.0624, "step": 4444 }, { "epoch": 0.09, "learning_rate": 4.8561807348174606e-05, "loss": 0.0513, "step": 4446 }, { "epoch": 0.09, "learning_rate": 4.856116038791737e-05, "loss": 0.0433, "step": 4448 }, { "epoch": 0.09, "learning_rate": 4.8560513427660144e-05, "loss": 0.0454, "step": 4450 }, { "epoch": 0.09, "learning_rate": 4.855986646740291e-05, "loss": 0.0302, "step": 4452 }, { "epoch": 0.09, "learning_rate": 4.8559219507145676e-05, "loss": 0.0899, "step": 4454 }, { "epoch": 0.09, "learning_rate": 4.8558572546888445e-05, "loss": 0.0556, "step": 4456 }, { "epoch": 0.09, "learning_rate": 4.8557925586631214e-05, "loss": 0.0481, "step": 4458 }, { "epoch": 0.09, "learning_rate": 4.855727862637399e-05, "loss": 0.0609, "step": 4460 }, { "epoch": 0.09, "learning_rate": 4.855663166611675e-05, "loss": 0.147, "step": 4462 }, { "epoch": 0.09, "learning_rate": 4.855598470585952e-05, "loss": 0.0299, "step": 4464 }, { "epoch": 0.09, "learning_rate": 4.855533774560229e-05, "loss": 0.0728, "step": 4466 }, { "epoch": 0.09, "learning_rate": 4.855469078534506e-05, "loss": 0.1251, "step": 4468 }, { "epoch": 0.09, "learning_rate": 4.855404382508782e-05, "loss": 0.0714, "step": 4470 }, { "epoch": 0.09, "learning_rate": 4.85533968648306e-05, "loss": 0.0378, "step": 4472 }, { "epoch": 0.09, "learning_rate": 4.855274990457337e-05, "loss": 0.0425, "step": 4474 }, { "epoch": 0.09, "learning_rate": 4.8552102944316136e-05, "loss": 0.0598, "step": 4476 }, { "epoch": 0.09, "learning_rate": 4.8551455984058905e-05, "loss": 0.0679, "step": 4478 }, { "epoch": 0.09, "learning_rate": 4.855080902380167e-05, "loss": 0.0704, "step": 4480 }, { "epoch": 0.09, "learning_rate": 4.855016206354444e-05, "loss": 0.0341, "step": 4482 }, { "epoch": 0.09, "learning_rate": 4.8549515103287205e-05, "loss": 0.0265, "step": 4484 }, { "epoch": 0.09, "learning_rate": 4.8548868143029974e-05, "loss": 0.029, "step": 4486 }, { "epoch": 0.09, "learning_rate": 4.8548221182772744e-05, "loss": 0.054, "step": 4488 }, { "epoch": 0.09, "learning_rate": 4.854757422251551e-05, "loss": 0.056, "step": 4490 }, { "epoch": 0.09, "learning_rate": 4.854692726225828e-05, "loss": 0.0369, "step": 4492 }, { "epoch": 0.09, "learning_rate": 4.854628030200105e-05, "loss": 0.0519, "step": 4494 }, { "epoch": 0.09, "learning_rate": 4.854563334174382e-05, "loss": 0.0848, "step": 4496 }, { "epoch": 0.09, "learning_rate": 4.854498638148659e-05, "loss": 0.0182, "step": 4498 }, { "epoch": 0.09, "learning_rate": 4.854433942122936e-05, "loss": 0.07, "step": 4500 }, { "epoch": 0.09, "learning_rate": 4.854369246097212e-05, "loss": 0.033, "step": 4502 }, { "epoch": 0.09, "learning_rate": 4.8543045500714896e-05, "loss": 0.0248, "step": 4504 }, { "epoch": 0.09, "learning_rate": 4.8542398540457665e-05, "loss": 0.0428, "step": 4506 }, { "epoch": 0.09, "learning_rate": 4.854175158020043e-05, "loss": 0.0239, "step": 4508 }, { "epoch": 0.09, "learning_rate": 4.8541104619943203e-05, "loss": 0.0571, "step": 4510 }, { "epoch": 0.09, "learning_rate": 4.8540457659685966e-05, "loss": 0.0632, "step": 4512 }, { "epoch": 0.09, "learning_rate": 4.8539810699428735e-05, "loss": 0.058, "step": 4514 }, { "epoch": 0.09, "learning_rate": 4.8539163739171504e-05, "loss": 0.0161, "step": 4516 }, { "epoch": 0.09, "learning_rate": 4.853851677891427e-05, "loss": 0.0442, "step": 4518 }, { "epoch": 0.09, "learning_rate": 4.853786981865704e-05, "loss": 0.0443, "step": 4520 }, { "epoch": 0.09, "learning_rate": 4.853722285839981e-05, "loss": 0.0434, "step": 4522 }, { "epoch": 0.09, "learning_rate": 4.853657589814258e-05, "loss": 0.0219, "step": 4524 }, { "epoch": 0.09, "learning_rate": 4.853592893788535e-05, "loss": 0.0948, "step": 4526 }, { "epoch": 0.09, "learning_rate": 4.853528197762812e-05, "loss": 0.0499, "step": 4528 }, { "epoch": 0.09, "learning_rate": 4.853463501737088e-05, "loss": 0.09, "step": 4530 }, { "epoch": 0.09, "learning_rate": 4.8533988057113657e-05, "loss": 0.045, "step": 4532 }, { "epoch": 0.09, "learning_rate": 4.853334109685642e-05, "loss": 0.0544, "step": 4534 }, { "epoch": 0.09, "learning_rate": 4.853269413659919e-05, "loss": 0.0413, "step": 4536 }, { "epoch": 0.09, "learning_rate": 4.853204717634196e-05, "loss": 0.0502, "step": 4538 }, { "epoch": 0.09, "learning_rate": 4.8531400216084726e-05, "loss": 0.0186, "step": 4540 }, { "epoch": 0.09, "learning_rate": 4.85307532558275e-05, "loss": 0.0418, "step": 4542 }, { "epoch": 0.09, "learning_rate": 4.8530106295570264e-05, "loss": 0.0361, "step": 4544 }, { "epoch": 0.09, "learning_rate": 4.8529459335313033e-05, "loss": 0.0313, "step": 4546 }, { "epoch": 0.09, "learning_rate": 4.85288123750558e-05, "loss": 0.0583, "step": 4548 }, { "epoch": 0.09, "learning_rate": 4.852816541479857e-05, "loss": 0.0446, "step": 4550 }, { "epoch": 0.09, "learning_rate": 4.852751845454134e-05, "loss": 0.0742, "step": 4552 }, { "epoch": 0.09, "learning_rate": 4.852687149428411e-05, "loss": 0.0308, "step": 4554 }, { "epoch": 0.09, "learning_rate": 4.852622453402688e-05, "loss": 0.0391, "step": 4556 }, { "epoch": 0.09, "learning_rate": 4.852557757376965e-05, "loss": 0.0444, "step": 4558 }, { "epoch": 0.09, "learning_rate": 4.852493061351242e-05, "loss": 0.0447, "step": 4560 }, { "epoch": 0.09, "learning_rate": 4.852428365325518e-05, "loss": 0.054, "step": 4562 }, { "epoch": 0.09, "learning_rate": 4.8523636692997955e-05, "loss": 0.0206, "step": 4564 }, { "epoch": 0.09, "learning_rate": 4.852298973274072e-05, "loss": 0.0288, "step": 4566 }, { "epoch": 0.09, "learning_rate": 4.8522342772483487e-05, "loss": 0.0597, "step": 4568 }, { "epoch": 0.09, "learning_rate": 4.8521695812226256e-05, "loss": 0.0797, "step": 4570 }, { "epoch": 0.09, "learning_rate": 4.8521048851969025e-05, "loss": 0.0298, "step": 4572 }, { "epoch": 0.09, "learning_rate": 4.8520401891711794e-05, "loss": 0.0358, "step": 4574 }, { "epoch": 0.09, "learning_rate": 4.851975493145456e-05, "loss": 0.0449, "step": 4576 }, { "epoch": 0.09, "learning_rate": 4.851910797119733e-05, "loss": 0.0479, "step": 4578 }, { "epoch": 0.09, "learning_rate": 4.85184610109401e-05, "loss": 0.042, "step": 4580 }, { "epoch": 0.09, "learning_rate": 4.851781405068287e-05, "loss": 0.024, "step": 4582 }, { "epoch": 0.09, "learning_rate": 4.851716709042564e-05, "loss": 0.0345, "step": 4584 }, { "epoch": 0.09, "learning_rate": 4.851652013016841e-05, "loss": 0.0812, "step": 4586 }, { "epoch": 0.09, "learning_rate": 4.851587316991118e-05, "loss": 0.0352, "step": 4588 }, { "epoch": 0.09, "learning_rate": 4.851522620965394e-05, "loss": 0.0421, "step": 4590 }, { "epoch": 0.09, "learning_rate": 4.8514579249396716e-05, "loss": 0.0665, "step": 4592 }, { "epoch": 0.09, "learning_rate": 4.851393228913948e-05, "loss": 0.0564, "step": 4594 }, { "epoch": 0.09, "learning_rate": 4.851328532888225e-05, "loss": 0.0442, "step": 4596 }, { "epoch": 0.09, "learning_rate": 4.8512638368625016e-05, "loss": 0.0271, "step": 4598 }, { "epoch": 0.09, "learning_rate": 4.8511991408367785e-05, "loss": 0.0442, "step": 4600 }, { "epoch": 0.09, "learning_rate": 4.8511344448110554e-05, "loss": 0.0494, "step": 4602 }, { "epoch": 0.09, "learning_rate": 4.851069748785332e-05, "loss": 0.0318, "step": 4604 }, { "epoch": 0.09, "learning_rate": 4.851005052759609e-05, "loss": 0.0363, "step": 4606 }, { "epoch": 0.09, "learning_rate": 4.850940356733886e-05, "loss": 0.0482, "step": 4608 }, { "epoch": 0.09, "learning_rate": 4.850875660708163e-05, "loss": 0.0549, "step": 4610 }, { "epoch": 0.09, "learning_rate": 4.850810964682439e-05, "loss": 0.0395, "step": 4612 }, { "epoch": 0.09, "learning_rate": 4.850746268656717e-05, "loss": 0.0377, "step": 4614 }, { "epoch": 0.09, "learning_rate": 4.850681572630993e-05, "loss": 0.0472, "step": 4616 }, { "epoch": 0.09, "learning_rate": 4.850616876605271e-05, "loss": 0.076, "step": 4618 }, { "epoch": 0.09, "learning_rate": 4.8505521805795476e-05, "loss": 0.0847, "step": 4620 }, { "epoch": 0.09, "learning_rate": 4.850487484553824e-05, "loss": 0.034, "step": 4622 }, { "epoch": 0.09, "learning_rate": 4.8504227885281014e-05, "loss": 0.0598, "step": 4624 }, { "epoch": 0.09, "learning_rate": 4.8503580925023776e-05, "loss": 0.0852, "step": 4626 }, { "epoch": 0.09, "learning_rate": 4.8502933964766546e-05, "loss": 0.0391, "step": 4628 }, { "epoch": 0.09, "learning_rate": 4.8502287004509315e-05, "loss": 0.0653, "step": 4630 }, { "epoch": 0.09, "learning_rate": 4.8501640044252084e-05, "loss": 0.0333, "step": 4632 }, { "epoch": 0.09, "learning_rate": 4.850099308399485e-05, "loss": 0.0272, "step": 4634 }, { "epoch": 0.09, "learning_rate": 4.850034612373762e-05, "loss": 0.0706, "step": 4636 }, { "epoch": 0.09, "learning_rate": 4.849969916348039e-05, "loss": 0.0356, "step": 4638 }, { "epoch": 0.09, "learning_rate": 4.849905220322316e-05, "loss": 0.0613, "step": 4640 }, { "epoch": 0.09, "learning_rate": 4.849840524296593e-05, "loss": 0.026, "step": 4642 }, { "epoch": 0.09, "learning_rate": 4.849775828270869e-05, "loss": 0.064, "step": 4644 }, { "epoch": 0.09, "learning_rate": 4.849711132245147e-05, "loss": 0.0656, "step": 4646 }, { "epoch": 0.09, "learning_rate": 4.849646436219423e-05, "loss": 0.0304, "step": 4648 }, { "epoch": 0.09, "learning_rate": 4.8495817401937e-05, "loss": 0.0398, "step": 4650 }, { "epoch": 0.09, "learning_rate": 4.8495170441679775e-05, "loss": 0.0255, "step": 4652 }, { "epoch": 0.09, "learning_rate": 4.849452348142254e-05, "loss": 0.0465, "step": 4654 }, { "epoch": 0.09, "learning_rate": 4.8493876521165306e-05, "loss": 0.0362, "step": 4656 }, { "epoch": 0.09, "learning_rate": 4.8493229560908075e-05, "loss": 0.0393, "step": 4658 }, { "epoch": 0.09, "learning_rate": 4.8492582600650844e-05, "loss": 0.0527, "step": 4660 }, { "epoch": 0.09, "learning_rate": 4.849193564039361e-05, "loss": 0.0617, "step": 4662 }, { "epoch": 0.09, "learning_rate": 4.849128868013638e-05, "loss": 0.0319, "step": 4664 }, { "epoch": 0.09, "learning_rate": 4.849064171987915e-05, "loss": 0.0505, "step": 4666 }, { "epoch": 0.09, "learning_rate": 4.848999475962192e-05, "loss": 0.0479, "step": 4668 }, { "epoch": 0.09, "learning_rate": 4.848934779936469e-05, "loss": 0.0525, "step": 4670 }, { "epoch": 0.09, "learning_rate": 4.848870083910745e-05, "loss": 0.0355, "step": 4672 }, { "epoch": 0.09, "learning_rate": 4.848805387885023e-05, "loss": 0.0275, "step": 4674 }, { "epoch": 0.09, "learning_rate": 4.848740691859299e-05, "loss": 0.0541, "step": 4676 }, { "epoch": 0.09, "learning_rate": 4.848675995833576e-05, "loss": 0.0617, "step": 4678 }, { "epoch": 0.09, "learning_rate": 4.848611299807853e-05, "loss": 0.0729, "step": 4680 }, { "epoch": 0.09, "learning_rate": 4.84854660378213e-05, "loss": 0.0395, "step": 4682 }, { "epoch": 0.09, "learning_rate": 4.848481907756407e-05, "loss": 0.0939, "step": 4684 }, { "epoch": 0.09, "learning_rate": 4.8484172117306835e-05, "loss": 0.0377, "step": 4686 }, { "epoch": 0.09, "learning_rate": 4.8483525157049605e-05, "loss": 0.0289, "step": 4688 }, { "epoch": 0.09, "learning_rate": 4.8482878196792374e-05, "loss": 0.0854, "step": 4690 }, { "epoch": 0.09, "learning_rate": 4.848223123653514e-05, "loss": 0.061, "step": 4692 }, { "epoch": 0.09, "learning_rate": 4.8481584276277905e-05, "loss": 0.0522, "step": 4694 }, { "epoch": 0.09, "learning_rate": 4.848093731602068e-05, "loss": 0.0294, "step": 4696 }, { "epoch": 0.09, "learning_rate": 4.848029035576345e-05, "loss": 0.0324, "step": 4698 }, { "epoch": 0.09, "learning_rate": 4.847964339550622e-05, "loss": 0.0443, "step": 4700 }, { "epoch": 0.09, "learning_rate": 4.847899643524899e-05, "loss": 0.0312, "step": 4702 }, { "epoch": 0.09, "learning_rate": 4.847834947499175e-05, "loss": 0.037, "step": 4704 }, { "epoch": 0.09, "learning_rate": 4.8477702514734526e-05, "loss": 0.0401, "step": 4706 }, { "epoch": 0.09, "learning_rate": 4.847705555447729e-05, "loss": 0.0591, "step": 4708 }, { "epoch": 0.09, "learning_rate": 4.847640859422006e-05, "loss": 0.0186, "step": 4710 }, { "epoch": 0.09, "learning_rate": 4.847576163396283e-05, "loss": 0.0437, "step": 4712 }, { "epoch": 0.09, "learning_rate": 4.8475114673705596e-05, "loss": 0.1263, "step": 4714 }, { "epoch": 0.09, "learning_rate": 4.8474467713448365e-05, "loss": 0.0197, "step": 4716 }, { "epoch": 0.09, "learning_rate": 4.8473820753191134e-05, "loss": 0.0513, "step": 4718 }, { "epoch": 0.09, "learning_rate": 4.84731737929339e-05, "loss": 0.0864, "step": 4720 }, { "epoch": 0.09, "learning_rate": 4.847252683267667e-05, "loss": 0.0652, "step": 4722 }, { "epoch": 0.09, "learning_rate": 4.847187987241944e-05, "loss": 0.0279, "step": 4724 }, { "epoch": 0.09, "learning_rate": 4.8471232912162204e-05, "loss": 0.0469, "step": 4726 }, { "epoch": 0.09, "learning_rate": 4.847058595190498e-05, "loss": 0.0412, "step": 4728 }, { "epoch": 0.09, "learning_rate": 4.846993899164775e-05, "loss": 0.0669, "step": 4730 }, { "epoch": 0.09, "learning_rate": 4.846929203139051e-05, "loss": 0.0666, "step": 4732 }, { "epoch": 0.09, "learning_rate": 4.846864507113329e-05, "loss": 0.0267, "step": 4734 }, { "epoch": 0.09, "learning_rate": 4.846799811087605e-05, "loss": 0.0249, "step": 4736 }, { "epoch": 0.09, "learning_rate": 4.846735115061882e-05, "loss": 0.164, "step": 4738 }, { "epoch": 0.09, "learning_rate": 4.846670419036159e-05, "loss": 0.0586, "step": 4740 }, { "epoch": 0.09, "learning_rate": 4.8466057230104356e-05, "loss": 0.0741, "step": 4742 }, { "epoch": 0.09, "learning_rate": 4.8465410269847125e-05, "loss": 0.0452, "step": 4744 }, { "epoch": 0.09, "learning_rate": 4.8464763309589894e-05, "loss": 0.041, "step": 4746 }, { "epoch": 0.09, "learning_rate": 4.8464116349332664e-05, "loss": 0.0401, "step": 4748 }, { "epoch": 0.09, "learning_rate": 4.846346938907543e-05, "loss": 0.038, "step": 4750 }, { "epoch": 0.09, "learning_rate": 4.84628224288182e-05, "loss": 0.0527, "step": 4752 }, { "epoch": 0.09, "learning_rate": 4.8462175468560964e-05, "loss": 0.0464, "step": 4754 }, { "epoch": 0.09, "learning_rate": 4.846152850830374e-05, "loss": 0.0481, "step": 4756 }, { "epoch": 0.09, "learning_rate": 4.84608815480465e-05, "loss": 0.0344, "step": 4758 }, { "epoch": 0.09, "learning_rate": 4.846023458778928e-05, "loss": 0.0223, "step": 4760 }, { "epoch": 0.09, "learning_rate": 4.845958762753205e-05, "loss": 0.0406, "step": 4762 }, { "epoch": 0.09, "learning_rate": 4.845894066727481e-05, "loss": 0.0483, "step": 4764 }, { "epoch": 0.09, "learning_rate": 4.8458293707017585e-05, "loss": 0.0295, "step": 4766 }, { "epoch": 0.09, "learning_rate": 4.845764674676035e-05, "loss": 0.0414, "step": 4768 }, { "epoch": 0.09, "learning_rate": 4.845699978650312e-05, "loss": 0.0519, "step": 4770 }, { "epoch": 0.09, "learning_rate": 4.8456352826245886e-05, "loss": 0.0325, "step": 4772 }, { "epoch": 0.09, "learning_rate": 4.8455705865988655e-05, "loss": 0.0332, "step": 4774 }, { "epoch": 0.09, "learning_rate": 4.8455058905731424e-05, "loss": 0.0404, "step": 4776 }, { "epoch": 0.09, "learning_rate": 4.845441194547419e-05, "loss": 0.0737, "step": 4778 }, { "epoch": 0.09, "learning_rate": 4.845376498521696e-05, "loss": 0.0452, "step": 4780 }, { "epoch": 0.09, "learning_rate": 4.845311802495973e-05, "loss": 0.0384, "step": 4782 }, { "epoch": 0.09, "learning_rate": 4.84524710647025e-05, "loss": 0.0639, "step": 4784 }, { "epoch": 0.09, "learning_rate": 4.845182410444526e-05, "loss": 0.0273, "step": 4786 }, { "epoch": 0.09, "learning_rate": 4.845117714418804e-05, "loss": 0.0392, "step": 4788 }, { "epoch": 0.09, "learning_rate": 4.84505301839308e-05, "loss": 0.0544, "step": 4790 }, { "epoch": 0.09, "learning_rate": 4.844988322367357e-05, "loss": 0.0447, "step": 4792 }, { "epoch": 0.09, "learning_rate": 4.844923626341634e-05, "loss": 0.0448, "step": 4794 }, { "epoch": 0.09, "learning_rate": 4.844858930315911e-05, "loss": 0.0843, "step": 4796 }, { "epoch": 0.09, "learning_rate": 4.844794234290188e-05, "loss": 0.0646, "step": 4798 }, { "epoch": 0.09, "learning_rate": 4.8447295382644646e-05, "loss": 0.0722, "step": 4800 }, { "epoch": 0.09, "learning_rate": 4.8446648422387415e-05, "loss": 0.0918, "step": 4802 }, { "epoch": 0.09, "learning_rate": 4.8446001462130184e-05, "loss": 0.076, "step": 4804 }, { "epoch": 0.09, "learning_rate": 4.8445354501872953e-05, "loss": 0.0497, "step": 4806 }, { "epoch": 0.09, "learning_rate": 4.844470754161572e-05, "loss": 0.0292, "step": 4808 }, { "epoch": 0.09, "learning_rate": 4.844406058135849e-05, "loss": 0.0236, "step": 4810 }, { "epoch": 0.09, "learning_rate": 4.844341362110126e-05, "loss": 0.0311, "step": 4812 }, { "epoch": 0.09, "learning_rate": 4.844276666084402e-05, "loss": 0.0641, "step": 4814 }, { "epoch": 0.09, "learning_rate": 4.84421197005868e-05, "loss": 0.0994, "step": 4816 }, { "epoch": 0.09, "learning_rate": 4.844147274032956e-05, "loss": 0.0523, "step": 4818 }, { "epoch": 0.09, "learning_rate": 4.844082578007233e-05, "loss": 0.0489, "step": 4820 }, { "epoch": 0.09, "learning_rate": 4.84401788198151e-05, "loss": 0.0312, "step": 4822 }, { "epoch": 0.09, "learning_rate": 4.843953185955787e-05, "loss": 0.0772, "step": 4824 }, { "epoch": 0.09, "learning_rate": 4.843888489930064e-05, "loss": 0.0496, "step": 4826 }, { "epoch": 0.09, "learning_rate": 4.8438237939043407e-05, "loss": 0.0494, "step": 4828 }, { "epoch": 0.09, "learning_rate": 4.8437590978786176e-05, "loss": 0.0471, "step": 4830 }, { "epoch": 0.09, "learning_rate": 4.8436944018528945e-05, "loss": 0.063, "step": 4832 }, { "epoch": 0.09, "learning_rate": 4.8436297058271714e-05, "loss": 0.0295, "step": 4834 }, { "epoch": 0.09, "learning_rate": 4.8435650098014476e-05, "loss": 0.028, "step": 4836 }, { "epoch": 0.09, "learning_rate": 4.843500313775725e-05, "loss": 0.0214, "step": 4838 }, { "epoch": 0.09, "learning_rate": 4.8434356177500014e-05, "loss": 0.0305, "step": 4840 }, { "epoch": 0.09, "learning_rate": 4.843370921724279e-05, "loss": 0.0441, "step": 4842 }, { "epoch": 0.09, "learning_rate": 4.843306225698556e-05, "loss": 0.0524, "step": 4844 }, { "epoch": 0.09, "learning_rate": 4.843241529672832e-05, "loss": 0.0186, "step": 4846 }, { "epoch": 0.09, "learning_rate": 4.84317683364711e-05, "loss": 0.0184, "step": 4848 }, { "epoch": 0.09, "learning_rate": 4.843112137621386e-05, "loss": 0.0398, "step": 4850 }, { "epoch": 0.09, "learning_rate": 4.843047441595663e-05, "loss": 0.0569, "step": 4852 }, { "epoch": 0.09, "learning_rate": 4.84298274556994e-05, "loss": 0.0646, "step": 4854 }, { "epoch": 0.09, "learning_rate": 4.842918049544217e-05, "loss": 0.0413, "step": 4856 }, { "epoch": 0.09, "learning_rate": 4.8428533535184936e-05, "loss": 0.0401, "step": 4858 }, { "epoch": 0.09, "learning_rate": 4.8427886574927705e-05, "loss": 0.0409, "step": 4860 }, { "epoch": 0.09, "learning_rate": 4.8427239614670474e-05, "loss": 0.0693, "step": 4862 }, { "epoch": 0.09, "learning_rate": 4.842659265441324e-05, "loss": 0.042, "step": 4864 }, { "epoch": 0.09, "learning_rate": 4.842594569415601e-05, "loss": 0.0234, "step": 4866 }, { "epoch": 0.09, "learning_rate": 4.8425298733898775e-05, "loss": 0.0301, "step": 4868 }, { "epoch": 0.09, "learning_rate": 4.842465177364155e-05, "loss": 0.0394, "step": 4870 }, { "epoch": 0.09, "learning_rate": 4.842400481338431e-05, "loss": 0.0627, "step": 4872 }, { "epoch": 0.09, "learning_rate": 4.842335785312708e-05, "loss": 0.0476, "step": 4874 }, { "epoch": 0.09, "learning_rate": 4.842271089286986e-05, "loss": 0.0355, "step": 4876 }, { "epoch": 0.09, "learning_rate": 4.842206393261262e-05, "loss": 0.0626, "step": 4878 }, { "epoch": 0.09, "learning_rate": 4.842141697235539e-05, "loss": 0.0268, "step": 4880 }, { "epoch": 0.09, "learning_rate": 4.842077001209816e-05, "loss": 0.0146, "step": 4882 }, { "epoch": 0.09, "learning_rate": 4.842012305184093e-05, "loss": 0.0879, "step": 4884 }, { "epoch": 0.09, "learning_rate": 4.8419476091583696e-05, "loss": 0.0188, "step": 4886 }, { "epoch": 0.09, "learning_rate": 4.8418829131326466e-05, "loss": 0.0678, "step": 4888 }, { "epoch": 0.09, "learning_rate": 4.8418182171069235e-05, "loss": 0.0369, "step": 4890 }, { "epoch": 0.09, "learning_rate": 4.8417535210812004e-05, "loss": 0.0168, "step": 4892 }, { "epoch": 0.09, "learning_rate": 4.841688825055477e-05, "loss": 0.0177, "step": 4894 }, { "epoch": 0.1, "learning_rate": 4.8416241290297535e-05, "loss": 0.0449, "step": 4896 }, { "epoch": 0.1, "learning_rate": 4.841559433004031e-05, "loss": 0.0248, "step": 4898 }, { "epoch": 0.1, "learning_rate": 4.841494736978307e-05, "loss": 0.0361, "step": 4900 }, { "epoch": 0.1, "learning_rate": 4.841430040952585e-05, "loss": 0.0453, "step": 4902 }, { "epoch": 0.1, "learning_rate": 4.841365344926861e-05, "loss": 0.0786, "step": 4904 }, { "epoch": 0.1, "learning_rate": 4.841300648901138e-05, "loss": 0.0362, "step": 4906 }, { "epoch": 0.1, "learning_rate": 4.8412359528754156e-05, "loss": 0.0349, "step": 4908 }, { "epoch": 0.1, "learning_rate": 4.841171256849692e-05, "loss": 0.0414, "step": 4910 }, { "epoch": 0.1, "learning_rate": 4.841106560823969e-05, "loss": 0.0341, "step": 4912 }, { "epoch": 0.1, "learning_rate": 4.841041864798246e-05, "loss": 0.0364, "step": 4914 }, { "epoch": 0.1, "learning_rate": 4.8409771687725226e-05, "loss": 0.0358, "step": 4916 }, { "epoch": 0.1, "learning_rate": 4.840912472746799e-05, "loss": 0.0155, "step": 4918 }, { "epoch": 0.1, "learning_rate": 4.8408477767210764e-05, "loss": 0.0347, "step": 4920 }, { "epoch": 0.1, "learning_rate": 4.840783080695353e-05, "loss": 0.0129, "step": 4922 }, { "epoch": 0.1, "learning_rate": 4.84071838466963e-05, "loss": 0.0381, "step": 4924 }, { "epoch": 0.1, "learning_rate": 4.840653688643907e-05, "loss": 0.0467, "step": 4926 }, { "epoch": 0.1, "learning_rate": 4.8405889926181834e-05, "loss": 0.0477, "step": 4928 }, { "epoch": 0.1, "learning_rate": 4.840524296592461e-05, "loss": 0.0551, "step": 4930 }, { "epoch": 0.1, "learning_rate": 4.840459600566737e-05, "loss": 0.0545, "step": 4932 }, { "epoch": 0.1, "learning_rate": 4.840394904541014e-05, "loss": 0.0466, "step": 4934 }, { "epoch": 0.1, "learning_rate": 4.840330208515291e-05, "loss": 0.0269, "step": 4936 }, { "epoch": 0.1, "learning_rate": 4.840265512489568e-05, "loss": 0.0325, "step": 4938 }, { "epoch": 0.1, "learning_rate": 4.840200816463845e-05, "loss": 0.0431, "step": 4940 }, { "epoch": 0.1, "learning_rate": 4.840136120438122e-05, "loss": 0.0202, "step": 4942 }, { "epoch": 0.1, "learning_rate": 4.8400714244123986e-05, "loss": 0.0393, "step": 4944 }, { "epoch": 0.1, "learning_rate": 4.8400067283866755e-05, "loss": 0.0415, "step": 4946 }, { "epoch": 0.1, "learning_rate": 4.8399420323609525e-05, "loss": 0.1724, "step": 4948 }, { "epoch": 0.1, "learning_rate": 4.839877336335229e-05, "loss": 0.0428, "step": 4950 }, { "epoch": 0.1, "learning_rate": 4.839812640309506e-05, "loss": 0.0714, "step": 4952 }, { "epoch": 0.1, "learning_rate": 4.839747944283783e-05, "loss": 0.0368, "step": 4954 }, { "epoch": 0.1, "learning_rate": 4.8396832482580594e-05, "loss": 0.0337, "step": 4956 }, { "epoch": 0.1, "learning_rate": 4.839618552232337e-05, "loss": 0.0556, "step": 4958 }, { "epoch": 0.1, "learning_rate": 4.839553856206613e-05, "loss": 0.1047, "step": 4960 }, { "epoch": 0.1, "learning_rate": 4.83948916018089e-05, "loss": 0.0239, "step": 4962 }, { "epoch": 0.1, "learning_rate": 4.839424464155167e-05, "loss": 0.0903, "step": 4964 }, { "epoch": 0.1, "learning_rate": 4.839359768129444e-05, "loss": 0.0286, "step": 4966 }, { "epoch": 0.1, "learning_rate": 4.839295072103721e-05, "loss": 0.0524, "step": 4968 }, { "epoch": 0.1, "learning_rate": 4.839230376077998e-05, "loss": 0.0593, "step": 4970 }, { "epoch": 0.1, "learning_rate": 4.839165680052275e-05, "loss": 0.0346, "step": 4972 }, { "epoch": 0.1, "learning_rate": 4.8391009840265516e-05, "loss": 0.0593, "step": 4974 }, { "epoch": 0.1, "learning_rate": 4.8390362880008285e-05, "loss": 0.0781, "step": 4976 }, { "epoch": 0.1, "learning_rate": 4.838971591975105e-05, "loss": 0.0257, "step": 4978 }, { "epoch": 0.1, "learning_rate": 4.838906895949382e-05, "loss": 0.043, "step": 4980 }, { "epoch": 0.1, "learning_rate": 4.8388421999236585e-05, "loss": 0.0534, "step": 4982 }, { "epoch": 0.1, "learning_rate": 4.838777503897936e-05, "loss": 0.0395, "step": 4984 }, { "epoch": 0.1, "learning_rate": 4.838712807872213e-05, "loss": 0.038, "step": 4986 }, { "epoch": 0.1, "learning_rate": 4.838648111846489e-05, "loss": 0.0707, "step": 4988 }, { "epoch": 0.1, "learning_rate": 4.838583415820767e-05, "loss": 0.0186, "step": 4990 }, { "epoch": 0.1, "learning_rate": 4.838518719795043e-05, "loss": 0.0792, "step": 4992 }, { "epoch": 0.1, "learning_rate": 4.83845402376932e-05, "loss": 0.0148, "step": 4994 }, { "epoch": 0.1, "learning_rate": 4.838389327743597e-05, "loss": 0.0268, "step": 4996 }, { "epoch": 0.1, "learning_rate": 4.838324631717874e-05, "loss": 0.0261, "step": 4998 }, { "epoch": 0.1, "learning_rate": 4.838259935692151e-05, "loss": 0.0444, "step": 5000 }, { "epoch": 0.1, "learning_rate": 4.8381952396664276e-05, "loss": 0.0355, "step": 5002 }, { "epoch": 0.1, "learning_rate": 4.8381305436407045e-05, "loss": 0.0552, "step": 5004 }, { "epoch": 0.1, "learning_rate": 4.8380658476149814e-05, "loss": 0.046, "step": 5006 }, { "epoch": 0.1, "learning_rate": 4.8380011515892583e-05, "loss": 0.0267, "step": 5008 }, { "epoch": 0.1, "learning_rate": 4.8379364555635346e-05, "loss": 0.0293, "step": 5010 }, { "epoch": 0.1, "learning_rate": 4.837871759537812e-05, "loss": 0.0648, "step": 5012 }, { "epoch": 0.1, "learning_rate": 4.8378070635120884e-05, "loss": 0.0783, "step": 5014 }, { "epoch": 0.1, "learning_rate": 4.837742367486365e-05, "loss": 0.0236, "step": 5016 }, { "epoch": 0.1, "learning_rate": 4.837677671460642e-05, "loss": 0.0173, "step": 5018 }, { "epoch": 0.1, "learning_rate": 4.837612975434919e-05, "loss": 0.0277, "step": 5020 }, { "epoch": 0.1, "learning_rate": 4.837548279409196e-05, "loss": 0.0393, "step": 5022 }, { "epoch": 0.1, "learning_rate": 4.837483583383473e-05, "loss": 0.032, "step": 5024 }, { "epoch": 0.1, "learning_rate": 4.83741888735775e-05, "loss": 0.0568, "step": 5026 }, { "epoch": 0.1, "learning_rate": 4.837354191332027e-05, "loss": 0.0469, "step": 5028 }, { "epoch": 0.1, "learning_rate": 4.837289495306304e-05, "loss": 0.0416, "step": 5030 }, { "epoch": 0.1, "learning_rate": 4.8372247992805806e-05, "loss": 0.0414, "step": 5032 }, { "epoch": 0.1, "learning_rate": 4.8371601032548575e-05, "loss": 0.0331, "step": 5034 }, { "epoch": 0.1, "learning_rate": 4.8370954072291344e-05, "loss": 0.0273, "step": 5036 }, { "epoch": 0.1, "learning_rate": 4.8370307112034106e-05, "loss": 0.0384, "step": 5038 }, { "epoch": 0.1, "learning_rate": 4.836966015177688e-05, "loss": 0.0425, "step": 5040 }, { "epoch": 0.1, "learning_rate": 4.8369013191519644e-05, "loss": 0.0726, "step": 5042 }, { "epoch": 0.1, "learning_rate": 4.836836623126242e-05, "loss": 0.0523, "step": 5044 }, { "epoch": 0.1, "learning_rate": 4.836771927100518e-05, "loss": 0.0847, "step": 5046 }, { "epoch": 0.1, "learning_rate": 4.836707231074795e-05, "loss": 0.0253, "step": 5048 }, { "epoch": 0.1, "learning_rate": 4.836642535049072e-05, "loss": 0.0633, "step": 5050 }, { "epoch": 0.1, "learning_rate": 4.836577839023349e-05, "loss": 0.0331, "step": 5052 }, { "epoch": 0.1, "learning_rate": 4.836513142997626e-05, "loss": 0.0149, "step": 5054 }, { "epoch": 0.1, "learning_rate": 4.836448446971903e-05, "loss": 0.0708, "step": 5056 }, { "epoch": 0.1, "learning_rate": 4.83638375094618e-05, "loss": 0.0379, "step": 5058 }, { "epoch": 0.1, "learning_rate": 4.836319054920456e-05, "loss": 0.0353, "step": 5060 }, { "epoch": 0.1, "learning_rate": 4.8362543588947335e-05, "loss": 0.1067, "step": 5062 }, { "epoch": 0.1, "learning_rate": 4.8361896628690104e-05, "loss": 0.0355, "step": 5064 }, { "epoch": 0.1, "learning_rate": 4.836124966843287e-05, "loss": 0.0173, "step": 5066 }, { "epoch": 0.1, "learning_rate": 4.836060270817564e-05, "loss": 0.0936, "step": 5068 }, { "epoch": 0.1, "learning_rate": 4.8359955747918405e-05, "loss": 0.0425, "step": 5070 }, { "epoch": 0.1, "learning_rate": 4.835930878766118e-05, "loss": 0.074, "step": 5072 }, { "epoch": 0.1, "learning_rate": 4.835866182740394e-05, "loss": 0.0559, "step": 5074 }, { "epoch": 0.1, "learning_rate": 4.835801486714671e-05, "loss": 0.072, "step": 5076 }, { "epoch": 0.1, "learning_rate": 4.835736790688948e-05, "loss": 0.0481, "step": 5078 }, { "epoch": 0.1, "learning_rate": 4.835672094663225e-05, "loss": 0.0217, "step": 5080 }, { "epoch": 0.1, "learning_rate": 4.835607398637502e-05, "loss": 0.0256, "step": 5082 }, { "epoch": 0.1, "learning_rate": 4.835542702611779e-05, "loss": 0.02, "step": 5084 }, { "epoch": 0.1, "learning_rate": 4.835478006586056e-05, "loss": 0.034, "step": 5086 }, { "epoch": 0.1, "learning_rate": 4.8354133105603327e-05, "loss": 0.029, "step": 5088 }, { "epoch": 0.1, "learning_rate": 4.8353486145346096e-05, "loss": 0.0481, "step": 5090 }, { "epoch": 0.1, "learning_rate": 4.835283918508886e-05, "loss": 0.0458, "step": 5092 }, { "epoch": 0.1, "learning_rate": 4.8352192224831634e-05, "loss": 0.0328, "step": 5094 }, { "epoch": 0.1, "learning_rate": 4.8351545264574396e-05, "loss": 0.0836, "step": 5096 }, { "epoch": 0.1, "learning_rate": 4.8350898304317165e-05, "loss": 0.0384, "step": 5098 }, { "epoch": 0.1, "learning_rate": 4.835025134405994e-05, "loss": 0.0566, "step": 5100 }, { "epoch": 0.1, "learning_rate": 4.83496043838027e-05, "loss": 0.0264, "step": 5102 }, { "epoch": 0.1, "learning_rate": 4.834895742354547e-05, "loss": 0.043, "step": 5104 }, { "epoch": 0.1, "learning_rate": 4.834831046328824e-05, "loss": 0.0347, "step": 5106 }, { "epoch": 0.1, "learning_rate": 4.834766350303101e-05, "loss": 0.0258, "step": 5108 }, { "epoch": 0.1, "learning_rate": 4.834701654277378e-05, "loss": 0.0648, "step": 5110 }, { "epoch": 0.1, "learning_rate": 4.834636958251655e-05, "loss": 0.0981, "step": 5112 }, { "epoch": 0.1, "learning_rate": 4.834572262225932e-05, "loss": 0.0388, "step": 5114 }, { "epoch": 0.1, "learning_rate": 4.834507566200209e-05, "loss": 0.0515, "step": 5116 }, { "epoch": 0.1, "learning_rate": 4.8344428701744856e-05, "loss": 0.0249, "step": 5118 }, { "epoch": 0.1, "learning_rate": 4.834378174148762e-05, "loss": 0.0232, "step": 5120 }, { "epoch": 0.1, "learning_rate": 4.8343134781230394e-05, "loss": 0.0123, "step": 5122 }, { "epoch": 0.1, "learning_rate": 4.8342487820973156e-05, "loss": 0.026, "step": 5124 }, { "epoch": 0.1, "learning_rate": 4.834184086071593e-05, "loss": 0.0432, "step": 5126 }, { "epoch": 0.1, "learning_rate": 4.8341193900458695e-05, "loss": 0.0492, "step": 5128 }, { "epoch": 0.1, "learning_rate": 4.8340546940201464e-05, "loss": 0.0143, "step": 5130 }, { "epoch": 0.1, "learning_rate": 4.833989997994424e-05, "loss": 0.0297, "step": 5132 }, { "epoch": 0.1, "learning_rate": 4.8339253019687e-05, "loss": 0.0438, "step": 5134 }, { "epoch": 0.1, "learning_rate": 4.833860605942977e-05, "loss": 0.038, "step": 5136 }, { "epoch": 0.1, "learning_rate": 4.833795909917254e-05, "loss": 0.1027, "step": 5138 }, { "epoch": 0.1, "learning_rate": 4.833731213891531e-05, "loss": 0.0344, "step": 5140 }, { "epoch": 0.1, "learning_rate": 4.833666517865808e-05, "loss": 0.0369, "step": 5142 }, { "epoch": 0.1, "learning_rate": 4.833601821840085e-05, "loss": 0.0341, "step": 5144 }, { "epoch": 0.1, "learning_rate": 4.8335371258143616e-05, "loss": 0.021, "step": 5146 }, { "epoch": 0.1, "learning_rate": 4.8334724297886386e-05, "loss": 0.0147, "step": 5148 }, { "epoch": 0.1, "learning_rate": 4.8334077337629155e-05, "loss": 0.0331, "step": 5150 }, { "epoch": 0.1, "learning_rate": 4.833343037737192e-05, "loss": 0.0332, "step": 5152 }, { "epoch": 0.1, "learning_rate": 4.833278341711469e-05, "loss": 0.0439, "step": 5154 }, { "epoch": 0.1, "learning_rate": 4.8332136456857455e-05, "loss": 0.0359, "step": 5156 }, { "epoch": 0.1, "learning_rate": 4.8331489496600224e-05, "loss": 0.053, "step": 5158 }, { "epoch": 0.1, "learning_rate": 4.833084253634299e-05, "loss": 0.0372, "step": 5160 }, { "epoch": 0.1, "learning_rate": 4.833019557608576e-05, "loss": 0.0514, "step": 5162 }, { "epoch": 0.1, "learning_rate": 4.832954861582853e-05, "loss": 0.0809, "step": 5164 }, { "epoch": 0.1, "learning_rate": 4.83289016555713e-05, "loss": 0.0291, "step": 5166 }, { "epoch": 0.1, "learning_rate": 4.832825469531407e-05, "loss": 0.0763, "step": 5168 }, { "epoch": 0.1, "learning_rate": 4.832760773505684e-05, "loss": 0.0295, "step": 5170 }, { "epoch": 0.1, "learning_rate": 4.832696077479961e-05, "loss": 0.0277, "step": 5172 }, { "epoch": 0.1, "learning_rate": 4.832631381454237e-05, "loss": 0.0196, "step": 5174 }, { "epoch": 0.1, "learning_rate": 4.8325666854285146e-05, "loss": 0.021, "step": 5176 }, { "epoch": 0.1, "learning_rate": 4.8325019894027915e-05, "loss": 0.0263, "step": 5178 }, { "epoch": 0.1, "learning_rate": 4.832437293377068e-05, "loss": 0.0297, "step": 5180 }, { "epoch": 0.1, "learning_rate": 4.832372597351345e-05, "loss": 0.0499, "step": 5182 }, { "epoch": 0.1, "learning_rate": 4.8323079013256215e-05, "loss": 0.0352, "step": 5184 }, { "epoch": 0.1, "learning_rate": 4.832243205299899e-05, "loss": 0.0718, "step": 5186 }, { "epoch": 0.1, "learning_rate": 4.8321785092741754e-05, "loss": 0.037, "step": 5188 }, { "epoch": 0.1, "learning_rate": 4.832113813248452e-05, "loss": 0.0465, "step": 5190 }, { "epoch": 0.1, "learning_rate": 4.832049117222729e-05, "loss": 0.0362, "step": 5192 }, { "epoch": 0.1, "learning_rate": 4.831984421197006e-05, "loss": 0.0345, "step": 5194 }, { "epoch": 0.1, "learning_rate": 4.831919725171283e-05, "loss": 0.0309, "step": 5196 }, { "epoch": 0.1, "learning_rate": 4.83185502914556e-05, "loss": 0.0111, "step": 5198 }, { "epoch": 0.1, "learning_rate": 4.831790333119837e-05, "loss": 0.0304, "step": 5200 }, { "epoch": 0.1, "learning_rate": 4.831725637094113e-05, "loss": 0.0133, "step": 5202 }, { "epoch": 0.1, "learning_rate": 4.8316609410683906e-05, "loss": 0.0401, "step": 5204 }, { "epoch": 0.1, "learning_rate": 4.831596245042667e-05, "loss": 0.0251, "step": 5206 }, { "epoch": 0.1, "learning_rate": 4.8315315490169444e-05, "loss": 0.0437, "step": 5208 }, { "epoch": 0.1, "learning_rate": 4.8314668529912214e-05, "loss": 0.0509, "step": 5210 }, { "epoch": 0.1, "learning_rate": 4.8314021569654976e-05, "loss": 0.0208, "step": 5212 }, { "epoch": 0.1, "learning_rate": 4.831337460939775e-05, "loss": 0.0245, "step": 5214 }, { "epoch": 0.1, "learning_rate": 4.8312727649140514e-05, "loss": 0.0405, "step": 5216 }, { "epoch": 0.1, "learning_rate": 4.831208068888328e-05, "loss": 0.0356, "step": 5218 }, { "epoch": 0.1, "learning_rate": 4.831143372862605e-05, "loss": 0.0379, "step": 5220 }, { "epoch": 0.1, "learning_rate": 4.831078676836882e-05, "loss": 0.0231, "step": 5222 }, { "epoch": 0.1, "learning_rate": 4.831013980811159e-05, "loss": 0.0159, "step": 5224 }, { "epoch": 0.1, "learning_rate": 4.830949284785436e-05, "loss": 0.0363, "step": 5226 }, { "epoch": 0.1, "learning_rate": 4.830884588759713e-05, "loss": 0.0433, "step": 5228 }, { "epoch": 0.1, "learning_rate": 4.83081989273399e-05, "loss": 0.0325, "step": 5230 }, { "epoch": 0.1, "learning_rate": 4.830755196708267e-05, "loss": 0.0353, "step": 5232 }, { "epoch": 0.1, "learning_rate": 4.830690500682543e-05, "loss": 0.0364, "step": 5234 }, { "epoch": 0.1, "learning_rate": 4.8306258046568205e-05, "loss": 0.0792, "step": 5236 }, { "epoch": 0.1, "learning_rate": 4.830561108631097e-05, "loss": 0.0185, "step": 5238 }, { "epoch": 0.1, "learning_rate": 4.8304964126053736e-05, "loss": 0.0264, "step": 5240 }, { "epoch": 0.1, "learning_rate": 4.830431716579651e-05, "loss": 0.046, "step": 5242 }, { "epoch": 0.1, "learning_rate": 4.8303670205539274e-05, "loss": 0.0876, "step": 5244 }, { "epoch": 0.1, "learning_rate": 4.8303023245282044e-05, "loss": 0.0431, "step": 5246 }, { "epoch": 0.1, "learning_rate": 4.830237628502481e-05, "loss": 0.0534, "step": 5248 }, { "epoch": 0.1, "learning_rate": 4.830172932476758e-05, "loss": 0.0242, "step": 5250 }, { "epoch": 0.1, "learning_rate": 4.830108236451035e-05, "loss": 0.0294, "step": 5252 }, { "epoch": 0.1, "learning_rate": 4.830043540425312e-05, "loss": 0.0255, "step": 5254 }, { "epoch": 0.1, "learning_rate": 4.829978844399589e-05, "loss": 0.056, "step": 5256 }, { "epoch": 0.1, "learning_rate": 4.829914148373866e-05, "loss": 0.038, "step": 5258 }, { "epoch": 0.1, "learning_rate": 4.829849452348143e-05, "loss": 0.0353, "step": 5260 }, { "epoch": 0.1, "learning_rate": 4.829784756322419e-05, "loss": 0.0377, "step": 5262 }, { "epoch": 0.1, "learning_rate": 4.8297200602966965e-05, "loss": 0.0535, "step": 5264 }, { "epoch": 0.1, "learning_rate": 4.829655364270973e-05, "loss": 0.0366, "step": 5266 }, { "epoch": 0.1, "learning_rate": 4.8295906682452503e-05, "loss": 0.0204, "step": 5268 }, { "epoch": 0.1, "learning_rate": 4.8295259722195266e-05, "loss": 0.0684, "step": 5270 }, { "epoch": 0.1, "learning_rate": 4.8294612761938035e-05, "loss": 0.0678, "step": 5272 }, { "epoch": 0.1, "learning_rate": 4.8293965801680804e-05, "loss": 0.0194, "step": 5274 }, { "epoch": 0.1, "learning_rate": 4.829331884142357e-05, "loss": 0.0637, "step": 5276 }, { "epoch": 0.1, "learning_rate": 4.829267188116634e-05, "loss": 0.0408, "step": 5278 }, { "epoch": 0.1, "learning_rate": 4.829202492090911e-05, "loss": 0.0387, "step": 5280 }, { "epoch": 0.1, "learning_rate": 4.829137796065188e-05, "loss": 0.0156, "step": 5282 }, { "epoch": 0.1, "learning_rate": 4.829073100039464e-05, "loss": 0.0387, "step": 5284 }, { "epoch": 0.1, "learning_rate": 4.829008404013742e-05, "loss": 0.025, "step": 5286 }, { "epoch": 0.1, "learning_rate": 4.828943707988019e-05, "loss": 0.0248, "step": 5288 }, { "epoch": 0.1, "learning_rate": 4.828879011962296e-05, "loss": 0.0247, "step": 5290 }, { "epoch": 0.1, "learning_rate": 4.8288143159365726e-05, "loss": 0.0411, "step": 5292 }, { "epoch": 0.1, "learning_rate": 4.828749619910849e-05, "loss": 0.0615, "step": 5294 }, { "epoch": 0.1, "learning_rate": 4.8286849238851264e-05, "loss": 0.0514, "step": 5296 }, { "epoch": 0.1, "learning_rate": 4.8286202278594026e-05, "loss": 0.0279, "step": 5298 }, { "epoch": 0.1, "learning_rate": 4.8285555318336795e-05, "loss": 0.0377, "step": 5300 }, { "epoch": 0.1, "learning_rate": 4.8284908358079564e-05, "loss": 0.027, "step": 5302 }, { "epoch": 0.1, "learning_rate": 4.8284261397822333e-05, "loss": 0.018, "step": 5304 }, { "epoch": 0.1, "learning_rate": 4.82836144375651e-05, "loss": 0.0159, "step": 5306 }, { "epoch": 0.1, "learning_rate": 4.828296747730787e-05, "loss": 0.0119, "step": 5308 }, { "epoch": 0.1, "learning_rate": 4.828232051705064e-05, "loss": 0.032, "step": 5310 }, { "epoch": 0.1, "learning_rate": 4.828167355679341e-05, "loss": 0.0362, "step": 5312 }, { "epoch": 0.1, "learning_rate": 4.828102659653618e-05, "loss": 0.0239, "step": 5314 }, { "epoch": 0.1, "learning_rate": 4.828037963627894e-05, "loss": 0.0339, "step": 5316 }, { "epoch": 0.1, "learning_rate": 4.827973267602172e-05, "loss": 0.0445, "step": 5318 }, { "epoch": 0.1, "learning_rate": 4.827908571576448e-05, "loss": 0.058, "step": 5320 }, { "epoch": 0.1, "learning_rate": 4.827843875550725e-05, "loss": 0.0147, "step": 5322 }, { "epoch": 0.1, "learning_rate": 4.8277791795250024e-05, "loss": 0.0206, "step": 5324 }, { "epoch": 0.1, "learning_rate": 4.8277144834992787e-05, "loss": 0.0207, "step": 5326 }, { "epoch": 0.1, "learning_rate": 4.827649787473556e-05, "loss": 0.0315, "step": 5328 }, { "epoch": 0.1, "learning_rate": 4.8275850914478325e-05, "loss": 0.025, "step": 5330 }, { "epoch": 0.1, "learning_rate": 4.8275203954221094e-05, "loss": 0.0226, "step": 5332 }, { "epoch": 0.1, "learning_rate": 4.827455699396386e-05, "loss": 0.0351, "step": 5334 }, { "epoch": 0.1, "learning_rate": 4.827391003370663e-05, "loss": 0.0496, "step": 5336 }, { "epoch": 0.1, "learning_rate": 4.82732630734494e-05, "loss": 0.0235, "step": 5338 }, { "epoch": 0.1, "learning_rate": 4.827261611319217e-05, "loss": 0.0326, "step": 5340 }, { "epoch": 0.1, "learning_rate": 4.827196915293494e-05, "loss": 0.0294, "step": 5342 }, { "epoch": 0.1, "learning_rate": 4.82713221926777e-05, "loss": 0.0234, "step": 5344 }, { "epoch": 0.1, "learning_rate": 4.827067523242048e-05, "loss": 0.0265, "step": 5346 }, { "epoch": 0.1, "learning_rate": 4.827002827216324e-05, "loss": 0.0197, "step": 5348 }, { "epoch": 0.1, "learning_rate": 4.8269381311906016e-05, "loss": 0.0332, "step": 5350 }, { "epoch": 0.1, "learning_rate": 4.826873435164878e-05, "loss": 0.0271, "step": 5352 }, { "epoch": 0.1, "learning_rate": 4.826808739139155e-05, "loss": 0.0168, "step": 5354 }, { "epoch": 0.1, "learning_rate": 4.826744043113432e-05, "loss": 0.1713, "step": 5356 }, { "epoch": 0.1, "learning_rate": 4.8266793470877085e-05, "loss": 0.036, "step": 5358 }, { "epoch": 0.1, "learning_rate": 4.8266146510619854e-05, "loss": 0.0187, "step": 5360 }, { "epoch": 0.1, "learning_rate": 4.826549955036262e-05, "loss": 0.0163, "step": 5362 }, { "epoch": 0.1, "learning_rate": 4.826485259010539e-05, "loss": 0.0393, "step": 5364 }, { "epoch": 0.1, "learning_rate": 4.826420562984816e-05, "loss": 0.0465, "step": 5366 }, { "epoch": 0.1, "learning_rate": 4.826355866959093e-05, "loss": 0.0541, "step": 5368 }, { "epoch": 0.1, "learning_rate": 4.82629117093337e-05, "loss": 0.0249, "step": 5370 }, { "epoch": 0.1, "learning_rate": 4.826226474907647e-05, "loss": 0.0239, "step": 5372 }, { "epoch": 0.1, "learning_rate": 4.826161778881924e-05, "loss": 0.0241, "step": 5374 }, { "epoch": 0.1, "learning_rate": 4.8260970828562e-05, "loss": 0.0512, "step": 5376 }, { "epoch": 0.1, "learning_rate": 4.8260323868304776e-05, "loss": 0.0324, "step": 5378 }, { "epoch": 0.1, "learning_rate": 4.825967690804754e-05, "loss": 0.0464, "step": 5380 }, { "epoch": 0.1, "learning_rate": 4.825902994779031e-05, "loss": 0.0341, "step": 5382 }, { "epoch": 0.1, "learning_rate": 4.8258382987533076e-05, "loss": 0.0222, "step": 5384 }, { "epoch": 0.1, "learning_rate": 4.8257736027275846e-05, "loss": 0.0096, "step": 5386 }, { "epoch": 0.1, "learning_rate": 4.8257089067018615e-05, "loss": 0.0319, "step": 5388 }, { "epoch": 0.1, "learning_rate": 4.8256442106761384e-05, "loss": 0.0265, "step": 5390 }, { "epoch": 0.1, "learning_rate": 4.825579514650415e-05, "loss": 0.0081, "step": 5392 }, { "epoch": 0.1, "learning_rate": 4.825514818624692e-05, "loss": 0.0355, "step": 5394 }, { "epoch": 0.1, "learning_rate": 4.825450122598969e-05, "loss": 0.0243, "step": 5396 }, { "epoch": 0.1, "learning_rate": 4.825385426573245e-05, "loss": 0.044, "step": 5398 }, { "epoch": 0.1, "learning_rate": 4.825320730547523e-05, "loss": 0.0548, "step": 5400 }, { "epoch": 0.1, "learning_rate": 4.8252560345218e-05, "loss": 0.0418, "step": 5402 }, { "epoch": 0.1, "learning_rate": 4.825191338496076e-05, "loss": 0.0371, "step": 5404 }, { "epoch": 0.1, "learning_rate": 4.8251266424703536e-05, "loss": 0.0379, "step": 5406 }, { "epoch": 0.1, "learning_rate": 4.82506194644463e-05, "loss": 0.0501, "step": 5408 }, { "epoch": 0.11, "learning_rate": 4.8249972504189075e-05, "loss": 0.0253, "step": 5410 }, { "epoch": 0.11, "learning_rate": 4.824932554393184e-05, "loss": 0.0419, "step": 5412 }, { "epoch": 0.11, "learning_rate": 4.8248678583674606e-05, "loss": 0.039, "step": 5414 }, { "epoch": 0.11, "learning_rate": 4.8248031623417375e-05, "loss": 0.0712, "step": 5416 }, { "epoch": 0.11, "learning_rate": 4.8247384663160144e-05, "loss": 0.0641, "step": 5418 }, { "epoch": 0.11, "learning_rate": 4.824673770290291e-05, "loss": 0.0417, "step": 5420 }, { "epoch": 0.11, "learning_rate": 4.824609074264568e-05, "loss": 0.0685, "step": 5422 }, { "epoch": 0.11, "learning_rate": 4.824544378238845e-05, "loss": 0.021, "step": 5424 }, { "epoch": 0.11, "learning_rate": 4.8244796822131214e-05, "loss": 0.0169, "step": 5426 }, { "epoch": 0.11, "learning_rate": 4.824414986187399e-05, "loss": 0.0188, "step": 5428 }, { "epoch": 0.11, "learning_rate": 4.824350290161675e-05, "loss": 0.0466, "step": 5430 }, { "epoch": 0.11, "learning_rate": 4.824285594135953e-05, "loss": 0.0355, "step": 5432 }, { "epoch": 0.11, "learning_rate": 4.82422089811023e-05, "loss": 0.0417, "step": 5434 }, { "epoch": 0.11, "learning_rate": 4.824156202084506e-05, "loss": 0.0402, "step": 5436 }, { "epoch": 0.11, "learning_rate": 4.8240915060587835e-05, "loss": 0.0332, "step": 5438 }, { "epoch": 0.11, "learning_rate": 4.82402681003306e-05, "loss": 0.0708, "step": 5440 }, { "epoch": 0.11, "learning_rate": 4.8239621140073366e-05, "loss": 0.0165, "step": 5442 }, { "epoch": 0.11, "learning_rate": 4.8238974179816135e-05, "loss": 0.0261, "step": 5444 }, { "epoch": 0.11, "learning_rate": 4.8238327219558905e-05, "loss": 0.0287, "step": 5446 }, { "epoch": 0.11, "learning_rate": 4.8237680259301674e-05, "loss": 0.0507, "step": 5448 }, { "epoch": 0.11, "learning_rate": 4.823703329904444e-05, "loss": 0.0492, "step": 5450 }, { "epoch": 0.11, "learning_rate": 4.823638633878721e-05, "loss": 0.0312, "step": 5452 }, { "epoch": 0.11, "learning_rate": 4.823573937852998e-05, "loss": 0.0599, "step": 5454 }, { "epoch": 0.11, "learning_rate": 4.823509241827275e-05, "loss": 0.0389, "step": 5456 }, { "epoch": 0.11, "learning_rate": 4.823444545801551e-05, "loss": 0.0244, "step": 5458 }, { "epoch": 0.11, "learning_rate": 4.823379849775829e-05, "loss": 0.0257, "step": 5460 }, { "epoch": 0.11, "learning_rate": 4.823315153750105e-05, "loss": 0.0345, "step": 5462 }, { "epoch": 0.11, "learning_rate": 4.823250457724382e-05, "loss": 0.0382, "step": 5464 }, { "epoch": 0.11, "learning_rate": 4.8231857616986595e-05, "loss": 0.0488, "step": 5466 }, { "epoch": 0.11, "learning_rate": 4.823121065672936e-05, "loss": 0.0234, "step": 5468 }, { "epoch": 0.11, "learning_rate": 4.8230563696472134e-05, "loss": 0.0434, "step": 5470 }, { "epoch": 0.11, "learning_rate": 4.8229916736214896e-05, "loss": 0.0367, "step": 5472 }, { "epoch": 0.11, "learning_rate": 4.8229269775957665e-05, "loss": 0.0218, "step": 5474 }, { "epoch": 0.11, "learning_rate": 4.8228622815700434e-05, "loss": 0.0355, "step": 5476 }, { "epoch": 0.11, "learning_rate": 4.82279758554432e-05, "loss": 0.0332, "step": 5478 }, { "epoch": 0.11, "learning_rate": 4.822732889518597e-05, "loss": 0.0244, "step": 5480 }, { "epoch": 0.11, "learning_rate": 4.822668193492874e-05, "loss": 0.0528, "step": 5482 }, { "epoch": 0.11, "learning_rate": 4.822603497467151e-05, "loss": 0.0631, "step": 5484 }, { "epoch": 0.11, "learning_rate": 4.822538801441427e-05, "loss": 0.0445, "step": 5486 }, { "epoch": 0.11, "learning_rate": 4.822474105415705e-05, "loss": 0.0353, "step": 5488 }, { "epoch": 0.11, "learning_rate": 4.822409409389981e-05, "loss": 0.0198, "step": 5490 }, { "epoch": 0.11, "learning_rate": 4.822344713364259e-05, "loss": 0.02, "step": 5492 }, { "epoch": 0.11, "learning_rate": 4.822280017338535e-05, "loss": 0.024, "step": 5494 }, { "epoch": 0.11, "learning_rate": 4.822215321312812e-05, "loss": 0.0491, "step": 5496 }, { "epoch": 0.11, "learning_rate": 4.822150625287089e-05, "loss": 0.0463, "step": 5498 }, { "epoch": 0.11, "learning_rate": 4.8220859292613656e-05, "loss": 0.033, "step": 5500 }, { "epoch": 0.11, "learning_rate": 4.8220212332356425e-05, "loss": 0.0149, "step": 5502 }, { "epoch": 0.11, "learning_rate": 4.8219565372099194e-05, "loss": 0.0307, "step": 5504 }, { "epoch": 0.11, "learning_rate": 4.8218918411841964e-05, "loss": 0.0392, "step": 5506 }, { "epoch": 0.11, "learning_rate": 4.8218271451584726e-05, "loss": 0.0389, "step": 5508 }, { "epoch": 0.11, "learning_rate": 4.82176244913275e-05, "loss": 0.0355, "step": 5510 }, { "epoch": 0.11, "learning_rate": 4.821697753107027e-05, "loss": 0.0375, "step": 5512 }, { "epoch": 0.11, "learning_rate": 4.821633057081304e-05, "loss": 0.0783, "step": 5514 }, { "epoch": 0.11, "learning_rate": 4.821568361055581e-05, "loss": 0.0214, "step": 5516 }, { "epoch": 0.11, "learning_rate": 4.821503665029857e-05, "loss": 0.0299, "step": 5518 }, { "epoch": 0.11, "learning_rate": 4.821438969004135e-05, "loss": 0.0507, "step": 5520 }, { "epoch": 0.11, "learning_rate": 4.821374272978411e-05, "loss": 0.0977, "step": 5522 }, { "epoch": 0.11, "learning_rate": 4.821309576952688e-05, "loss": 0.045, "step": 5524 }, { "epoch": 0.11, "learning_rate": 4.821244880926965e-05, "loss": 0.0518, "step": 5526 }, { "epoch": 0.11, "learning_rate": 4.821180184901242e-05, "loss": 0.0208, "step": 5528 }, { "epoch": 0.11, "learning_rate": 4.8211154888755186e-05, "loss": 0.0184, "step": 5530 }, { "epoch": 0.11, "learning_rate": 4.8210507928497955e-05, "loss": 0.0586, "step": 5532 }, { "epoch": 0.11, "learning_rate": 4.8209860968240724e-05, "loss": 0.0377, "step": 5534 }, { "epoch": 0.11, "learning_rate": 4.820921400798349e-05, "loss": 0.0258, "step": 5536 }, { "epoch": 0.11, "learning_rate": 4.820856704772626e-05, "loss": 0.0208, "step": 5538 }, { "epoch": 0.11, "learning_rate": 4.8207920087469024e-05, "loss": 0.0405, "step": 5540 }, { "epoch": 0.11, "learning_rate": 4.82072731272118e-05, "loss": 0.0171, "step": 5542 }, { "epoch": 0.11, "learning_rate": 4.820662616695457e-05, "loss": 0.016, "step": 5544 }, { "epoch": 0.11, "learning_rate": 4.820597920669733e-05, "loss": 0.0426, "step": 5546 }, { "epoch": 0.11, "learning_rate": 4.820533224644011e-05, "loss": 0.0248, "step": 5548 }, { "epoch": 0.11, "learning_rate": 4.820468528618287e-05, "loss": 0.0108, "step": 5550 }, { "epoch": 0.11, "learning_rate": 4.8204038325925646e-05, "loss": 0.0117, "step": 5552 }, { "epoch": 0.11, "learning_rate": 4.820339136566841e-05, "loss": 0.0207, "step": 5554 }, { "epoch": 0.11, "learning_rate": 4.820274440541118e-05, "loss": 0.0201, "step": 5556 }, { "epoch": 0.11, "learning_rate": 4.8202097445153946e-05, "loss": 0.0283, "step": 5558 }, { "epoch": 0.11, "learning_rate": 4.8201450484896715e-05, "loss": 0.0131, "step": 5560 }, { "epoch": 0.11, "learning_rate": 4.8200803524639484e-05, "loss": 0.0397, "step": 5562 }, { "epoch": 0.11, "learning_rate": 4.8200156564382253e-05, "loss": 0.0483, "step": 5564 }, { "epoch": 0.11, "learning_rate": 4.819950960412502e-05, "loss": 0.0737, "step": 5566 }, { "epoch": 0.11, "learning_rate": 4.8198862643867785e-05, "loss": 0.0298, "step": 5568 }, { "epoch": 0.11, "learning_rate": 4.819821568361056e-05, "loss": 0.0317, "step": 5570 }, { "epoch": 0.11, "learning_rate": 4.819756872335332e-05, "loss": 0.0094, "step": 5572 }, { "epoch": 0.11, "learning_rate": 4.81969217630961e-05, "loss": 0.0218, "step": 5574 }, { "epoch": 0.11, "learning_rate": 4.819627480283886e-05, "loss": 0.023, "step": 5576 }, { "epoch": 0.11, "learning_rate": 4.819562784258163e-05, "loss": 0.0203, "step": 5578 }, { "epoch": 0.11, "learning_rate": 4.8194980882324406e-05, "loss": 0.0221, "step": 5580 }, { "epoch": 0.11, "learning_rate": 4.819433392206717e-05, "loss": 0.0212, "step": 5582 }, { "epoch": 0.11, "learning_rate": 4.819368696180994e-05, "loss": 0.0223, "step": 5584 }, { "epoch": 0.11, "learning_rate": 4.8193040001552707e-05, "loss": 0.0338, "step": 5586 }, { "epoch": 0.11, "learning_rate": 4.8192393041295476e-05, "loss": 0.0404, "step": 5588 }, { "epoch": 0.11, "learning_rate": 4.8191746081038245e-05, "loss": 0.0136, "step": 5590 }, { "epoch": 0.11, "learning_rate": 4.8191099120781014e-05, "loss": 0.0339, "step": 5592 }, { "epoch": 0.11, "learning_rate": 4.819045216052378e-05, "loss": 0.0167, "step": 5594 }, { "epoch": 0.11, "learning_rate": 4.818980520026655e-05, "loss": 0.0189, "step": 5596 }, { "epoch": 0.11, "learning_rate": 4.818915824000932e-05, "loss": 0.0144, "step": 5598 }, { "epoch": 0.11, "learning_rate": 4.818851127975208e-05, "loss": 0.017, "step": 5600 }, { "epoch": 0.11, "learning_rate": 4.818786431949486e-05, "loss": 0.0595, "step": 5602 }, { "epoch": 0.11, "learning_rate": 4.818721735923762e-05, "loss": 0.0183, "step": 5604 }, { "epoch": 0.11, "learning_rate": 4.818657039898039e-05, "loss": 0.0646, "step": 5606 }, { "epoch": 0.11, "learning_rate": 4.818592343872316e-05, "loss": 0.046, "step": 5608 }, { "epoch": 0.11, "learning_rate": 4.818527647846593e-05, "loss": 0.0265, "step": 5610 }, { "epoch": 0.11, "learning_rate": 4.81846295182087e-05, "loss": 0.0122, "step": 5612 }, { "epoch": 0.11, "learning_rate": 4.818398255795147e-05, "loss": 0.021, "step": 5614 }, { "epoch": 0.11, "learning_rate": 4.8183335597694236e-05, "loss": 0.0264, "step": 5616 }, { "epoch": 0.11, "learning_rate": 4.8182688637437005e-05, "loss": 0.0266, "step": 5618 }, { "epoch": 0.11, "learning_rate": 4.8182041677179774e-05, "loss": 0.0089, "step": 5620 }, { "epoch": 0.11, "learning_rate": 4.818139471692254e-05, "loss": 0.0515, "step": 5622 }, { "epoch": 0.11, "learning_rate": 4.818074775666531e-05, "loss": 0.0285, "step": 5624 }, { "epoch": 0.11, "learning_rate": 4.818010079640808e-05, "loss": 0.0111, "step": 5626 }, { "epoch": 0.11, "learning_rate": 4.8179453836150844e-05, "loss": 0.0349, "step": 5628 }, { "epoch": 0.11, "learning_rate": 4.817880687589362e-05, "loss": 0.0298, "step": 5630 }, { "epoch": 0.11, "learning_rate": 4.817815991563638e-05, "loss": 0.0227, "step": 5632 }, { "epoch": 0.11, "learning_rate": 4.817751295537916e-05, "loss": 0.0376, "step": 5634 }, { "epoch": 0.11, "learning_rate": 4.817686599512192e-05, "loss": 0.0235, "step": 5636 }, { "epoch": 0.11, "learning_rate": 4.817621903486469e-05, "loss": 0.0315, "step": 5638 }, { "epoch": 0.11, "learning_rate": 4.817557207460746e-05, "loss": 0.0545, "step": 5640 }, { "epoch": 0.11, "learning_rate": 4.817492511435023e-05, "loss": 0.0305, "step": 5642 }, { "epoch": 0.11, "learning_rate": 4.8174278154092996e-05, "loss": 0.0127, "step": 5644 }, { "epoch": 0.11, "learning_rate": 4.8173631193835766e-05, "loss": 0.0329, "step": 5646 }, { "epoch": 0.11, "learning_rate": 4.8172984233578535e-05, "loss": 0.0277, "step": 5648 }, { "epoch": 0.11, "learning_rate": 4.81723372733213e-05, "loss": 0.0316, "step": 5650 }, { "epoch": 0.11, "learning_rate": 4.817169031306407e-05, "loss": 0.0246, "step": 5652 }, { "epoch": 0.11, "learning_rate": 4.8171043352806835e-05, "loss": 0.0186, "step": 5654 }, { "epoch": 0.11, "learning_rate": 4.817039639254961e-05, "loss": 0.0422, "step": 5656 }, { "epoch": 0.11, "learning_rate": 4.816974943229238e-05, "loss": 0.0305, "step": 5658 }, { "epoch": 0.11, "learning_rate": 4.816910247203514e-05, "loss": 0.0185, "step": 5660 }, { "epoch": 0.11, "learning_rate": 4.816845551177792e-05, "loss": 0.0484, "step": 5662 }, { "epoch": 0.11, "learning_rate": 4.816780855152068e-05, "loss": 0.0175, "step": 5664 }, { "epoch": 0.11, "learning_rate": 4.816716159126345e-05, "loss": 0.0265, "step": 5666 }, { "epoch": 0.11, "learning_rate": 4.816651463100622e-05, "loss": 0.0359, "step": 5668 }, { "epoch": 0.11, "learning_rate": 4.816586767074899e-05, "loss": 0.0391, "step": 5670 }, { "epoch": 0.11, "learning_rate": 4.816522071049176e-05, "loss": 0.0303, "step": 5672 }, { "epoch": 0.11, "learning_rate": 4.8164573750234526e-05, "loss": 0.0361, "step": 5674 }, { "epoch": 0.11, "learning_rate": 4.8163926789977295e-05, "loss": 0.031, "step": 5676 }, { "epoch": 0.11, "learning_rate": 4.8163279829720064e-05, "loss": 0.0098, "step": 5678 }, { "epoch": 0.11, "learning_rate": 4.816263286946283e-05, "loss": 0.0395, "step": 5680 }, { "epoch": 0.11, "learning_rate": 4.8161985909205596e-05, "loss": 0.0445, "step": 5682 }, { "epoch": 0.11, "learning_rate": 4.816133894894837e-05, "loss": 0.0493, "step": 5684 }, { "epoch": 0.11, "learning_rate": 4.8160691988691134e-05, "loss": 0.054, "step": 5686 }, { "epoch": 0.11, "learning_rate": 4.81600450284339e-05, "loss": 0.0256, "step": 5688 }, { "epoch": 0.11, "learning_rate": 4.815939806817668e-05, "loss": 0.0588, "step": 5690 }, { "epoch": 0.11, "learning_rate": 4.815875110791944e-05, "loss": 0.0151, "step": 5692 }, { "epoch": 0.11, "learning_rate": 4.815810414766222e-05, "loss": 0.0262, "step": 5694 }, { "epoch": 0.11, "learning_rate": 4.815745718740498e-05, "loss": 0.0272, "step": 5696 }, { "epoch": 0.11, "learning_rate": 4.815681022714775e-05, "loss": 0.0226, "step": 5698 }, { "epoch": 0.11, "learning_rate": 4.815616326689052e-05, "loss": 0.0278, "step": 5700 }, { "epoch": 0.11, "learning_rate": 4.8155516306633286e-05, "loss": 0.0452, "step": 5702 }, { "epoch": 0.11, "learning_rate": 4.8154869346376055e-05, "loss": 0.0217, "step": 5704 }, { "epoch": 0.11, "learning_rate": 4.8154222386118825e-05, "loss": 0.0474, "step": 5706 }, { "epoch": 0.11, "learning_rate": 4.8153575425861594e-05, "loss": 0.023, "step": 5708 }, { "epoch": 0.11, "learning_rate": 4.8152928465604356e-05, "loss": 0.0331, "step": 5710 }, { "epoch": 0.11, "learning_rate": 4.815228150534713e-05, "loss": 0.0347, "step": 5712 }, { "epoch": 0.11, "learning_rate": 4.8151634545089894e-05, "loss": 0.0308, "step": 5714 }, { "epoch": 0.11, "learning_rate": 4.815098758483267e-05, "loss": 0.0143, "step": 5716 }, { "epoch": 0.11, "learning_rate": 4.815034062457543e-05, "loss": 0.0282, "step": 5718 }, { "epoch": 0.11, "learning_rate": 4.81496936643182e-05, "loss": 0.0399, "step": 5720 }, { "epoch": 0.11, "learning_rate": 4.814904670406098e-05, "loss": 0.0189, "step": 5722 }, { "epoch": 0.11, "learning_rate": 4.814839974380374e-05, "loss": 0.0094, "step": 5724 }, { "epoch": 0.11, "learning_rate": 4.814775278354651e-05, "loss": 0.0707, "step": 5726 }, { "epoch": 0.11, "learning_rate": 4.814710582328928e-05, "loss": 0.0382, "step": 5728 }, { "epoch": 0.11, "learning_rate": 4.814645886303205e-05, "loss": 0.0363, "step": 5730 }, { "epoch": 0.11, "learning_rate": 4.814581190277481e-05, "loss": 0.0192, "step": 5732 }, { "epoch": 0.11, "learning_rate": 4.8145164942517585e-05, "loss": 0.029, "step": 5734 }, { "epoch": 0.11, "learning_rate": 4.8144517982260354e-05, "loss": 0.0677, "step": 5736 }, { "epoch": 0.11, "learning_rate": 4.814387102200312e-05, "loss": 0.0188, "step": 5738 }, { "epoch": 0.11, "learning_rate": 4.814322406174589e-05, "loss": 0.0078, "step": 5740 }, { "epoch": 0.11, "learning_rate": 4.8142577101488654e-05, "loss": 0.0225, "step": 5742 }, { "epoch": 0.11, "learning_rate": 4.814193014123143e-05, "loss": 0.0289, "step": 5744 }, { "epoch": 0.11, "learning_rate": 4.814128318097419e-05, "loss": 0.0369, "step": 5746 }, { "epoch": 0.11, "learning_rate": 4.814063622071696e-05, "loss": 0.0159, "step": 5748 }, { "epoch": 0.11, "learning_rate": 4.813998926045973e-05, "loss": 0.0132, "step": 5750 }, { "epoch": 0.11, "learning_rate": 4.81393423002025e-05, "loss": 0.0353, "step": 5752 }, { "epoch": 0.11, "learning_rate": 4.813869533994527e-05, "loss": 0.0216, "step": 5754 }, { "epoch": 0.11, "learning_rate": 4.813804837968804e-05, "loss": 0.0265, "step": 5756 }, { "epoch": 0.11, "learning_rate": 4.813740141943081e-05, "loss": 0.0146, "step": 5758 }, { "epoch": 0.11, "learning_rate": 4.8136754459173576e-05, "loss": 0.0301, "step": 5760 }, { "epoch": 0.11, "learning_rate": 4.8136107498916345e-05, "loss": 0.0195, "step": 5762 }, { "epoch": 0.11, "learning_rate": 4.813546053865911e-05, "loss": 0.0182, "step": 5764 }, { "epoch": 0.11, "learning_rate": 4.8134813578401883e-05, "loss": 0.0424, "step": 5766 }, { "epoch": 0.11, "learning_rate": 4.813416661814465e-05, "loss": 0.0362, "step": 5768 }, { "epoch": 0.11, "learning_rate": 4.8133519657887415e-05, "loss": 0.0334, "step": 5770 }, { "epoch": 0.11, "learning_rate": 4.813287269763019e-05, "loss": 0.0389, "step": 5772 }, { "epoch": 0.11, "learning_rate": 4.813222573737295e-05, "loss": 0.0189, "step": 5774 }, { "epoch": 0.11, "learning_rate": 4.813157877711573e-05, "loss": 0.0178, "step": 5776 }, { "epoch": 0.11, "learning_rate": 4.813093181685849e-05, "loss": 0.0233, "step": 5778 }, { "epoch": 0.11, "learning_rate": 4.813028485660126e-05, "loss": 0.0057, "step": 5780 }, { "epoch": 0.11, "learning_rate": 4.812963789634403e-05, "loss": 0.0304, "step": 5782 }, { "epoch": 0.11, "learning_rate": 4.81289909360868e-05, "loss": 0.0399, "step": 5784 }, { "epoch": 0.11, "learning_rate": 4.812834397582957e-05, "loss": 0.0357, "step": 5786 }, { "epoch": 0.11, "learning_rate": 4.812769701557234e-05, "loss": 0.0367, "step": 5788 }, { "epoch": 0.11, "learning_rate": 4.8127050055315106e-05, "loss": 0.0697, "step": 5790 }, { "epoch": 0.11, "learning_rate": 4.812640309505787e-05, "loss": 0.0139, "step": 5792 }, { "epoch": 0.11, "learning_rate": 4.8125756134800644e-05, "loss": 0.0447, "step": 5794 }, { "epoch": 0.11, "learning_rate": 4.8125109174543406e-05, "loss": 0.0443, "step": 5796 }, { "epoch": 0.11, "learning_rate": 4.812446221428618e-05, "loss": 0.0341, "step": 5798 }, { "epoch": 0.11, "learning_rate": 4.8123815254028944e-05, "loss": 0.0211, "step": 5800 }, { "epoch": 0.11, "learning_rate": 4.8123168293771713e-05, "loss": 0.0177, "step": 5802 }, { "epoch": 0.11, "learning_rate": 4.812252133351449e-05, "loss": 0.0214, "step": 5804 }, { "epoch": 0.11, "learning_rate": 4.812187437325725e-05, "loss": 0.0155, "step": 5806 }, { "epoch": 0.11, "learning_rate": 4.812122741300002e-05, "loss": 0.0232, "step": 5808 }, { "epoch": 0.11, "learning_rate": 4.812058045274279e-05, "loss": 0.0168, "step": 5810 }, { "epoch": 0.11, "learning_rate": 4.811993349248556e-05, "loss": 0.0066, "step": 5812 }, { "epoch": 0.11, "learning_rate": 4.811928653222833e-05, "loss": 0.0204, "step": 5814 }, { "epoch": 0.11, "learning_rate": 4.81186395719711e-05, "loss": 0.0196, "step": 5816 }, { "epoch": 0.11, "learning_rate": 4.8117992611713866e-05, "loss": 0.0283, "step": 5818 }, { "epoch": 0.11, "learning_rate": 4.8117345651456635e-05, "loss": 0.0118, "step": 5820 }, { "epoch": 0.11, "learning_rate": 4.8116698691199404e-05, "loss": 0.0224, "step": 5822 }, { "epoch": 0.11, "learning_rate": 4.811605173094217e-05, "loss": 0.0295, "step": 5824 }, { "epoch": 0.11, "learning_rate": 4.811540477068494e-05, "loss": 0.0661, "step": 5826 }, { "epoch": 0.11, "learning_rate": 4.8114757810427705e-05, "loss": 0.0265, "step": 5828 }, { "epoch": 0.11, "learning_rate": 4.8114110850170474e-05, "loss": 0.0365, "step": 5830 }, { "epoch": 0.11, "learning_rate": 4.811346388991324e-05, "loss": 0.0459, "step": 5832 }, { "epoch": 0.11, "learning_rate": 4.811281692965601e-05, "loss": 0.0598, "step": 5834 }, { "epoch": 0.11, "learning_rate": 4.811216996939879e-05, "loss": 0.0358, "step": 5836 }, { "epoch": 0.11, "learning_rate": 4.811152300914155e-05, "loss": 0.0724, "step": 5838 }, { "epoch": 0.11, "learning_rate": 4.811087604888432e-05, "loss": 0.0244, "step": 5840 }, { "epoch": 0.11, "learning_rate": 4.811022908862709e-05, "loss": 0.0171, "step": 5842 }, { "epoch": 0.11, "learning_rate": 4.810958212836986e-05, "loss": 0.0195, "step": 5844 }, { "epoch": 0.11, "learning_rate": 4.8108935168112627e-05, "loss": 0.0175, "step": 5846 }, { "epoch": 0.11, "learning_rate": 4.8108288207855396e-05, "loss": 0.0317, "step": 5848 }, { "epoch": 0.11, "learning_rate": 4.8107641247598165e-05, "loss": 0.0105, "step": 5850 }, { "epoch": 0.11, "learning_rate": 4.810699428734093e-05, "loss": 0.0716, "step": 5852 }, { "epoch": 0.11, "learning_rate": 4.81063473270837e-05, "loss": 0.0087, "step": 5854 }, { "epoch": 0.11, "learning_rate": 4.8105700366826465e-05, "loss": 0.0073, "step": 5856 }, { "epoch": 0.11, "learning_rate": 4.810505340656924e-05, "loss": 0.0397, "step": 5858 }, { "epoch": 0.11, "learning_rate": 4.8104406446312e-05, "loss": 0.042, "step": 5860 }, { "epoch": 0.11, "learning_rate": 4.810375948605477e-05, "loss": 0.1009, "step": 5862 }, { "epoch": 0.11, "learning_rate": 4.810311252579754e-05, "loss": 0.0317, "step": 5864 }, { "epoch": 0.11, "learning_rate": 4.810246556554031e-05, "loss": 0.0337, "step": 5866 }, { "epoch": 0.11, "learning_rate": 4.810181860528308e-05, "loss": 0.0406, "step": 5868 }, { "epoch": 0.11, "learning_rate": 4.810117164502585e-05, "loss": 0.0289, "step": 5870 }, { "epoch": 0.11, "learning_rate": 4.810052468476862e-05, "loss": 0.0724, "step": 5872 }, { "epoch": 0.11, "learning_rate": 4.809987772451138e-05, "loss": 0.0403, "step": 5874 }, { "epoch": 0.11, "learning_rate": 4.8099230764254156e-05, "loss": 0.0336, "step": 5876 }, { "epoch": 0.11, "learning_rate": 4.809858380399692e-05, "loss": 0.0386, "step": 5878 }, { "epoch": 0.11, "learning_rate": 4.8097936843739694e-05, "loss": 0.0411, "step": 5880 }, { "epoch": 0.11, "learning_rate": 4.809728988348246e-05, "loss": 0.0277, "step": 5882 }, { "epoch": 0.11, "learning_rate": 4.8096642923225226e-05, "loss": 0.0376, "step": 5884 }, { "epoch": 0.11, "learning_rate": 4.8095995962968e-05, "loss": 0.0276, "step": 5886 }, { "epoch": 0.11, "learning_rate": 4.8095349002710764e-05, "loss": 0.0279, "step": 5888 }, { "epoch": 0.11, "learning_rate": 4.809470204245353e-05, "loss": 0.0344, "step": 5890 }, { "epoch": 0.11, "learning_rate": 4.80940550821963e-05, "loss": 0.0569, "step": 5892 }, { "epoch": 0.11, "learning_rate": 4.809340812193907e-05, "loss": 0.0335, "step": 5894 }, { "epoch": 0.11, "learning_rate": 4.809276116168184e-05, "loss": 0.0137, "step": 5896 }, { "epoch": 0.11, "learning_rate": 4.809211420142461e-05, "loss": 0.0158, "step": 5898 }, { "epoch": 0.11, "learning_rate": 4.809146724116738e-05, "loss": 0.0333, "step": 5900 }, { "epoch": 0.11, "learning_rate": 4.809082028091015e-05, "loss": 0.0513, "step": 5902 }, { "epoch": 0.11, "learning_rate": 4.8090173320652916e-05, "loss": 0.0464, "step": 5904 }, { "epoch": 0.11, "learning_rate": 4.808952636039568e-05, "loss": 0.0292, "step": 5906 }, { "epoch": 0.11, "learning_rate": 4.8088879400138455e-05, "loss": 0.0416, "step": 5908 }, { "epoch": 0.11, "learning_rate": 4.808823243988122e-05, "loss": 0.0437, "step": 5910 }, { "epoch": 0.11, "learning_rate": 4.8087585479623986e-05, "loss": 0.0278, "step": 5912 }, { "epoch": 0.11, "learning_rate": 4.808693851936676e-05, "loss": 0.0185, "step": 5914 }, { "epoch": 0.11, "learning_rate": 4.8086291559109524e-05, "loss": 0.031, "step": 5916 }, { "epoch": 0.11, "learning_rate": 4.80856445988523e-05, "loss": 0.017, "step": 5918 }, { "epoch": 0.11, "learning_rate": 4.808499763859506e-05, "loss": 0.0267, "step": 5920 }, { "epoch": 0.11, "learning_rate": 4.808435067833783e-05, "loss": 0.0133, "step": 5922 }, { "epoch": 0.11, "learning_rate": 4.80837037180806e-05, "loss": 0.0566, "step": 5924 }, { "epoch": 0.12, "learning_rate": 4.808305675782337e-05, "loss": 0.0261, "step": 5926 }, { "epoch": 0.12, "learning_rate": 4.808240979756614e-05, "loss": 0.0126, "step": 5928 }, { "epoch": 0.12, "learning_rate": 4.808176283730891e-05, "loss": 0.0264, "step": 5930 }, { "epoch": 0.12, "learning_rate": 4.808111587705168e-05, "loss": 0.0558, "step": 5932 }, { "epoch": 0.12, "learning_rate": 4.808046891679444e-05, "loss": 0.0282, "step": 5934 }, { "epoch": 0.12, "learning_rate": 4.8079821956537215e-05, "loss": 0.0203, "step": 5936 }, { "epoch": 0.12, "learning_rate": 4.807917499627998e-05, "loss": 0.0297, "step": 5938 }, { "epoch": 0.12, "learning_rate": 4.807852803602275e-05, "loss": 0.0178, "step": 5940 }, { "epoch": 0.12, "learning_rate": 4.8077881075765515e-05, "loss": 0.024, "step": 5942 }, { "epoch": 0.12, "learning_rate": 4.8077234115508285e-05, "loss": 0.0134, "step": 5944 }, { "epoch": 0.12, "learning_rate": 4.807658715525106e-05, "loss": 0.0181, "step": 5946 }, { "epoch": 0.12, "learning_rate": 4.807594019499382e-05, "loss": 0.021, "step": 5948 }, { "epoch": 0.12, "learning_rate": 4.807529323473659e-05, "loss": 0.036, "step": 5950 }, { "epoch": 0.12, "learning_rate": 4.807464627447936e-05, "loss": 0.0258, "step": 5952 }, { "epoch": 0.12, "learning_rate": 4.807399931422213e-05, "loss": 0.0196, "step": 5954 }, { "epoch": 0.12, "learning_rate": 4.807335235396489e-05, "loss": 0.0172, "step": 5956 }, { "epoch": 0.12, "learning_rate": 4.807270539370767e-05, "loss": 0.0364, "step": 5958 }, { "epoch": 0.12, "learning_rate": 4.807205843345044e-05, "loss": 0.0651, "step": 5960 }, { "epoch": 0.12, "learning_rate": 4.8071411473193206e-05, "loss": 0.0712, "step": 5962 }, { "epoch": 0.12, "learning_rate": 4.8070764512935975e-05, "loss": 0.0249, "step": 5964 }, { "epoch": 0.12, "learning_rate": 4.807011755267874e-05, "loss": 0.0161, "step": 5966 }, { "epoch": 0.12, "learning_rate": 4.8069470592421514e-05, "loss": 0.0223, "step": 5968 }, { "epoch": 0.12, "learning_rate": 4.8068823632164276e-05, "loss": 0.0322, "step": 5970 }, { "epoch": 0.12, "learning_rate": 4.8068176671907045e-05, "loss": 0.019, "step": 5972 }, { "epoch": 0.12, "learning_rate": 4.8067529711649814e-05, "loss": 0.0186, "step": 5974 }, { "epoch": 0.12, "learning_rate": 4.806688275139258e-05, "loss": 0.0268, "step": 5976 }, { "epoch": 0.12, "learning_rate": 4.806623579113535e-05, "loss": 0.0851, "step": 5978 }, { "epoch": 0.12, "learning_rate": 4.806558883087812e-05, "loss": 0.0606, "step": 5980 }, { "epoch": 0.12, "learning_rate": 4.806494187062089e-05, "loss": 0.0069, "step": 5982 }, { "epoch": 0.12, "learning_rate": 4.806429491036366e-05, "loss": 0.0179, "step": 5984 }, { "epoch": 0.12, "learning_rate": 4.806364795010643e-05, "loss": 0.0534, "step": 5986 }, { "epoch": 0.12, "learning_rate": 4.806300098984919e-05, "loss": 0.0278, "step": 5988 }, { "epoch": 0.12, "learning_rate": 4.806235402959197e-05, "loss": 0.0432, "step": 5990 }, { "epoch": 0.12, "learning_rate": 4.8061707069334736e-05, "loss": 0.0398, "step": 5992 }, { "epoch": 0.12, "learning_rate": 4.80610601090775e-05, "loss": 0.0272, "step": 5994 }, { "epoch": 0.12, "learning_rate": 4.8060413148820274e-05, "loss": 0.0281, "step": 5996 }, { "epoch": 0.12, "learning_rate": 4.8059766188563036e-05, "loss": 0.0314, "step": 5998 }, { "epoch": 0.12, "learning_rate": 4.805911922830581e-05, "loss": 0.0317, "step": 6000 }, { "epoch": 0.12, "learning_rate": 4.8058472268048574e-05, "loss": 0.0369, "step": 6002 }, { "epoch": 0.12, "learning_rate": 4.8057825307791344e-05, "loss": 0.024, "step": 6004 }, { "epoch": 0.12, "learning_rate": 4.805717834753411e-05, "loss": 0.039, "step": 6006 }, { "epoch": 0.12, "learning_rate": 4.805653138727688e-05, "loss": 0.0397, "step": 6008 }, { "epoch": 0.12, "learning_rate": 4.805588442701965e-05, "loss": 0.0606, "step": 6010 }, { "epoch": 0.12, "learning_rate": 4.805523746676242e-05, "loss": 0.0182, "step": 6012 }, { "epoch": 0.12, "learning_rate": 4.805459050650519e-05, "loss": 0.0287, "step": 6014 }, { "epoch": 0.12, "learning_rate": 4.805394354624795e-05, "loss": 0.025, "step": 6016 }, { "epoch": 0.12, "learning_rate": 4.805329658599073e-05, "loss": 0.0135, "step": 6018 }, { "epoch": 0.12, "learning_rate": 4.805264962573349e-05, "loss": 0.0264, "step": 6020 }, { "epoch": 0.12, "learning_rate": 4.8052002665476265e-05, "loss": 0.0354, "step": 6022 }, { "epoch": 0.12, "learning_rate": 4.8051355705219034e-05, "loss": 0.017, "step": 6024 }, { "epoch": 0.12, "learning_rate": 4.80507087449618e-05, "loss": 0.0399, "step": 6026 }, { "epoch": 0.12, "learning_rate": 4.805006178470457e-05, "loss": 0.0212, "step": 6028 }, { "epoch": 0.12, "learning_rate": 4.8049414824447335e-05, "loss": 0.0391, "step": 6030 }, { "epoch": 0.12, "learning_rate": 4.8048767864190104e-05, "loss": 0.0252, "step": 6032 }, { "epoch": 0.12, "learning_rate": 4.804812090393287e-05, "loss": 0.008, "step": 6034 }, { "epoch": 0.12, "learning_rate": 4.804747394367564e-05, "loss": 0.0274, "step": 6036 }, { "epoch": 0.12, "learning_rate": 4.804682698341841e-05, "loss": 0.0671, "step": 6038 }, { "epoch": 0.12, "learning_rate": 4.804618002316118e-05, "loss": 0.0395, "step": 6040 }, { "epoch": 0.12, "learning_rate": 4.804553306290395e-05, "loss": 0.0443, "step": 6042 }, { "epoch": 0.12, "learning_rate": 4.804488610264672e-05, "loss": 0.0493, "step": 6044 }, { "epoch": 0.12, "learning_rate": 4.804423914238949e-05, "loss": 0.0406, "step": 6046 }, { "epoch": 0.12, "learning_rate": 4.804359218213225e-05, "loss": 0.0161, "step": 6048 }, { "epoch": 0.12, "learning_rate": 4.8042945221875026e-05, "loss": 0.0667, "step": 6050 }, { "epoch": 0.12, "learning_rate": 4.804229826161779e-05, "loss": 0.0274, "step": 6052 }, { "epoch": 0.12, "learning_rate": 4.804165130136056e-05, "loss": 0.0615, "step": 6054 }, { "epoch": 0.12, "learning_rate": 4.8041004341103326e-05, "loss": 0.0231, "step": 6056 }, { "epoch": 0.12, "learning_rate": 4.8040357380846095e-05, "loss": 0.0462, "step": 6058 }, { "epoch": 0.12, "learning_rate": 4.803971042058887e-05, "loss": 0.0414, "step": 6060 }, { "epoch": 0.12, "learning_rate": 4.8039063460331633e-05, "loss": 0.0458, "step": 6062 }, { "epoch": 0.12, "learning_rate": 4.80384165000744e-05, "loss": 0.0161, "step": 6064 }, { "epoch": 0.12, "learning_rate": 4.803776953981717e-05, "loss": 0.0098, "step": 6066 }, { "epoch": 0.12, "learning_rate": 4.803712257955994e-05, "loss": 0.0226, "step": 6068 }, { "epoch": 0.12, "learning_rate": 4.803647561930271e-05, "loss": 0.0388, "step": 6070 }, { "epoch": 0.12, "learning_rate": 4.803582865904548e-05, "loss": 0.0222, "step": 6072 }, { "epoch": 0.12, "learning_rate": 4.803518169878825e-05, "loss": 0.0409, "step": 6074 }, { "epoch": 0.12, "learning_rate": 4.803453473853101e-05, "loss": 0.0157, "step": 6076 }, { "epoch": 0.12, "learning_rate": 4.8033887778273786e-05, "loss": 0.0307, "step": 6078 }, { "epoch": 0.12, "learning_rate": 4.803324081801655e-05, "loss": 0.0186, "step": 6080 }, { "epoch": 0.12, "learning_rate": 4.8032593857759324e-05, "loss": 0.0254, "step": 6082 }, { "epoch": 0.12, "learning_rate": 4.8031946897502087e-05, "loss": 0.0328, "step": 6084 }, { "epoch": 0.12, "learning_rate": 4.8031299937244856e-05, "loss": 0.016, "step": 6086 }, { "epoch": 0.12, "learning_rate": 4.8030652976987625e-05, "loss": 0.0824, "step": 6088 }, { "epoch": 0.12, "learning_rate": 4.8030006016730394e-05, "loss": 0.0433, "step": 6090 }, { "epoch": 0.12, "learning_rate": 4.802935905647316e-05, "loss": 0.0346, "step": 6092 }, { "epoch": 0.12, "learning_rate": 4.802871209621593e-05, "loss": 0.0228, "step": 6094 }, { "epoch": 0.12, "learning_rate": 4.80280651359587e-05, "loss": 0.0206, "step": 6096 }, { "epoch": 0.12, "learning_rate": 4.8027418175701463e-05, "loss": 0.0146, "step": 6098 }, { "epoch": 0.12, "learning_rate": 4.802677121544424e-05, "loss": 0.0283, "step": 6100 }, { "epoch": 0.12, "learning_rate": 4.802612425518701e-05, "loss": 0.0272, "step": 6102 }, { "epoch": 0.12, "learning_rate": 4.802547729492978e-05, "loss": 0.0649, "step": 6104 }, { "epoch": 0.12, "learning_rate": 4.8024830334672547e-05, "loss": 0.0158, "step": 6106 }, { "epoch": 0.12, "learning_rate": 4.802418337441531e-05, "loss": 0.0119, "step": 6108 }, { "epoch": 0.12, "learning_rate": 4.8023536414158085e-05, "loss": 0.0377, "step": 6110 }, { "epoch": 0.12, "learning_rate": 4.802288945390085e-05, "loss": 0.0151, "step": 6112 }, { "epoch": 0.12, "learning_rate": 4.8022242493643616e-05, "loss": 0.0343, "step": 6114 }, { "epoch": 0.12, "learning_rate": 4.8021595533386385e-05, "loss": 0.0427, "step": 6116 }, { "epoch": 0.12, "learning_rate": 4.8020948573129154e-05, "loss": 0.0194, "step": 6118 }, { "epoch": 0.12, "learning_rate": 4.802030161287192e-05, "loss": 0.0118, "step": 6120 }, { "epoch": 0.12, "learning_rate": 4.801965465261469e-05, "loss": 0.0562, "step": 6122 }, { "epoch": 0.12, "learning_rate": 4.801900769235746e-05, "loss": 0.0247, "step": 6124 }, { "epoch": 0.12, "learning_rate": 4.801836073210023e-05, "loss": 0.0154, "step": 6126 }, { "epoch": 0.12, "learning_rate": 4.8017713771843e-05, "loss": 0.0213, "step": 6128 }, { "epoch": 0.12, "learning_rate": 4.801706681158576e-05, "loss": 0.0303, "step": 6130 }, { "epoch": 0.12, "learning_rate": 4.801641985132854e-05, "loss": 0.0341, "step": 6132 }, { "epoch": 0.12, "learning_rate": 4.80157728910713e-05, "loss": 0.0369, "step": 6134 }, { "epoch": 0.12, "learning_rate": 4.801512593081407e-05, "loss": 0.0925, "step": 6136 }, { "epoch": 0.12, "learning_rate": 4.8014478970556845e-05, "loss": 0.0157, "step": 6138 }, { "epoch": 0.12, "learning_rate": 4.801383201029961e-05, "loss": 0.0288, "step": 6140 }, { "epoch": 0.12, "learning_rate": 4.801318505004238e-05, "loss": 0.0625, "step": 6142 }, { "epoch": 0.12, "learning_rate": 4.8012538089785146e-05, "loss": 0.0293, "step": 6144 }, { "epoch": 0.12, "learning_rate": 4.8011891129527915e-05, "loss": 0.022, "step": 6146 }, { "epoch": 0.12, "learning_rate": 4.8011244169270684e-05, "loss": 0.0081, "step": 6148 }, { "epoch": 0.12, "learning_rate": 4.801059720901345e-05, "loss": 0.0425, "step": 6150 }, { "epoch": 0.12, "learning_rate": 4.800995024875622e-05, "loss": 0.0284, "step": 6152 }, { "epoch": 0.12, "learning_rate": 4.800930328849899e-05, "loss": 0.0264, "step": 6154 }, { "epoch": 0.12, "learning_rate": 4.800865632824176e-05, "loss": 0.0176, "step": 6156 }, { "epoch": 0.12, "learning_rate": 4.800800936798452e-05, "loss": 0.0273, "step": 6158 }, { "epoch": 0.12, "learning_rate": 4.80073624077273e-05, "loss": 0.0159, "step": 6160 }, { "epoch": 0.12, "learning_rate": 4.800671544747006e-05, "loss": 0.0372, "step": 6162 }, { "epoch": 0.12, "learning_rate": 4.8006068487212836e-05, "loss": 0.0174, "step": 6164 }, { "epoch": 0.12, "learning_rate": 4.80054215269556e-05, "loss": 0.0383, "step": 6166 }, { "epoch": 0.12, "learning_rate": 4.800477456669837e-05, "loss": 0.0266, "step": 6168 }, { "epoch": 0.12, "learning_rate": 4.8004127606441144e-05, "loss": 0.0082, "step": 6170 }, { "epoch": 0.12, "learning_rate": 4.8003480646183906e-05, "loss": 0.019, "step": 6172 }, { "epoch": 0.12, "learning_rate": 4.8002833685926675e-05, "loss": 0.0381, "step": 6174 }, { "epoch": 0.12, "learning_rate": 4.8002186725669444e-05, "loss": 0.0366, "step": 6176 }, { "epoch": 0.12, "learning_rate": 4.800153976541221e-05, "loss": 0.0105, "step": 6178 }, { "epoch": 0.12, "learning_rate": 4.8000892805154976e-05, "loss": 0.0266, "step": 6180 }, { "epoch": 0.12, "learning_rate": 4.800024584489775e-05, "loss": 0.0488, "step": 6182 }, { "epoch": 0.12, "learning_rate": 4.799959888464052e-05, "loss": 0.2202, "step": 6184 }, { "epoch": 0.12, "learning_rate": 4.799895192438329e-05, "loss": 0.0218, "step": 6186 }, { "epoch": 0.12, "learning_rate": 4.799830496412606e-05, "loss": 0.0153, "step": 6188 }, { "epoch": 0.12, "learning_rate": 4.799765800386882e-05, "loss": 0.0731, "step": 6190 }, { "epoch": 0.12, "learning_rate": 4.79970110436116e-05, "loss": 0.031, "step": 6192 }, { "epoch": 0.12, "learning_rate": 4.799636408335436e-05, "loss": 0.0168, "step": 6194 }, { "epoch": 0.12, "learning_rate": 4.799571712309713e-05, "loss": 0.044, "step": 6196 }, { "epoch": 0.12, "learning_rate": 4.79950701628399e-05, "loss": 0.0221, "step": 6198 }, { "epoch": 0.12, "learning_rate": 4.7994423202582666e-05, "loss": 0.0172, "step": 6200 }, { "epoch": 0.12, "learning_rate": 4.799377624232544e-05, "loss": 0.0297, "step": 6202 }, { "epoch": 0.12, "learning_rate": 4.7993129282068205e-05, "loss": 0.0439, "step": 6204 }, { "epoch": 0.12, "learning_rate": 4.7992482321810974e-05, "loss": 0.0182, "step": 6206 }, { "epoch": 0.12, "learning_rate": 4.799183536155374e-05, "loss": 0.0192, "step": 6208 }, { "epoch": 0.12, "learning_rate": 4.799118840129651e-05, "loss": 0.0318, "step": 6210 }, { "epoch": 0.12, "learning_rate": 4.7990541441039274e-05, "loss": 0.0096, "step": 6212 }, { "epoch": 0.12, "learning_rate": 4.798989448078205e-05, "loss": 0.0119, "step": 6214 }, { "epoch": 0.12, "learning_rate": 4.798924752052482e-05, "loss": 0.0264, "step": 6216 }, { "epoch": 0.12, "learning_rate": 4.798860056026758e-05, "loss": 0.0161, "step": 6218 }, { "epoch": 0.12, "learning_rate": 4.798795360001036e-05, "loss": 0.0513, "step": 6220 }, { "epoch": 0.12, "learning_rate": 4.798730663975312e-05, "loss": 0.0262, "step": 6222 }, { "epoch": 0.12, "learning_rate": 4.7986659679495895e-05, "loss": 0.1103, "step": 6224 }, { "epoch": 0.12, "learning_rate": 4.798601271923866e-05, "loss": 0.0102, "step": 6226 }, { "epoch": 0.12, "learning_rate": 4.798536575898143e-05, "loss": 0.0507, "step": 6228 }, { "epoch": 0.12, "learning_rate": 4.7984718798724196e-05, "loss": 0.0172, "step": 6230 }, { "epoch": 0.12, "learning_rate": 4.7984071838466965e-05, "loss": 0.0268, "step": 6232 }, { "epoch": 0.12, "learning_rate": 4.7983424878209734e-05, "loss": 0.028, "step": 6234 }, { "epoch": 0.12, "learning_rate": 4.79827779179525e-05, "loss": 0.0175, "step": 6236 }, { "epoch": 0.12, "learning_rate": 4.798213095769527e-05, "loss": 0.0157, "step": 6238 }, { "epoch": 0.12, "learning_rate": 4.7981483997438035e-05, "loss": 0.0122, "step": 6240 }, { "epoch": 0.12, "learning_rate": 4.798083703718081e-05, "loss": 0.0184, "step": 6242 }, { "epoch": 0.12, "learning_rate": 4.798019007692357e-05, "loss": 0.0277, "step": 6244 }, { "epoch": 0.12, "learning_rate": 4.797954311666635e-05, "loss": 0.021, "step": 6246 }, { "epoch": 0.12, "learning_rate": 4.797889615640912e-05, "loss": 0.0209, "step": 6248 }, { "epoch": 0.12, "learning_rate": 4.797824919615188e-05, "loss": 0.0111, "step": 6250 }, { "epoch": 0.12, "learning_rate": 4.7977602235894656e-05, "loss": 0.0484, "step": 6252 }, { "epoch": 0.12, "learning_rate": 4.797695527563742e-05, "loss": 0.0279, "step": 6254 }, { "epoch": 0.12, "learning_rate": 4.797630831538019e-05, "loss": 0.019, "step": 6256 }, { "epoch": 0.12, "learning_rate": 4.7975661355122956e-05, "loss": 0.016, "step": 6258 }, { "epoch": 0.12, "learning_rate": 4.7975014394865725e-05, "loss": 0.0327, "step": 6260 }, { "epoch": 0.12, "learning_rate": 4.7974367434608494e-05, "loss": 0.0213, "step": 6262 }, { "epoch": 0.12, "learning_rate": 4.7973720474351264e-05, "loss": 0.0155, "step": 6264 }, { "epoch": 0.12, "learning_rate": 4.797307351409403e-05, "loss": 0.1003, "step": 6266 }, { "epoch": 0.12, "learning_rate": 4.79724265538368e-05, "loss": 0.039, "step": 6268 }, { "epoch": 0.12, "learning_rate": 4.797177959357957e-05, "loss": 0.0465, "step": 6270 }, { "epoch": 0.12, "learning_rate": 4.797113263332233e-05, "loss": 0.0394, "step": 6272 }, { "epoch": 0.12, "learning_rate": 4.797048567306511e-05, "loss": 0.055, "step": 6274 }, { "epoch": 0.12, "learning_rate": 4.796983871280787e-05, "loss": 0.0176, "step": 6276 }, { "epoch": 0.12, "learning_rate": 4.796919175255064e-05, "loss": 0.0193, "step": 6278 }, { "epoch": 0.12, "learning_rate": 4.796854479229341e-05, "loss": 0.0457, "step": 6280 }, { "epoch": 0.12, "learning_rate": 4.796789783203618e-05, "loss": 0.1201, "step": 6282 }, { "epoch": 0.12, "learning_rate": 4.7967250871778954e-05, "loss": 0.0282, "step": 6284 }, { "epoch": 0.12, "learning_rate": 4.796660391152172e-05, "loss": 0.0325, "step": 6286 }, { "epoch": 0.12, "learning_rate": 4.7965956951264486e-05, "loss": 0.0147, "step": 6288 }, { "epoch": 0.12, "learning_rate": 4.7965309991007255e-05, "loss": 0.013, "step": 6290 }, { "epoch": 0.12, "learning_rate": 4.7964663030750024e-05, "loss": 0.0177, "step": 6292 }, { "epoch": 0.12, "learning_rate": 4.796401607049279e-05, "loss": 0.0398, "step": 6294 }, { "epoch": 0.12, "learning_rate": 4.796336911023556e-05, "loss": 0.0183, "step": 6296 }, { "epoch": 0.12, "learning_rate": 4.796272214997833e-05, "loss": 0.024, "step": 6298 }, { "epoch": 0.12, "learning_rate": 4.7962075189721093e-05, "loss": 0.0355, "step": 6300 }, { "epoch": 0.12, "learning_rate": 4.796142822946387e-05, "loss": 0.0332, "step": 6302 }, { "epoch": 0.12, "learning_rate": 4.796078126920663e-05, "loss": 0.0191, "step": 6304 }, { "epoch": 0.12, "learning_rate": 4.796013430894941e-05, "loss": 0.0172, "step": 6306 }, { "epoch": 0.12, "learning_rate": 4.795948734869217e-05, "loss": 0.0138, "step": 6308 }, { "epoch": 0.12, "learning_rate": 4.795884038843494e-05, "loss": 0.0627, "step": 6310 }, { "epoch": 0.12, "learning_rate": 4.795819342817771e-05, "loss": 0.0069, "step": 6312 }, { "epoch": 0.12, "learning_rate": 4.795754646792048e-05, "loss": 0.0507, "step": 6314 }, { "epoch": 0.12, "learning_rate": 4.7956899507663246e-05, "loss": 0.0292, "step": 6316 }, { "epoch": 0.12, "learning_rate": 4.7956252547406015e-05, "loss": 0.0665, "step": 6318 }, { "epoch": 0.12, "learning_rate": 4.7955605587148784e-05, "loss": 0.0321, "step": 6320 }, { "epoch": 0.12, "learning_rate": 4.795495862689155e-05, "loss": 0.0292, "step": 6322 }, { "epoch": 0.12, "learning_rate": 4.795431166663432e-05, "loss": 0.0221, "step": 6324 }, { "epoch": 0.12, "learning_rate": 4.795366470637709e-05, "loss": 0.0391, "step": 6326 }, { "epoch": 0.12, "learning_rate": 4.795301774611986e-05, "loss": 0.0208, "step": 6328 }, { "epoch": 0.12, "learning_rate": 4.795237078586263e-05, "loss": 0.0174, "step": 6330 }, { "epoch": 0.12, "learning_rate": 4.795172382560539e-05, "loss": 0.0288, "step": 6332 }, { "epoch": 0.12, "learning_rate": 4.795107686534817e-05, "loss": 0.0582, "step": 6334 }, { "epoch": 0.12, "learning_rate": 4.795042990509093e-05, "loss": 0.0228, "step": 6336 }, { "epoch": 0.12, "learning_rate": 4.79497829448337e-05, "loss": 0.0326, "step": 6338 }, { "epoch": 0.12, "learning_rate": 4.794913598457647e-05, "loss": 0.0336, "step": 6340 }, { "epoch": 0.12, "learning_rate": 4.794848902431924e-05, "loss": 0.0563, "step": 6342 }, { "epoch": 0.12, "learning_rate": 4.7947842064062007e-05, "loss": 0.0347, "step": 6344 }, { "epoch": 0.12, "learning_rate": 4.7947195103804776e-05, "loss": 0.0401, "step": 6346 }, { "epoch": 0.12, "learning_rate": 4.7946548143547545e-05, "loss": 0.0236, "step": 6348 }, { "epoch": 0.12, "learning_rate": 4.7945901183290314e-05, "loss": 0.0185, "step": 6350 }, { "epoch": 0.12, "learning_rate": 4.794525422303308e-05, "loss": 0.0327, "step": 6352 }, { "epoch": 0.12, "learning_rate": 4.7944607262775845e-05, "loss": 0.0144, "step": 6354 }, { "epoch": 0.12, "learning_rate": 4.794396030251862e-05, "loss": 0.0094, "step": 6356 }, { "epoch": 0.12, "learning_rate": 4.7943313342261383e-05, "loss": 0.0143, "step": 6358 }, { "epoch": 0.12, "learning_rate": 4.794266638200415e-05, "loss": 0.013, "step": 6360 }, { "epoch": 0.12, "learning_rate": 4.794201942174693e-05, "loss": 0.0295, "step": 6362 }, { "epoch": 0.12, "learning_rate": 4.794137246148969e-05, "loss": 0.0252, "step": 6364 }, { "epoch": 0.12, "learning_rate": 4.7940725501232467e-05, "loss": 0.0117, "step": 6366 }, { "epoch": 0.12, "learning_rate": 4.794007854097523e-05, "loss": 0.0307, "step": 6368 }, { "epoch": 0.12, "learning_rate": 4.7939431580718e-05, "loss": 0.0208, "step": 6370 }, { "epoch": 0.12, "learning_rate": 4.793878462046077e-05, "loss": 0.0205, "step": 6372 }, { "epoch": 0.12, "learning_rate": 4.7938137660203536e-05, "loss": 0.041, "step": 6374 }, { "epoch": 0.12, "learning_rate": 4.7937490699946305e-05, "loss": 0.0361, "step": 6376 }, { "epoch": 0.12, "learning_rate": 4.7936843739689074e-05, "loss": 0.022, "step": 6378 }, { "epoch": 0.12, "learning_rate": 4.793619677943184e-05, "loss": 0.0801, "step": 6380 }, { "epoch": 0.12, "learning_rate": 4.7935549819174606e-05, "loss": 0.0189, "step": 6382 }, { "epoch": 0.12, "learning_rate": 4.793490285891738e-05, "loss": 0.0057, "step": 6384 }, { "epoch": 0.12, "learning_rate": 4.7934255898660144e-05, "loss": 0.0227, "step": 6386 }, { "epoch": 0.12, "learning_rate": 4.793360893840292e-05, "loss": 0.0493, "step": 6388 }, { "epoch": 0.12, "learning_rate": 4.793296197814568e-05, "loss": 0.0314, "step": 6390 }, { "epoch": 0.12, "learning_rate": 4.793231501788845e-05, "loss": 0.0392, "step": 6392 }, { "epoch": 0.12, "learning_rate": 4.793166805763123e-05, "loss": 0.016, "step": 6394 }, { "epoch": 0.12, "learning_rate": 4.793102109737399e-05, "loss": 0.0265, "step": 6396 }, { "epoch": 0.12, "learning_rate": 4.793037413711676e-05, "loss": 0.0229, "step": 6398 }, { "epoch": 0.12, "learning_rate": 4.792972717685953e-05, "loss": 0.0228, "step": 6400 }, { "epoch": 0.12, "learning_rate": 4.792940369673091e-05, "loss": 0.1248, "step": 6402 }, { "epoch": 0.12, "learning_rate": 4.792875673647368e-05, "loss": 0.0366, "step": 6404 }, { "epoch": 0.12, "learning_rate": 4.792810977621645e-05, "loss": 0.026, "step": 6406 }, { "epoch": 0.12, "learning_rate": 4.792746281595922e-05, "loss": 0.0101, "step": 6408 }, { "epoch": 0.12, "learning_rate": 4.792681585570198e-05, "loss": 0.0495, "step": 6410 }, { "epoch": 0.12, "learning_rate": 4.792616889544476e-05, "loss": 0.0343, "step": 6412 }, { "epoch": 0.12, "learning_rate": 4.792552193518752e-05, "loss": 0.0165, "step": 6414 }, { "epoch": 0.12, "learning_rate": 4.7924874974930296e-05, "loss": 0.017, "step": 6416 }, { "epoch": 0.12, "learning_rate": 4.7924228014673065e-05, "loss": 0.0271, "step": 6418 }, { "epoch": 0.12, "learning_rate": 4.792358105441583e-05, "loss": 0.0199, "step": 6420 }, { "epoch": 0.12, "learning_rate": 4.79229340941586e-05, "loss": 0.0222, "step": 6422 }, { "epoch": 0.12, "learning_rate": 4.7922287133901365e-05, "loss": 0.0174, "step": 6424 }, { "epoch": 0.12, "learning_rate": 4.7921640173644134e-05, "loss": 0.0259, "step": 6426 }, { "epoch": 0.12, "learning_rate": 4.79209932133869e-05, "loss": 0.0146, "step": 6428 }, { "epoch": 0.12, "learning_rate": 4.792034625312967e-05, "loss": 0.0477, "step": 6430 }, { "epoch": 0.12, "learning_rate": 4.791969929287244e-05, "loss": 0.0122, "step": 6432 }, { "epoch": 0.12, "learning_rate": 4.791905233261521e-05, "loss": 0.0123, "step": 6434 }, { "epoch": 0.12, "learning_rate": 4.791840537235798e-05, "loss": 0.0222, "step": 6436 }, { "epoch": 0.12, "learning_rate": 4.791775841210075e-05, "loss": 0.0351, "step": 6438 }, { "epoch": 0.12, "learning_rate": 4.791711145184352e-05, "loss": 0.015, "step": 6440 }, { "epoch": 0.13, "learning_rate": 4.791646449158628e-05, "loss": 0.0248, "step": 6442 }, { "epoch": 0.13, "learning_rate": 4.7915817531329056e-05, "loss": 0.0305, "step": 6444 }, { "epoch": 0.13, "learning_rate": 4.791517057107182e-05, "loss": 0.0218, "step": 6446 }, { "epoch": 0.13, "learning_rate": 4.791452361081459e-05, "loss": 0.0188, "step": 6448 }, { "epoch": 0.13, "learning_rate": 4.791387665055736e-05, "loss": 0.0156, "step": 6450 }, { "epoch": 0.13, "learning_rate": 4.7913229690300125e-05, "loss": 0.0217, "step": 6452 }, { "epoch": 0.13, "learning_rate": 4.7912582730042895e-05, "loss": 0.0162, "step": 6454 }, { "epoch": 0.13, "learning_rate": 4.7911935769785664e-05, "loss": 0.0118, "step": 6456 }, { "epoch": 0.13, "learning_rate": 4.791128880952843e-05, "loss": 0.0201, "step": 6458 }, { "epoch": 0.13, "learning_rate": 4.79106418492712e-05, "loss": 0.0165, "step": 6460 }, { "epoch": 0.13, "learning_rate": 4.790999488901397e-05, "loss": 0.0172, "step": 6462 }, { "epoch": 0.13, "learning_rate": 4.790934792875674e-05, "loss": 0.067, "step": 6464 }, { "epoch": 0.13, "learning_rate": 4.790870096849951e-05, "loss": 0.042, "step": 6466 }, { "epoch": 0.13, "learning_rate": 4.790805400824228e-05, "loss": 0.01, "step": 6468 }, { "epoch": 0.13, "learning_rate": 4.790740704798504e-05, "loss": 0.0215, "step": 6470 }, { "epoch": 0.13, "learning_rate": 4.7906760087727816e-05, "loss": 0.0101, "step": 6472 }, { "epoch": 0.13, "learning_rate": 4.790611312747058e-05, "loss": 0.0477, "step": 6474 }, { "epoch": 0.13, "learning_rate": 4.7905466167213355e-05, "loss": 0.0185, "step": 6476 }, { "epoch": 0.13, "learning_rate": 4.790481920695612e-05, "loss": 0.0117, "step": 6478 }, { "epoch": 0.13, "learning_rate": 4.7904172246698886e-05, "loss": 0.0208, "step": 6480 }, { "epoch": 0.13, "learning_rate": 4.790352528644166e-05, "loss": 0.0351, "step": 6482 }, { "epoch": 0.13, "learning_rate": 4.7902878326184424e-05, "loss": 0.0408, "step": 6484 }, { "epoch": 0.13, "learning_rate": 4.790223136592719e-05, "loss": 0.022, "step": 6486 }, { "epoch": 0.13, "learning_rate": 4.790158440566996e-05, "loss": 0.0304, "step": 6488 }, { "epoch": 0.13, "learning_rate": 4.790093744541273e-05, "loss": 0.0132, "step": 6490 }, { "epoch": 0.13, "learning_rate": 4.7900290485155494e-05, "loss": 0.0073, "step": 6492 }, { "epoch": 0.13, "learning_rate": 4.789964352489827e-05, "loss": 0.0239, "step": 6494 }, { "epoch": 0.13, "learning_rate": 4.789899656464104e-05, "loss": 0.0513, "step": 6496 }, { "epoch": 0.13, "learning_rate": 4.789834960438381e-05, "loss": 0.0198, "step": 6498 }, { "epoch": 0.13, "learning_rate": 4.789770264412658e-05, "loss": 0.0174, "step": 6500 }, { "epoch": 0.13, "learning_rate": 4.789705568386934e-05, "loss": 0.0214, "step": 6502 }, { "epoch": 0.13, "learning_rate": 4.7896408723612115e-05, "loss": 0.0105, "step": 6504 }, { "epoch": 0.13, "learning_rate": 4.789576176335488e-05, "loss": 0.0419, "step": 6506 }, { "epoch": 0.13, "learning_rate": 4.7895114803097646e-05, "loss": 0.0175, "step": 6508 }, { "epoch": 0.13, "learning_rate": 4.7894467842840415e-05, "loss": 0.023, "step": 6510 }, { "epoch": 0.13, "learning_rate": 4.7893820882583184e-05, "loss": 0.0116, "step": 6512 }, { "epoch": 0.13, "learning_rate": 4.7893173922325954e-05, "loss": 0.0383, "step": 6514 }, { "epoch": 0.13, "learning_rate": 4.789252696206872e-05, "loss": 0.023, "step": 6516 }, { "epoch": 0.13, "learning_rate": 4.789188000181149e-05, "loss": 0.0404, "step": 6518 }, { "epoch": 0.13, "learning_rate": 4.789123304155426e-05, "loss": 0.028, "step": 6520 }, { "epoch": 0.13, "learning_rate": 4.789058608129703e-05, "loss": 0.0937, "step": 6522 }, { "epoch": 0.13, "learning_rate": 4.788993912103979e-05, "loss": 0.0146, "step": 6524 }, { "epoch": 0.13, "learning_rate": 4.788929216078257e-05, "loss": 0.0218, "step": 6526 }, { "epoch": 0.13, "learning_rate": 4.788864520052534e-05, "loss": 0.0434, "step": 6528 }, { "epoch": 0.13, "learning_rate": 4.78879982402681e-05, "loss": 0.0386, "step": 6530 }, { "epoch": 0.13, "learning_rate": 4.7887351280010875e-05, "loss": 0.0271, "step": 6532 }, { "epoch": 0.13, "learning_rate": 4.788670431975364e-05, "loss": 0.0319, "step": 6534 }, { "epoch": 0.13, "learning_rate": 4.788605735949641e-05, "loss": 0.0476, "step": 6536 }, { "epoch": 0.13, "learning_rate": 4.7885410399239176e-05, "loss": 0.0168, "step": 6538 }, { "epoch": 0.13, "learning_rate": 4.7884763438981945e-05, "loss": 0.0389, "step": 6540 }, { "epoch": 0.13, "learning_rate": 4.7884116478724714e-05, "loss": 0.0219, "step": 6542 }, { "epoch": 0.13, "learning_rate": 4.788346951846748e-05, "loss": 0.0381, "step": 6544 }, { "epoch": 0.13, "learning_rate": 4.788282255821025e-05, "loss": 0.0149, "step": 6546 }, { "epoch": 0.13, "learning_rate": 4.788217559795302e-05, "loss": 0.0276, "step": 6548 }, { "epoch": 0.13, "learning_rate": 4.788152863769579e-05, "loss": 0.0165, "step": 6550 }, { "epoch": 0.13, "learning_rate": 4.788088167743855e-05, "loss": 0.0201, "step": 6552 }, { "epoch": 0.13, "learning_rate": 4.788023471718133e-05, "loss": 0.0069, "step": 6554 }, { "epoch": 0.13, "learning_rate": 4.787958775692409e-05, "loss": 0.0116, "step": 6556 }, { "epoch": 0.13, "learning_rate": 4.787894079666687e-05, "loss": 0.039, "step": 6558 }, { "epoch": 0.13, "learning_rate": 4.787829383640963e-05, "loss": 0.0227, "step": 6560 }, { "epoch": 0.13, "learning_rate": 4.78776468761524e-05, "loss": 0.0313, "step": 6562 }, { "epoch": 0.13, "learning_rate": 4.7876999915895174e-05, "loss": 0.0436, "step": 6564 }, { "epoch": 0.13, "learning_rate": 4.7876352955637936e-05, "loss": 0.0251, "step": 6566 }, { "epoch": 0.13, "learning_rate": 4.7875705995380705e-05, "loss": 0.0131, "step": 6568 }, { "epoch": 0.13, "learning_rate": 4.7875059035123474e-05, "loss": 0.0181, "step": 6570 }, { "epoch": 0.13, "learning_rate": 4.7874412074866243e-05, "loss": 0.0247, "step": 6572 }, { "epoch": 0.13, "learning_rate": 4.787376511460901e-05, "loss": 0.0203, "step": 6574 }, { "epoch": 0.13, "learning_rate": 4.787311815435178e-05, "loss": 0.0121, "step": 6576 }, { "epoch": 0.13, "learning_rate": 4.787247119409455e-05, "loss": 0.0224, "step": 6578 }, { "epoch": 0.13, "learning_rate": 4.787182423383732e-05, "loss": 0.0356, "step": 6580 }, { "epoch": 0.13, "learning_rate": 4.787117727358009e-05, "loss": 0.013, "step": 6582 }, { "epoch": 0.13, "learning_rate": 4.787053031332285e-05, "loss": 0.0486, "step": 6584 }, { "epoch": 0.13, "learning_rate": 4.786988335306563e-05, "loss": 0.0158, "step": 6586 }, { "epoch": 0.13, "learning_rate": 4.786923639280839e-05, "loss": 0.0202, "step": 6588 }, { "epoch": 0.13, "learning_rate": 4.786858943255116e-05, "loss": 0.0261, "step": 6590 }, { "epoch": 0.13, "learning_rate": 4.786794247229393e-05, "loss": 0.0174, "step": 6592 }, { "epoch": 0.13, "learning_rate": 4.7867295512036697e-05, "loss": 0.0271, "step": 6594 }, { "epoch": 0.13, "learning_rate": 4.7866648551779466e-05, "loss": 0.0192, "step": 6596 }, { "epoch": 0.13, "learning_rate": 4.7866001591522235e-05, "loss": 0.0158, "step": 6598 }, { "epoch": 0.13, "learning_rate": 4.7865354631265004e-05, "loss": 0.0197, "step": 6600 }, { "epoch": 0.13, "learning_rate": 4.786470767100777e-05, "loss": 0.0412, "step": 6602 }, { "epoch": 0.13, "learning_rate": 4.786406071075054e-05, "loss": 0.0224, "step": 6604 }, { "epoch": 0.13, "learning_rate": 4.786341375049331e-05, "loss": 0.0217, "step": 6606 }, { "epoch": 0.13, "learning_rate": 4.786276679023608e-05, "loss": 0.0069, "step": 6608 }, { "epoch": 0.13, "learning_rate": 4.786211982997885e-05, "loss": 0.0293, "step": 6610 }, { "epoch": 0.13, "learning_rate": 4.786147286972161e-05, "loss": 0.0172, "step": 6612 }, { "epoch": 0.13, "learning_rate": 4.786082590946439e-05, "loss": 0.0151, "step": 6614 }, { "epoch": 0.13, "learning_rate": 4.786017894920715e-05, "loss": 0.0372, "step": 6616 }, { "epoch": 0.13, "learning_rate": 4.7859531988949926e-05, "loss": 0.0048, "step": 6618 }, { "epoch": 0.13, "learning_rate": 4.785888502869269e-05, "loss": 0.0292, "step": 6620 }, { "epoch": 0.13, "learning_rate": 4.785823806843546e-05, "loss": 0.0166, "step": 6622 }, { "epoch": 0.13, "learning_rate": 4.7857591108178226e-05, "loss": 0.0387, "step": 6624 }, { "epoch": 0.13, "learning_rate": 4.7856944147920995e-05, "loss": 0.0166, "step": 6626 }, { "epoch": 0.13, "learning_rate": 4.7856297187663764e-05, "loss": 0.0291, "step": 6628 }, { "epoch": 0.13, "learning_rate": 4.785565022740653e-05, "loss": 0.0327, "step": 6630 }, { "epoch": 0.13, "learning_rate": 4.78550032671493e-05, "loss": 0.0497, "step": 6632 }, { "epoch": 0.13, "learning_rate": 4.7854356306892065e-05, "loss": 0.0394, "step": 6634 }, { "epoch": 0.13, "learning_rate": 4.785370934663484e-05, "loss": 0.0526, "step": 6636 }, { "epoch": 0.13, "learning_rate": 4.78530623863776e-05, "loss": 0.0191, "step": 6638 }, { "epoch": 0.13, "learning_rate": 4.785241542612038e-05, "loss": 0.0124, "step": 6640 }, { "epoch": 0.13, "learning_rate": 4.785176846586315e-05, "loss": 0.0294, "step": 6642 }, { "epoch": 0.13, "learning_rate": 4.785112150560591e-05, "loss": 0.041, "step": 6644 }, { "epoch": 0.13, "learning_rate": 4.7850474545348686e-05, "loss": 0.0274, "step": 6646 }, { "epoch": 0.13, "learning_rate": 4.784982758509145e-05, "loss": 0.0358, "step": 6648 }, { "epoch": 0.13, "learning_rate": 4.784918062483422e-05, "loss": 0.0363, "step": 6650 }, { "epoch": 0.13, "learning_rate": 4.7848533664576986e-05, "loss": 0.0204, "step": 6652 }, { "epoch": 0.13, "learning_rate": 4.7847886704319756e-05, "loss": 0.0483, "step": 6654 }, { "epoch": 0.13, "learning_rate": 4.7847239744062525e-05, "loss": 0.0157, "step": 6656 }, { "epoch": 0.13, "learning_rate": 4.7846592783805294e-05, "loss": 0.0125, "step": 6658 }, { "epoch": 0.13, "learning_rate": 4.784594582354806e-05, "loss": 0.0151, "step": 6660 }, { "epoch": 0.13, "learning_rate": 4.784529886329083e-05, "loss": 0.0135, "step": 6662 }, { "epoch": 0.13, "learning_rate": 4.78446519030336e-05, "loss": 0.0103, "step": 6664 }, { "epoch": 0.13, "learning_rate": 4.784400494277636e-05, "loss": 0.0123, "step": 6666 }, { "epoch": 0.13, "learning_rate": 4.784335798251914e-05, "loss": 0.0335, "step": 6668 }, { "epoch": 0.13, "learning_rate": 4.78427110222619e-05, "loss": 0.0508, "step": 6670 }, { "epoch": 0.13, "learning_rate": 4.784206406200467e-05, "loss": 0.0219, "step": 6672 }, { "epoch": 0.13, "learning_rate": 4.7841417101747446e-05, "loss": 0.015, "step": 6674 }, { "epoch": 0.13, "learning_rate": 4.784077014149021e-05, "loss": 0.0445, "step": 6676 }, { "epoch": 0.13, "learning_rate": 4.784012318123298e-05, "loss": 0.0388, "step": 6678 }, { "epoch": 0.13, "learning_rate": 4.783947622097575e-05, "loss": 0.0142, "step": 6680 }, { "epoch": 0.13, "learning_rate": 4.7838829260718516e-05, "loss": 0.0115, "step": 6682 }, { "epoch": 0.13, "learning_rate": 4.7838182300461285e-05, "loss": 0.0229, "step": 6684 }, { "epoch": 0.13, "learning_rate": 4.7837535340204054e-05, "loss": 0.0291, "step": 6686 }, { "epoch": 0.13, "learning_rate": 4.783688837994682e-05, "loss": 0.0303, "step": 6688 }, { "epoch": 0.13, "learning_rate": 4.783624141968959e-05, "loss": 0.0255, "step": 6690 }, { "epoch": 0.13, "learning_rate": 4.783559445943236e-05, "loss": 0.0364, "step": 6692 }, { "epoch": 0.13, "learning_rate": 4.7834947499175124e-05, "loss": 0.0239, "step": 6694 }, { "epoch": 0.13, "learning_rate": 4.78343005389179e-05, "loss": 0.0277, "step": 6696 }, { "epoch": 0.13, "learning_rate": 4.783365357866066e-05, "loss": 0.015, "step": 6698 }, { "epoch": 0.13, "learning_rate": 4.783300661840344e-05, "loss": 0.0234, "step": 6700 }, { "epoch": 0.13, "learning_rate": 4.78323596581462e-05, "loss": 0.0112, "step": 6702 }, { "epoch": 0.13, "learning_rate": 4.783171269788897e-05, "loss": 0.0213, "step": 6704 }, { "epoch": 0.13, "learning_rate": 4.7831065737631745e-05, "loss": 0.0365, "step": 6706 }, { "epoch": 0.13, "learning_rate": 4.783041877737451e-05, "loss": 0.0369, "step": 6708 }, { "epoch": 0.13, "learning_rate": 4.7829771817117276e-05, "loss": 0.0123, "step": 6710 }, { "epoch": 0.13, "learning_rate": 4.7829124856860045e-05, "loss": 0.0128, "step": 6712 }, { "epoch": 0.13, "learning_rate": 4.7828477896602815e-05, "loss": 0.0169, "step": 6714 }, { "epoch": 0.13, "learning_rate": 4.782783093634558e-05, "loss": 0.0272, "step": 6716 }, { "epoch": 0.13, "learning_rate": 4.782718397608835e-05, "loss": 0.0273, "step": 6718 }, { "epoch": 0.13, "learning_rate": 4.782653701583112e-05, "loss": 0.0236, "step": 6720 }, { "epoch": 0.13, "learning_rate": 4.782589005557389e-05, "loss": 0.0202, "step": 6722 }, { "epoch": 0.13, "learning_rate": 4.782524309531666e-05, "loss": 0.0229, "step": 6724 }, { "epoch": 0.13, "learning_rate": 4.782459613505942e-05, "loss": 0.0428, "step": 6726 }, { "epoch": 0.13, "learning_rate": 4.78239491748022e-05, "loss": 0.0292, "step": 6728 }, { "epoch": 0.13, "learning_rate": 4.782330221454496e-05, "loss": 0.0324, "step": 6730 }, { "epoch": 0.13, "learning_rate": 4.782265525428773e-05, "loss": 0.0188, "step": 6732 }, { "epoch": 0.13, "learning_rate": 4.78220082940305e-05, "loss": 0.0109, "step": 6734 }, { "epoch": 0.13, "learning_rate": 4.782136133377327e-05, "loss": 0.0379, "step": 6736 }, { "epoch": 0.13, "learning_rate": 4.782071437351604e-05, "loss": 0.0413, "step": 6738 }, { "epoch": 0.13, "learning_rate": 4.7820067413258806e-05, "loss": 0.0515, "step": 6740 }, { "epoch": 0.13, "learning_rate": 4.7819420453001575e-05, "loss": 0.0483, "step": 6742 }, { "epoch": 0.13, "learning_rate": 4.7818773492744344e-05, "loss": 0.0153, "step": 6744 }, { "epoch": 0.13, "learning_rate": 4.781812653248711e-05, "loss": 0.0177, "step": 6746 }, { "epoch": 0.13, "learning_rate": 4.7817479572229875e-05, "loss": 0.0173, "step": 6748 }, { "epoch": 0.13, "learning_rate": 4.781683261197265e-05, "loss": 0.0181, "step": 6750 }, { "epoch": 0.13, "learning_rate": 4.781618565171542e-05, "loss": 0.0189, "step": 6752 }, { "epoch": 0.13, "learning_rate": 4.781553869145818e-05, "loss": 0.0207, "step": 6754 }, { "epoch": 0.13, "learning_rate": 4.781489173120096e-05, "loss": 0.0163, "step": 6756 }, { "epoch": 0.13, "learning_rate": 4.781424477094372e-05, "loss": 0.015, "step": 6758 }, { "epoch": 0.13, "learning_rate": 4.78135978106865e-05, "loss": 0.007, "step": 6760 }, { "epoch": 0.13, "learning_rate": 4.781295085042926e-05, "loss": 0.0162, "step": 6762 }, { "epoch": 0.13, "learning_rate": 4.781230389017203e-05, "loss": 0.0266, "step": 6764 }, { "epoch": 0.13, "learning_rate": 4.78116569299148e-05, "loss": 0.0128, "step": 6766 }, { "epoch": 0.13, "learning_rate": 4.7811009969657566e-05, "loss": 0.0189, "step": 6768 }, { "epoch": 0.13, "learning_rate": 4.7810363009400335e-05, "loss": 0.0127, "step": 6770 }, { "epoch": 0.13, "learning_rate": 4.7809716049143104e-05, "loss": 0.0219, "step": 6772 }, { "epoch": 0.13, "learning_rate": 4.7809069088885874e-05, "loss": 0.083, "step": 6774 }, { "epoch": 0.13, "learning_rate": 4.7808422128628636e-05, "loss": 0.0124, "step": 6776 }, { "epoch": 0.13, "learning_rate": 4.780777516837141e-05, "loss": 0.0322, "step": 6778 }, { "epoch": 0.13, "learning_rate": 4.7807128208114174e-05, "loss": 0.0286, "step": 6780 }, { "epoch": 0.13, "learning_rate": 4.780648124785695e-05, "loss": 0.0257, "step": 6782 }, { "epoch": 0.13, "learning_rate": 4.780583428759972e-05, "loss": 0.0231, "step": 6784 }, { "epoch": 0.13, "learning_rate": 4.780518732734248e-05, "loss": 0.0088, "step": 6786 }, { "epoch": 0.13, "learning_rate": 4.780454036708526e-05, "loss": 0.0203, "step": 6788 }, { "epoch": 0.13, "learning_rate": 4.780389340682802e-05, "loss": 0.0332, "step": 6790 }, { "epoch": 0.13, "learning_rate": 4.780324644657079e-05, "loss": 0.0193, "step": 6792 }, { "epoch": 0.13, "learning_rate": 4.780259948631356e-05, "loss": 0.0116, "step": 6794 }, { "epoch": 0.13, "learning_rate": 4.780195252605633e-05, "loss": 0.0129, "step": 6796 }, { "epoch": 0.13, "learning_rate": 4.7801305565799096e-05, "loss": 0.0193, "step": 6798 }, { "epoch": 0.13, "learning_rate": 4.7800658605541865e-05, "loss": 0.0146, "step": 6800 }, { "epoch": 0.13, "learning_rate": 4.7800011645284634e-05, "loss": 0.0169, "step": 6802 }, { "epoch": 0.13, "learning_rate": 4.77993646850274e-05, "loss": 0.0153, "step": 6804 }, { "epoch": 0.13, "learning_rate": 4.779871772477017e-05, "loss": 0.0237, "step": 6806 }, { "epoch": 0.13, "learning_rate": 4.7798070764512934e-05, "loss": 0.0085, "step": 6808 }, { "epoch": 0.13, "learning_rate": 4.779742380425571e-05, "loss": 0.0526, "step": 6810 }, { "epoch": 0.13, "learning_rate": 4.779677684399847e-05, "loss": 0.0113, "step": 6812 }, { "epoch": 0.13, "learning_rate": 4.779612988374124e-05, "loss": 0.0172, "step": 6814 }, { "epoch": 0.13, "learning_rate": 4.779548292348401e-05, "loss": 0.0079, "step": 6816 }, { "epoch": 0.13, "learning_rate": 4.779483596322678e-05, "loss": 0.0429, "step": 6818 }, { "epoch": 0.13, "learning_rate": 4.779418900296955e-05, "loss": 0.0163, "step": 6820 }, { "epoch": 0.13, "learning_rate": 4.779354204271232e-05, "loss": 0.0043, "step": 6822 }, { "epoch": 0.13, "learning_rate": 4.779289508245509e-05, "loss": 0.0096, "step": 6824 }, { "epoch": 0.13, "learning_rate": 4.7792248122197856e-05, "loss": 0.0476, "step": 6826 }, { "epoch": 0.13, "learning_rate": 4.7791601161940625e-05, "loss": 0.0169, "step": 6828 }, { "epoch": 0.13, "learning_rate": 4.7790954201683394e-05, "loss": 0.0396, "step": 6830 }, { "epoch": 0.13, "learning_rate": 4.7790307241426163e-05, "loss": 0.0101, "step": 6832 }, { "epoch": 0.13, "learning_rate": 4.778966028116893e-05, "loss": 0.029, "step": 6834 }, { "epoch": 0.13, "learning_rate": 4.7789013320911695e-05, "loss": 0.0169, "step": 6836 }, { "epoch": 0.13, "learning_rate": 4.778836636065447e-05, "loss": 0.0816, "step": 6838 }, { "epoch": 0.13, "learning_rate": 4.778771940039723e-05, "loss": 0.0233, "step": 6840 }, { "epoch": 0.13, "learning_rate": 4.778707244014001e-05, "loss": 0.006, "step": 6842 }, { "epoch": 0.13, "learning_rate": 4.778642547988277e-05, "loss": 0.018, "step": 6844 }, { "epoch": 0.13, "learning_rate": 4.778577851962554e-05, "loss": 0.03, "step": 6846 }, { "epoch": 0.13, "learning_rate": 4.778513155936831e-05, "loss": 0.0205, "step": 6848 }, { "epoch": 0.13, "learning_rate": 4.778448459911108e-05, "loss": 0.0056, "step": 6850 }, { "epoch": 0.13, "learning_rate": 4.778383763885385e-05, "loss": 0.015, "step": 6852 }, { "epoch": 0.13, "learning_rate": 4.7783190678596617e-05, "loss": 0.0148, "step": 6854 }, { "epoch": 0.13, "learning_rate": 4.7782543718339386e-05, "loss": 0.0457, "step": 6856 }, { "epoch": 0.13, "learning_rate": 4.778189675808215e-05, "loss": 0.0365, "step": 6858 }, { "epoch": 0.13, "learning_rate": 4.7781249797824924e-05, "loss": 0.026, "step": 6860 }, { "epoch": 0.13, "learning_rate": 4.7780602837567686e-05, "loss": 0.0242, "step": 6862 }, { "epoch": 0.13, "learning_rate": 4.777995587731046e-05, "loss": 0.0284, "step": 6864 }, { "epoch": 0.13, "learning_rate": 4.777930891705323e-05, "loss": 0.0149, "step": 6866 }, { "epoch": 0.13, "learning_rate": 4.7778661956795993e-05, "loss": 0.0269, "step": 6868 }, { "epoch": 0.13, "learning_rate": 4.777801499653877e-05, "loss": 0.0117, "step": 6870 }, { "epoch": 0.13, "learning_rate": 4.777736803628153e-05, "loss": 0.0238, "step": 6872 }, { "epoch": 0.13, "learning_rate": 4.77767210760243e-05, "loss": 0.0352, "step": 6874 }, { "epoch": 0.13, "learning_rate": 4.777607411576707e-05, "loss": 0.0222, "step": 6876 }, { "epoch": 0.13, "learning_rate": 4.777542715550984e-05, "loss": 0.0582, "step": 6878 }, { "epoch": 0.13, "learning_rate": 4.777478019525261e-05, "loss": 0.1249, "step": 6880 }, { "epoch": 0.13, "learning_rate": 4.777413323499538e-05, "loss": 0.021, "step": 6882 }, { "epoch": 0.13, "learning_rate": 4.7773486274738146e-05, "loss": 0.0414, "step": 6884 }, { "epoch": 0.13, "learning_rate": 4.7772839314480915e-05, "loss": 0.0255, "step": 6886 }, { "epoch": 0.13, "learning_rate": 4.7772192354223684e-05, "loss": 0.0253, "step": 6888 }, { "epoch": 0.13, "learning_rate": 4.7771545393966447e-05, "loss": 0.0274, "step": 6890 }, { "epoch": 0.13, "learning_rate": 4.777089843370922e-05, "loss": 0.0193, "step": 6892 }, { "epoch": 0.13, "learning_rate": 4.7770251473451985e-05, "loss": 0.0182, "step": 6894 }, { "epoch": 0.13, "learning_rate": 4.7769604513194754e-05, "loss": 0.0404, "step": 6896 }, { "epoch": 0.13, "learning_rate": 4.776895755293753e-05, "loss": 0.0106, "step": 6898 }, { "epoch": 0.13, "learning_rate": 4.776831059268029e-05, "loss": 0.0211, "step": 6900 }, { "epoch": 0.13, "learning_rate": 4.776766363242307e-05, "loss": 0.0184, "step": 6902 }, { "epoch": 0.13, "learning_rate": 4.776701667216583e-05, "loss": 0.0205, "step": 6904 }, { "epoch": 0.13, "learning_rate": 4.77663697119086e-05, "loss": 0.0225, "step": 6906 }, { "epoch": 0.13, "learning_rate": 4.776572275165137e-05, "loss": 0.0231, "step": 6908 }, { "epoch": 0.13, "learning_rate": 4.776507579139414e-05, "loss": 0.0089, "step": 6910 }, { "epoch": 0.13, "learning_rate": 4.7764428831136906e-05, "loss": 0.0126, "step": 6912 }, { "epoch": 0.13, "learning_rate": 4.7763781870879676e-05, "loss": 0.0238, "step": 6914 }, { "epoch": 0.13, "learning_rate": 4.7763134910622445e-05, "loss": 0.0191, "step": 6916 }, { "epoch": 0.13, "learning_rate": 4.776248795036521e-05, "loss": 0.0224, "step": 6918 }, { "epoch": 0.13, "learning_rate": 4.776184099010798e-05, "loss": 0.0422, "step": 6920 }, { "epoch": 0.13, "learning_rate": 4.7761194029850745e-05, "loss": 0.0286, "step": 6922 }, { "epoch": 0.13, "learning_rate": 4.776054706959352e-05, "loss": 0.0162, "step": 6924 }, { "epoch": 0.13, "learning_rate": 4.775990010933628e-05, "loss": 0.0306, "step": 6926 }, { "epoch": 0.13, "learning_rate": 4.775925314907905e-05, "loss": 0.0135, "step": 6928 }, { "epoch": 0.13, "learning_rate": 4.775860618882183e-05, "loss": 0.0306, "step": 6930 }, { "epoch": 0.13, "learning_rate": 4.775795922856459e-05, "loss": 0.0348, "step": 6932 }, { "epoch": 0.13, "learning_rate": 4.775731226830736e-05, "loss": 0.0239, "step": 6934 }, { "epoch": 0.13, "learning_rate": 4.775666530805013e-05, "loss": 0.0232, "step": 6936 }, { "epoch": 0.13, "learning_rate": 4.77560183477929e-05, "loss": 0.0386, "step": 6938 }, { "epoch": 0.13, "learning_rate": 4.775537138753566e-05, "loss": 0.0145, "step": 6940 }, { "epoch": 0.13, "learning_rate": 4.7754724427278436e-05, "loss": 0.0552, "step": 6942 }, { "epoch": 0.13, "learning_rate": 4.7754077467021205e-05, "loss": 0.0081, "step": 6944 }, { "epoch": 0.13, "learning_rate": 4.7753430506763974e-05, "loss": 0.0304, "step": 6946 }, { "epoch": 0.13, "learning_rate": 4.775278354650674e-05, "loss": 0.0167, "step": 6948 }, { "epoch": 0.13, "learning_rate": 4.7752136586249506e-05, "loss": 0.0123, "step": 6950 }, { "epoch": 0.13, "learning_rate": 4.775148962599228e-05, "loss": 0.0122, "step": 6952 }, { "epoch": 0.13, "learning_rate": 4.7750842665735044e-05, "loss": 0.0289, "step": 6954 }, { "epoch": 0.14, "learning_rate": 4.775019570547781e-05, "loss": 0.0274, "step": 6956 }, { "epoch": 0.14, "learning_rate": 4.774954874522058e-05, "loss": 0.02, "step": 6958 }, { "epoch": 0.14, "learning_rate": 4.774890178496335e-05, "loss": 0.0344, "step": 6960 }, { "epoch": 0.14, "learning_rate": 4.774825482470612e-05, "loss": 0.021, "step": 6962 }, { "epoch": 0.14, "learning_rate": 4.774760786444889e-05, "loss": 0.047, "step": 6964 }, { "epoch": 0.14, "learning_rate": 4.774696090419166e-05, "loss": 0.0466, "step": 6966 }, { "epoch": 0.14, "learning_rate": 4.774631394393443e-05, "loss": 0.0157, "step": 6968 }, { "epoch": 0.14, "learning_rate": 4.7745666983677196e-05, "loss": 0.0239, "step": 6970 }, { "epoch": 0.14, "learning_rate": 4.774502002341996e-05, "loss": 0.0326, "step": 6972 }, { "epoch": 0.14, "learning_rate": 4.7744373063162735e-05, "loss": 0.0128, "step": 6974 }, { "epoch": 0.14, "learning_rate": 4.7743726102905504e-05, "loss": 0.0214, "step": 6976 }, { "epoch": 0.14, "learning_rate": 4.7743079142648266e-05, "loss": 0.0116, "step": 6978 }, { "epoch": 0.14, "learning_rate": 4.774243218239104e-05, "loss": 0.0174, "step": 6980 }, { "epoch": 0.14, "learning_rate": 4.7741785222133804e-05, "loss": 0.0141, "step": 6982 }, { "epoch": 0.14, "learning_rate": 4.774113826187658e-05, "loss": 0.0174, "step": 6984 }, { "epoch": 0.14, "learning_rate": 4.774049130161934e-05, "loss": 0.0232, "step": 6986 }, { "epoch": 0.14, "learning_rate": 4.773984434136211e-05, "loss": 0.0188, "step": 6988 }, { "epoch": 0.14, "learning_rate": 4.773919738110488e-05, "loss": 0.0468, "step": 6990 }, { "epoch": 0.14, "learning_rate": 4.773855042084765e-05, "loss": 0.019, "step": 6992 }, { "epoch": 0.14, "learning_rate": 4.773790346059042e-05, "loss": 0.0157, "step": 6994 }, { "epoch": 0.14, "learning_rate": 4.773725650033319e-05, "loss": 0.029, "step": 6996 }, { "epoch": 0.14, "learning_rate": 4.773660954007596e-05, "loss": 0.029, "step": 6998 }, { "epoch": 0.14, "learning_rate": 4.773596257981872e-05, "loss": 0.0361, "step": 7000 }, { "epoch": 0.14, "learning_rate": 4.7735315619561495e-05, "loss": 0.0092, "step": 7002 }, { "epoch": 0.14, "learning_rate": 4.773466865930426e-05, "loss": 0.0357, "step": 7004 }, { "epoch": 0.14, "learning_rate": 4.773402169904703e-05, "loss": 0.0049, "step": 7006 }, { "epoch": 0.14, "learning_rate": 4.77333747387898e-05, "loss": 0.0279, "step": 7008 }, { "epoch": 0.14, "learning_rate": 4.7732727778532565e-05, "loss": 0.0284, "step": 7010 }, { "epoch": 0.14, "learning_rate": 4.773208081827534e-05, "loss": 0.0144, "step": 7012 }, { "epoch": 0.14, "learning_rate": 4.77314338580181e-05, "loss": 0.0225, "step": 7014 }, { "epoch": 0.14, "learning_rate": 4.773078689776087e-05, "loss": 0.0224, "step": 7016 }, { "epoch": 0.14, "learning_rate": 4.773013993750364e-05, "loss": 0.0049, "step": 7018 }, { "epoch": 0.14, "learning_rate": 4.772949297724641e-05, "loss": 0.0256, "step": 7020 }, { "epoch": 0.14, "learning_rate": 4.772884601698918e-05, "loss": 0.0386, "step": 7022 }, { "epoch": 0.14, "learning_rate": 4.772819905673195e-05, "loss": 0.008, "step": 7024 }, { "epoch": 0.14, "learning_rate": 4.772755209647472e-05, "loss": 0.0058, "step": 7026 }, { "epoch": 0.14, "learning_rate": 4.7726905136217486e-05, "loss": 0.018, "step": 7028 }, { "epoch": 0.14, "learning_rate": 4.7726258175960255e-05, "loss": 0.0099, "step": 7030 }, { "epoch": 0.14, "learning_rate": 4.772561121570302e-05, "loss": 0.0134, "step": 7032 }, { "epoch": 0.14, "learning_rate": 4.7724964255445794e-05, "loss": 0.0244, "step": 7034 }, { "epoch": 0.14, "learning_rate": 4.7724317295188556e-05, "loss": 0.0359, "step": 7036 }, { "epoch": 0.14, "learning_rate": 4.7723670334931325e-05, "loss": 0.0332, "step": 7038 }, { "epoch": 0.14, "learning_rate": 4.7723023374674094e-05, "loss": 0.0409, "step": 7040 }, { "epoch": 0.14, "learning_rate": 4.772237641441686e-05, "loss": 0.017, "step": 7042 }, { "epoch": 0.14, "learning_rate": 4.772172945415964e-05, "loss": 0.0132, "step": 7044 }, { "epoch": 0.14, "learning_rate": 4.77210824939024e-05, "loss": 0.0231, "step": 7046 }, { "epoch": 0.14, "learning_rate": 4.772043553364517e-05, "loss": 0.0383, "step": 7048 }, { "epoch": 0.14, "learning_rate": 4.771978857338794e-05, "loss": 0.045, "step": 7050 }, { "epoch": 0.14, "learning_rate": 4.771914161313071e-05, "loss": 0.0073, "step": 7052 }, { "epoch": 0.14, "learning_rate": 4.771849465287348e-05, "loss": 0.0294, "step": 7054 }, { "epoch": 0.14, "learning_rate": 4.771784769261625e-05, "loss": 0.0233, "step": 7056 }, { "epoch": 0.14, "learning_rate": 4.7717200732359016e-05, "loss": 0.0175, "step": 7058 }, { "epoch": 0.14, "learning_rate": 4.771655377210178e-05, "loss": 0.0152, "step": 7060 }, { "epoch": 0.14, "learning_rate": 4.7715906811844554e-05, "loss": 0.0178, "step": 7062 }, { "epoch": 0.14, "learning_rate": 4.7715259851587316e-05, "loss": 0.0496, "step": 7064 }, { "epoch": 0.14, "learning_rate": 4.771461289133009e-05, "loss": 0.0457, "step": 7066 }, { "epoch": 0.14, "learning_rate": 4.7713965931072854e-05, "loss": 0.0258, "step": 7068 }, { "epoch": 0.14, "learning_rate": 4.7713318970815623e-05, "loss": 0.0272, "step": 7070 }, { "epoch": 0.14, "learning_rate": 4.771267201055839e-05, "loss": 0.0247, "step": 7072 }, { "epoch": 0.14, "learning_rate": 4.771202505030116e-05, "loss": 0.0488, "step": 7074 }, { "epoch": 0.14, "learning_rate": 4.771137809004393e-05, "loss": 0.02, "step": 7076 }, { "epoch": 0.14, "learning_rate": 4.77107311297867e-05, "loss": 0.0497, "step": 7078 }, { "epoch": 0.14, "learning_rate": 4.771008416952947e-05, "loss": 0.0424, "step": 7080 }, { "epoch": 0.14, "learning_rate": 4.770943720927223e-05, "loss": 0.0198, "step": 7082 }, { "epoch": 0.14, "learning_rate": 4.770879024901501e-05, "loss": 0.006, "step": 7084 }, { "epoch": 0.14, "learning_rate": 4.7708143288757776e-05, "loss": 0.0182, "step": 7086 }, { "epoch": 0.14, "learning_rate": 4.7707496328500545e-05, "loss": 0.0232, "step": 7088 }, { "epoch": 0.14, "learning_rate": 4.7706849368243314e-05, "loss": 0.0359, "step": 7090 }, { "epoch": 0.14, "learning_rate": 4.770620240798608e-05, "loss": 0.0559, "step": 7092 }, { "epoch": 0.14, "learning_rate": 4.770555544772885e-05, "loss": 0.0132, "step": 7094 }, { "epoch": 0.14, "learning_rate": 4.7704908487471615e-05, "loss": 0.0339, "step": 7096 }, { "epoch": 0.14, "learning_rate": 4.7704261527214384e-05, "loss": 0.0324, "step": 7098 }, { "epoch": 0.14, "learning_rate": 4.770361456695715e-05, "loss": 0.0264, "step": 7100 }, { "epoch": 0.14, "learning_rate": 4.770296760669992e-05, "loss": 0.0158, "step": 7102 }, { "epoch": 0.14, "learning_rate": 4.770232064644269e-05, "loss": 0.0045, "step": 7104 }, { "epoch": 0.14, "learning_rate": 4.770167368618546e-05, "loss": 0.0077, "step": 7106 }, { "epoch": 0.14, "learning_rate": 4.770102672592823e-05, "loss": 0.0217, "step": 7108 }, { "epoch": 0.14, "learning_rate": 4.7700379765671e-05, "loss": 0.0153, "step": 7110 }, { "epoch": 0.14, "learning_rate": 4.769973280541377e-05, "loss": 0.02, "step": 7112 }, { "epoch": 0.14, "learning_rate": 4.769908584515653e-05, "loss": 0.0135, "step": 7114 }, { "epoch": 0.14, "learning_rate": 4.7698438884899306e-05, "loss": 0.0139, "step": 7116 }, { "epoch": 0.14, "learning_rate": 4.769779192464207e-05, "loss": 0.0198, "step": 7118 }, { "epoch": 0.14, "learning_rate": 4.769714496438484e-05, "loss": 0.0117, "step": 7120 }, { "epoch": 0.14, "learning_rate": 4.769649800412761e-05, "loss": 0.0201, "step": 7122 }, { "epoch": 0.14, "learning_rate": 4.7695851043870375e-05, "loss": 0.0245, "step": 7124 }, { "epoch": 0.14, "learning_rate": 4.769520408361315e-05, "loss": 0.0053, "step": 7126 }, { "epoch": 0.14, "learning_rate": 4.769455712335591e-05, "loss": 0.0198, "step": 7128 }, { "epoch": 0.14, "learning_rate": 4.769391016309868e-05, "loss": 0.0052, "step": 7130 }, { "epoch": 0.14, "learning_rate": 4.769326320284145e-05, "loss": 0.024, "step": 7132 }, { "epoch": 0.14, "learning_rate": 4.769261624258422e-05, "loss": 0.023, "step": 7134 }, { "epoch": 0.14, "learning_rate": 4.769196928232699e-05, "loss": 0.1199, "step": 7136 }, { "epoch": 0.14, "learning_rate": 4.769132232206976e-05, "loss": 0.0178, "step": 7138 }, { "epoch": 0.14, "learning_rate": 4.769067536181253e-05, "loss": 0.0132, "step": 7140 }, { "epoch": 0.14, "learning_rate": 4.769002840155529e-05, "loss": 0.0278, "step": 7142 }, { "epoch": 0.14, "learning_rate": 4.7689381441298066e-05, "loss": 0.0099, "step": 7144 }, { "epoch": 0.14, "learning_rate": 4.768873448104083e-05, "loss": 0.0147, "step": 7146 }, { "epoch": 0.14, "learning_rate": 4.7688087520783604e-05, "loss": 0.0102, "step": 7148 }, { "epoch": 0.14, "learning_rate": 4.7687440560526367e-05, "loss": 0.0437, "step": 7150 }, { "epoch": 0.14, "learning_rate": 4.7686793600269136e-05, "loss": 0.0446, "step": 7152 }, { "epoch": 0.14, "learning_rate": 4.768614664001191e-05, "loss": 0.03, "step": 7154 }, { "epoch": 0.14, "learning_rate": 4.7685499679754674e-05, "loss": 0.0201, "step": 7156 }, { "epoch": 0.14, "learning_rate": 4.768485271949744e-05, "loss": 0.0124, "step": 7158 }, { "epoch": 0.14, "learning_rate": 4.768420575924021e-05, "loss": 0.0102, "step": 7160 }, { "epoch": 0.14, "learning_rate": 4.768355879898298e-05, "loss": 0.0119, "step": 7162 }, { "epoch": 0.14, "learning_rate": 4.768291183872575e-05, "loss": 0.0498, "step": 7164 }, { "epoch": 0.14, "learning_rate": 4.768226487846852e-05, "loss": 0.0372, "step": 7166 }, { "epoch": 0.14, "learning_rate": 4.768161791821129e-05, "loss": 0.0268, "step": 7168 }, { "epoch": 0.14, "learning_rate": 4.768097095795406e-05, "loss": 0.0177, "step": 7170 }, { "epoch": 0.14, "learning_rate": 4.7680323997696826e-05, "loss": 0.0334, "step": 7172 }, { "epoch": 0.14, "learning_rate": 4.767967703743959e-05, "loss": 0.0132, "step": 7174 }, { "epoch": 0.14, "learning_rate": 4.7679030077182365e-05, "loss": 0.0117, "step": 7176 }, { "epoch": 0.14, "learning_rate": 4.767838311692513e-05, "loss": 0.0133, "step": 7178 }, { "epoch": 0.14, "learning_rate": 4.7677736156667896e-05, "loss": 0.0264, "step": 7180 }, { "epoch": 0.14, "learning_rate": 4.7677089196410665e-05, "loss": 0.0223, "step": 7182 }, { "epoch": 0.14, "learning_rate": 4.7676442236153434e-05, "loss": 0.0218, "step": 7184 }, { "epoch": 0.14, "learning_rate": 4.76757952758962e-05, "loss": 0.0357, "step": 7186 }, { "epoch": 0.14, "learning_rate": 4.767514831563897e-05, "loss": 0.0187, "step": 7188 }, { "epoch": 0.14, "learning_rate": 4.767450135538174e-05, "loss": 0.0039, "step": 7190 }, { "epoch": 0.14, "learning_rate": 4.767385439512451e-05, "loss": 0.0356, "step": 7192 }, { "epoch": 0.14, "learning_rate": 4.767320743486728e-05, "loss": 0.0125, "step": 7194 }, { "epoch": 0.14, "learning_rate": 4.767256047461004e-05, "loss": 0.0235, "step": 7196 }, { "epoch": 0.14, "learning_rate": 4.767191351435282e-05, "loss": 0.0293, "step": 7198 }, { "epoch": 0.14, "learning_rate": 4.767126655409559e-05, "loss": 0.0128, "step": 7200 }, { "epoch": 0.14, "learning_rate": 4.767061959383835e-05, "loss": 0.0172, "step": 7202 }, { "epoch": 0.14, "learning_rate": 4.7669972633581125e-05, "loss": 0.0113, "step": 7204 }, { "epoch": 0.14, "learning_rate": 4.766932567332389e-05, "loss": 0.0265, "step": 7206 }, { "epoch": 0.14, "learning_rate": 4.766867871306666e-05, "loss": 0.0068, "step": 7208 }, { "epoch": 0.14, "learning_rate": 4.7668031752809426e-05, "loss": 0.0198, "step": 7210 }, { "epoch": 0.14, "learning_rate": 4.7667384792552195e-05, "loss": 0.0261, "step": 7212 }, { "epoch": 0.14, "learning_rate": 4.7666737832294964e-05, "loss": 0.0053, "step": 7214 }, { "epoch": 0.14, "learning_rate": 4.766609087203773e-05, "loss": 0.0319, "step": 7216 }, { "epoch": 0.14, "learning_rate": 4.76654439117805e-05, "loss": 0.0185, "step": 7218 }, { "epoch": 0.14, "learning_rate": 4.766479695152327e-05, "loss": 0.018, "step": 7220 }, { "epoch": 0.14, "learning_rate": 4.766414999126604e-05, "loss": 0.0281, "step": 7222 }, { "epoch": 0.14, "learning_rate": 4.76635030310088e-05, "loss": 0.0219, "step": 7224 }, { "epoch": 0.14, "learning_rate": 4.766285607075158e-05, "loss": 0.0404, "step": 7226 }, { "epoch": 0.14, "learning_rate": 4.766220911049434e-05, "loss": 0.0525, "step": 7228 }, { "epoch": 0.14, "learning_rate": 4.7661562150237116e-05, "loss": 0.0215, "step": 7230 }, { "epoch": 0.14, "learning_rate": 4.7660915189979885e-05, "loss": 0.0255, "step": 7232 }, { "epoch": 0.14, "learning_rate": 4.766026822972265e-05, "loss": 0.0077, "step": 7234 }, { "epoch": 0.14, "learning_rate": 4.7659621269465424e-05, "loss": 0.0136, "step": 7236 }, { "epoch": 0.14, "learning_rate": 4.7658974309208186e-05, "loss": 0.0437, "step": 7238 }, { "epoch": 0.14, "learning_rate": 4.7658327348950955e-05, "loss": 0.0134, "step": 7240 }, { "epoch": 0.14, "learning_rate": 4.7657680388693724e-05, "loss": 0.0084, "step": 7242 }, { "epoch": 0.14, "learning_rate": 4.765703342843649e-05, "loss": 0.0262, "step": 7244 }, { "epoch": 0.14, "learning_rate": 4.765638646817926e-05, "loss": 0.0246, "step": 7246 }, { "epoch": 0.14, "learning_rate": 4.765573950792203e-05, "loss": 0.0285, "step": 7248 }, { "epoch": 0.14, "learning_rate": 4.76550925476648e-05, "loss": 0.0191, "step": 7250 }, { "epoch": 0.14, "learning_rate": 4.765444558740757e-05, "loss": 0.0124, "step": 7252 }, { "epoch": 0.14, "learning_rate": 4.765379862715034e-05, "loss": 0.0206, "step": 7254 }, { "epoch": 0.14, "learning_rate": 4.76531516668931e-05, "loss": 0.0395, "step": 7256 }, { "epoch": 0.14, "learning_rate": 4.765250470663588e-05, "loss": 0.0092, "step": 7258 }, { "epoch": 0.14, "learning_rate": 4.765185774637864e-05, "loss": 0.0073, "step": 7260 }, { "epoch": 0.14, "learning_rate": 4.765121078612141e-05, "loss": 0.0113, "step": 7262 }, { "epoch": 0.14, "learning_rate": 4.7650563825864184e-05, "loss": 0.0454, "step": 7264 }, { "epoch": 0.14, "learning_rate": 4.7649916865606946e-05, "loss": 0.0121, "step": 7266 }, { "epoch": 0.14, "learning_rate": 4.764926990534972e-05, "loss": 0.0094, "step": 7268 }, { "epoch": 0.14, "learning_rate": 4.7648622945092484e-05, "loss": 0.0101, "step": 7270 }, { "epoch": 0.14, "learning_rate": 4.7647975984835254e-05, "loss": 0.0424, "step": 7272 }, { "epoch": 0.14, "learning_rate": 4.764732902457802e-05, "loss": 0.007, "step": 7274 }, { "epoch": 0.14, "learning_rate": 4.764668206432079e-05, "loss": 0.0197, "step": 7276 }, { "epoch": 0.14, "learning_rate": 4.764603510406356e-05, "loss": 0.0225, "step": 7278 }, { "epoch": 0.14, "learning_rate": 4.764538814380633e-05, "loss": 0.0117, "step": 7280 }, { "epoch": 0.14, "learning_rate": 4.76447411835491e-05, "loss": 0.0154, "step": 7282 }, { "epoch": 0.14, "learning_rate": 4.764409422329186e-05, "loss": 0.009, "step": 7284 }, { "epoch": 0.14, "learning_rate": 4.764344726303464e-05, "loss": 0.014, "step": 7286 }, { "epoch": 0.14, "learning_rate": 4.76428003027774e-05, "loss": 0.009, "step": 7288 }, { "epoch": 0.14, "learning_rate": 4.7642153342520175e-05, "loss": 0.013, "step": 7290 }, { "epoch": 0.14, "learning_rate": 4.764150638226294e-05, "loss": 0.0145, "step": 7292 }, { "epoch": 0.14, "learning_rate": 4.764085942200571e-05, "loss": 0.0142, "step": 7294 }, { "epoch": 0.14, "learning_rate": 4.7640212461748476e-05, "loss": 0.0133, "step": 7296 }, { "epoch": 0.14, "learning_rate": 4.7639565501491245e-05, "loss": 0.0203, "step": 7298 }, { "epoch": 0.14, "learning_rate": 4.7638918541234014e-05, "loss": 0.0149, "step": 7300 }, { "epoch": 0.14, "learning_rate": 4.763827158097678e-05, "loss": 0.0238, "step": 7302 }, { "epoch": 0.14, "learning_rate": 4.763762462071955e-05, "loss": 0.0067, "step": 7304 }, { "epoch": 0.14, "learning_rate": 4.7636977660462314e-05, "loss": 0.0394, "step": 7306 }, { "epoch": 0.14, "learning_rate": 4.763633070020509e-05, "loss": 0.0141, "step": 7308 }, { "epoch": 0.14, "learning_rate": 4.763568373994786e-05, "loss": 0.0254, "step": 7310 }, { "epoch": 0.14, "learning_rate": 4.763503677969063e-05, "loss": 0.0362, "step": 7312 }, { "epoch": 0.14, "learning_rate": 4.76343898194334e-05, "loss": 0.0233, "step": 7314 }, { "epoch": 0.14, "learning_rate": 4.763374285917616e-05, "loss": 0.0141, "step": 7316 }, { "epoch": 0.14, "learning_rate": 4.7633095898918936e-05, "loss": 0.0241, "step": 7318 }, { "epoch": 0.14, "learning_rate": 4.76324489386617e-05, "loss": 0.0189, "step": 7320 }, { "epoch": 0.14, "learning_rate": 4.763180197840447e-05, "loss": 0.0048, "step": 7322 }, { "epoch": 0.14, "learning_rate": 4.7631155018147236e-05, "loss": 0.0152, "step": 7324 }, { "epoch": 0.14, "learning_rate": 4.7630508057890005e-05, "loss": 0.0058, "step": 7326 }, { "epoch": 0.14, "learning_rate": 4.7629861097632774e-05, "loss": 0.0112, "step": 7328 }, { "epoch": 0.14, "learning_rate": 4.7629214137375543e-05, "loss": 0.0188, "step": 7330 }, { "epoch": 0.14, "learning_rate": 4.762856717711831e-05, "loss": 0.0196, "step": 7332 }, { "epoch": 0.14, "learning_rate": 4.762792021686108e-05, "loss": 0.0035, "step": 7334 }, { "epoch": 0.14, "learning_rate": 4.762727325660385e-05, "loss": 0.0298, "step": 7336 }, { "epoch": 0.14, "learning_rate": 4.762662629634661e-05, "loss": 0.0216, "step": 7338 }, { "epoch": 0.14, "learning_rate": 4.762597933608939e-05, "loss": 0.021, "step": 7340 }, { "epoch": 0.14, "learning_rate": 4.762533237583215e-05, "loss": 0.0045, "step": 7342 }, { "epoch": 0.14, "learning_rate": 4.762468541557492e-05, "loss": 0.0367, "step": 7344 }, { "epoch": 0.14, "learning_rate": 4.7624038455317696e-05, "loss": 0.0312, "step": 7346 }, { "epoch": 0.14, "learning_rate": 4.762339149506046e-05, "loss": 0.0172, "step": 7348 }, { "epoch": 0.14, "learning_rate": 4.7622744534803234e-05, "loss": 0.0058, "step": 7350 }, { "epoch": 0.14, "learning_rate": 4.7622097574546e-05, "loss": 0.0062, "step": 7352 }, { "epoch": 0.14, "learning_rate": 4.7621450614288766e-05, "loss": 0.0361, "step": 7354 }, { "epoch": 0.14, "learning_rate": 4.7620803654031535e-05, "loss": 0.0119, "step": 7356 }, { "epoch": 0.14, "learning_rate": 4.7620156693774304e-05, "loss": 0.0505, "step": 7358 }, { "epoch": 0.14, "learning_rate": 4.761950973351707e-05, "loss": 0.0059, "step": 7360 }, { "epoch": 0.14, "learning_rate": 4.761886277325984e-05, "loss": 0.0066, "step": 7362 }, { "epoch": 0.14, "learning_rate": 4.761821581300261e-05, "loss": 0.0189, "step": 7364 }, { "epoch": 0.14, "learning_rate": 4.7617568852745373e-05, "loss": 0.0248, "step": 7366 }, { "epoch": 0.14, "learning_rate": 4.761692189248815e-05, "loss": 0.0136, "step": 7368 }, { "epoch": 0.14, "learning_rate": 4.761627493223091e-05, "loss": 0.0229, "step": 7370 }, { "epoch": 0.14, "learning_rate": 4.761562797197369e-05, "loss": 0.0176, "step": 7372 }, { "epoch": 0.14, "learning_rate": 4.761498101171645e-05, "loss": 0.0199, "step": 7374 }, { "epoch": 0.14, "learning_rate": 4.761433405145922e-05, "loss": 0.034, "step": 7376 }, { "epoch": 0.14, "learning_rate": 4.7613687091201995e-05, "loss": 0.0072, "step": 7378 }, { "epoch": 0.14, "learning_rate": 4.761304013094476e-05, "loss": 0.0256, "step": 7380 }, { "epoch": 0.14, "learning_rate": 4.7612393170687526e-05, "loss": 0.0271, "step": 7382 }, { "epoch": 0.14, "learning_rate": 4.7611746210430295e-05, "loss": 0.0123, "step": 7384 }, { "epoch": 0.14, "learning_rate": 4.7611099250173064e-05, "loss": 0.0096, "step": 7386 }, { "epoch": 0.14, "learning_rate": 4.761045228991583e-05, "loss": 0.0266, "step": 7388 }, { "epoch": 0.14, "learning_rate": 4.76098053296586e-05, "loss": 0.0093, "step": 7390 }, { "epoch": 0.14, "learning_rate": 4.760915836940137e-05, "loss": 0.0058, "step": 7392 }, { "epoch": 0.14, "learning_rate": 4.760851140914414e-05, "loss": 0.0431, "step": 7394 }, { "epoch": 0.14, "learning_rate": 4.760786444888691e-05, "loss": 0.025, "step": 7396 }, { "epoch": 0.14, "learning_rate": 4.760721748862967e-05, "loss": 0.003, "step": 7398 }, { "epoch": 0.14, "learning_rate": 4.760657052837245e-05, "loss": 0.0271, "step": 7400 }, { "epoch": 0.14, "learning_rate": 4.760592356811521e-05, "loss": 0.0149, "step": 7402 }, { "epoch": 0.14, "learning_rate": 4.760527660785798e-05, "loss": 0.0098, "step": 7404 }, { "epoch": 0.14, "learning_rate": 4.760462964760075e-05, "loss": 0.0462, "step": 7406 }, { "epoch": 0.14, "learning_rate": 4.760398268734352e-05, "loss": 0.0076, "step": 7408 }, { "epoch": 0.14, "learning_rate": 4.760333572708629e-05, "loss": 0.0189, "step": 7410 }, { "epoch": 0.14, "learning_rate": 4.7602688766829056e-05, "loss": 0.0276, "step": 7412 }, { "epoch": 0.14, "learning_rate": 4.7602041806571825e-05, "loss": 0.0286, "step": 7414 }, { "epoch": 0.14, "learning_rate": 4.7601394846314594e-05, "loss": 0.0456, "step": 7416 }, { "epoch": 0.14, "learning_rate": 4.760074788605736e-05, "loss": 0.015, "step": 7418 }, { "epoch": 0.14, "learning_rate": 4.7600100925800125e-05, "loss": 0.0113, "step": 7420 }, { "epoch": 0.14, "learning_rate": 4.75994539655429e-05, "loss": 0.0122, "step": 7422 }, { "epoch": 0.14, "learning_rate": 4.759880700528567e-05, "loss": 0.0167, "step": 7424 }, { "epoch": 0.14, "learning_rate": 4.759816004502843e-05, "loss": 0.0428, "step": 7426 }, { "epoch": 0.14, "learning_rate": 4.759751308477121e-05, "loss": 0.013, "step": 7428 }, { "epoch": 0.14, "learning_rate": 4.759686612451397e-05, "loss": 0.0386, "step": 7430 }, { "epoch": 0.14, "learning_rate": 4.7596219164256746e-05, "loss": 0.0126, "step": 7432 }, { "epoch": 0.14, "learning_rate": 4.759557220399951e-05, "loss": 0.0134, "step": 7434 }, { "epoch": 0.14, "learning_rate": 4.759492524374228e-05, "loss": 0.0197, "step": 7436 }, { "epoch": 0.14, "learning_rate": 4.759427828348505e-05, "loss": 0.0084, "step": 7438 }, { "epoch": 0.14, "learning_rate": 4.7593631323227816e-05, "loss": 0.012, "step": 7440 }, { "epoch": 0.14, "learning_rate": 4.7592984362970585e-05, "loss": 0.0242, "step": 7442 }, { "epoch": 0.14, "learning_rate": 4.7592337402713354e-05, "loss": 0.0445, "step": 7444 }, { "epoch": 0.14, "learning_rate": 4.759169044245612e-05, "loss": 0.0216, "step": 7446 }, { "epoch": 0.14, "learning_rate": 4.7591043482198886e-05, "loss": 0.0229, "step": 7448 }, { "epoch": 0.14, "learning_rate": 4.759039652194166e-05, "loss": 0.018, "step": 7450 }, { "epoch": 0.14, "learning_rate": 4.7589749561684424e-05, "loss": 0.0202, "step": 7452 }, { "epoch": 0.14, "learning_rate": 4.75891026014272e-05, "loss": 0.0139, "step": 7454 }, { "epoch": 0.14, "learning_rate": 4.758845564116997e-05, "loss": 0.0115, "step": 7456 }, { "epoch": 0.14, "learning_rate": 4.758780868091273e-05, "loss": 0.0129, "step": 7458 }, { "epoch": 0.14, "learning_rate": 4.758716172065551e-05, "loss": 0.0428, "step": 7460 }, { "epoch": 0.14, "learning_rate": 4.758651476039827e-05, "loss": 0.0141, "step": 7462 }, { "epoch": 0.14, "learning_rate": 4.758586780014104e-05, "loss": 0.009, "step": 7464 }, { "epoch": 0.14, "learning_rate": 4.758522083988381e-05, "loss": 0.0149, "step": 7466 }, { "epoch": 0.14, "learning_rate": 4.7584573879626576e-05, "loss": 0.0346, "step": 7468 }, { "epoch": 0.14, "learning_rate": 4.7583926919369345e-05, "loss": 0.0104, "step": 7470 }, { "epoch": 0.15, "learning_rate": 4.7583279959112115e-05, "loss": 0.0123, "step": 7472 }, { "epoch": 0.15, "learning_rate": 4.7582632998854884e-05, "loss": 0.028, "step": 7474 }, { "epoch": 0.15, "learning_rate": 4.758198603859765e-05, "loss": 0.0177, "step": 7476 }, { "epoch": 0.15, "learning_rate": 4.758133907834042e-05, "loss": 0.0228, "step": 7478 }, { "epoch": 0.15, "learning_rate": 4.7580692118083184e-05, "loss": 0.0191, "step": 7480 }, { "epoch": 0.15, "learning_rate": 4.758004515782596e-05, "loss": 0.0178, "step": 7482 }, { "epoch": 0.15, "learning_rate": 4.757939819756872e-05, "loss": 0.0355, "step": 7484 }, { "epoch": 0.15, "learning_rate": 4.757875123731149e-05, "loss": 0.0101, "step": 7486 }, { "epoch": 0.15, "learning_rate": 4.757810427705427e-05, "loss": 0.0326, "step": 7488 }, { "epoch": 0.15, "learning_rate": 4.757745731679703e-05, "loss": 0.0165, "step": 7490 }, { "epoch": 0.15, "learning_rate": 4.7576810356539805e-05, "loss": 0.0168, "step": 7492 }, { "epoch": 0.15, "learning_rate": 4.757616339628257e-05, "loss": 0.0124, "step": 7494 }, { "epoch": 0.15, "learning_rate": 4.757551643602534e-05, "loss": 0.0141, "step": 7496 }, { "epoch": 0.15, "learning_rate": 4.7574869475768106e-05, "loss": 0.0205, "step": 7498 }, { "epoch": 0.15, "learning_rate": 4.7574222515510875e-05, "loss": 0.0237, "step": 7500 }, { "epoch": 0.15, "learning_rate": 4.7573575555253644e-05, "loss": 0.0238, "step": 7502 }, { "epoch": 0.15, "learning_rate": 4.757292859499641e-05, "loss": 0.0148, "step": 7504 }, { "epoch": 0.15, "learning_rate": 4.757228163473918e-05, "loss": 0.0181, "step": 7506 }, { "epoch": 0.15, "learning_rate": 4.7571634674481945e-05, "loss": 0.0263, "step": 7508 }, { "epoch": 0.15, "learning_rate": 4.757098771422472e-05, "loss": 0.0135, "step": 7510 }, { "epoch": 0.15, "learning_rate": 4.757034075396748e-05, "loss": 0.0215, "step": 7512 }, { "epoch": 0.15, "learning_rate": 4.756969379371026e-05, "loss": 0.0127, "step": 7514 }, { "epoch": 0.15, "learning_rate": 4.756904683345302e-05, "loss": 0.0255, "step": 7516 }, { "epoch": 0.15, "learning_rate": 4.756839987319579e-05, "loss": 0.0123, "step": 7518 }, { "epoch": 0.15, "learning_rate": 4.756775291293856e-05, "loss": 0.0272, "step": 7520 }, { "epoch": 0.15, "learning_rate": 4.756710595268133e-05, "loss": 0.0141, "step": 7522 }, { "epoch": 0.15, "learning_rate": 4.75664589924241e-05, "loss": 0.0267, "step": 7524 }, { "epoch": 0.15, "learning_rate": 4.7565812032166866e-05, "loss": 0.0282, "step": 7526 }, { "epoch": 0.15, "learning_rate": 4.7565165071909635e-05, "loss": 0.0278, "step": 7528 }, { "epoch": 0.15, "learning_rate": 4.75645181116524e-05, "loss": 0.0209, "step": 7530 }, { "epoch": 0.15, "learning_rate": 4.7563871151395174e-05, "loss": 0.025, "step": 7532 }, { "epoch": 0.15, "learning_rate": 4.756322419113794e-05, "loss": 0.0293, "step": 7534 }, { "epoch": 0.15, "learning_rate": 4.756257723088071e-05, "loss": 0.0095, "step": 7536 }, { "epoch": 0.15, "learning_rate": 4.756193027062348e-05, "loss": 0.0168, "step": 7538 }, { "epoch": 0.15, "learning_rate": 4.756128331036624e-05, "loss": 0.0153, "step": 7540 }, { "epoch": 0.15, "learning_rate": 4.756063635010902e-05, "loss": 0.0236, "step": 7542 }, { "epoch": 0.15, "learning_rate": 4.755998938985178e-05, "loss": 0.0212, "step": 7544 }, { "epoch": 0.15, "learning_rate": 4.755934242959455e-05, "loss": 0.0173, "step": 7546 }, { "epoch": 0.15, "learning_rate": 4.755869546933732e-05, "loss": 0.0213, "step": 7548 }, { "epoch": 0.15, "learning_rate": 4.755804850908009e-05, "loss": 0.0232, "step": 7550 }, { "epoch": 0.15, "learning_rate": 4.755740154882286e-05, "loss": 0.009, "step": 7552 }, { "epoch": 0.15, "learning_rate": 4.755675458856563e-05, "loss": 0.0174, "step": 7554 }, { "epoch": 0.15, "learning_rate": 4.7556107628308396e-05, "loss": 0.0324, "step": 7556 }, { "epoch": 0.15, "learning_rate": 4.7555460668051165e-05, "loss": 0.0091, "step": 7558 }, { "epoch": 0.15, "learning_rate": 4.755513718792255e-05, "loss": 0.2548, "step": 7560 }, { "epoch": 0.15, "learning_rate": 4.755449022766532e-05, "loss": 0.0159, "step": 7562 }, { "epoch": 0.15, "learning_rate": 4.755384326740809e-05, "loss": 0.0326, "step": 7564 }, { "epoch": 0.15, "learning_rate": 4.755319630715086e-05, "loss": 0.0094, "step": 7566 }, { "epoch": 0.15, "learning_rate": 4.755254934689362e-05, "loss": 0.0368, "step": 7568 }, { "epoch": 0.15, "learning_rate": 4.7551902386636395e-05, "loss": 0.0263, "step": 7570 }, { "epoch": 0.15, "learning_rate": 4.755125542637916e-05, "loss": 0.0196, "step": 7572 }, { "epoch": 0.15, "learning_rate": 4.7550608466121926e-05, "loss": 0.0117, "step": 7574 }, { "epoch": 0.15, "learning_rate": 4.7549961505864695e-05, "loss": 0.0259, "step": 7576 }, { "epoch": 0.15, "learning_rate": 4.7549314545607464e-05, "loss": 0.0417, "step": 7578 }, { "epoch": 0.15, "learning_rate": 4.7548667585350233e-05, "loss": 0.0217, "step": 7580 }, { "epoch": 0.15, "learning_rate": 4.7548020625093e-05, "loss": 0.0112, "step": 7582 }, { "epoch": 0.15, "learning_rate": 4.754737366483577e-05, "loss": 0.026, "step": 7584 }, { "epoch": 0.15, "learning_rate": 4.754672670457854e-05, "loss": 0.0136, "step": 7586 }, { "epoch": 0.15, "learning_rate": 4.754607974432131e-05, "loss": 0.0064, "step": 7588 }, { "epoch": 0.15, "learning_rate": 4.754543278406408e-05, "loss": 0.0204, "step": 7590 }, { "epoch": 0.15, "learning_rate": 4.754478582380685e-05, "loss": 0.0381, "step": 7592 }, { "epoch": 0.15, "learning_rate": 4.754413886354962e-05, "loss": 0.0131, "step": 7594 }, { "epoch": 0.15, "learning_rate": 4.754349190329238e-05, "loss": 0.0174, "step": 7596 }, { "epoch": 0.15, "learning_rate": 4.7542844943035155e-05, "loss": 0.0107, "step": 7598 }, { "epoch": 0.15, "learning_rate": 4.754219798277792e-05, "loss": 0.011, "step": 7600 }, { "epoch": 0.15, "learning_rate": 4.754155102252069e-05, "loss": 0.0388, "step": 7602 }, { "epoch": 0.15, "learning_rate": 4.7540904062263456e-05, "loss": 0.0138, "step": 7604 }, { "epoch": 0.15, "learning_rate": 4.7540257102006225e-05, "loss": 0.0155, "step": 7606 }, { "epoch": 0.15, "learning_rate": 4.7539610141748994e-05, "loss": 0.0187, "step": 7608 }, { "epoch": 0.15, "learning_rate": 4.753896318149176e-05, "loss": 0.0245, "step": 7610 }, { "epoch": 0.15, "learning_rate": 4.753831622123453e-05, "loss": 0.0178, "step": 7612 }, { "epoch": 0.15, "learning_rate": 4.75376692609773e-05, "loss": 0.0224, "step": 7614 }, { "epoch": 0.15, "learning_rate": 4.753702230072007e-05, "loss": 0.0396, "step": 7616 }, { "epoch": 0.15, "learning_rate": 4.753637534046283e-05, "loss": 0.0166, "step": 7618 }, { "epoch": 0.15, "learning_rate": 4.753572838020561e-05, "loss": 0.0212, "step": 7620 }, { "epoch": 0.15, "learning_rate": 4.753508141994837e-05, "loss": 0.0086, "step": 7622 }, { "epoch": 0.15, "learning_rate": 4.7534434459691147e-05, "loss": 0.0205, "step": 7624 }, { "epoch": 0.15, "learning_rate": 4.7533787499433916e-05, "loss": 0.0659, "step": 7626 }, { "epoch": 0.15, "learning_rate": 4.753314053917668e-05, "loss": 0.0104, "step": 7628 }, { "epoch": 0.15, "learning_rate": 4.7532493578919454e-05, "loss": 0.0146, "step": 7630 }, { "epoch": 0.15, "learning_rate": 4.7531846618662216e-05, "loss": 0.0123, "step": 7632 }, { "epoch": 0.15, "learning_rate": 4.7531199658404985e-05, "loss": 0.0137, "step": 7634 }, { "epoch": 0.15, "learning_rate": 4.7530552698147754e-05, "loss": 0.0113, "step": 7636 }, { "epoch": 0.15, "learning_rate": 4.752990573789052e-05, "loss": 0.0076, "step": 7638 }, { "epoch": 0.15, "learning_rate": 4.752925877763329e-05, "loss": 0.0189, "step": 7640 }, { "epoch": 0.15, "learning_rate": 4.752861181737606e-05, "loss": 0.0083, "step": 7642 }, { "epoch": 0.15, "learning_rate": 4.752796485711883e-05, "loss": 0.0312, "step": 7644 }, { "epoch": 0.15, "learning_rate": 4.75273178968616e-05, "loss": 0.0229, "step": 7646 }, { "epoch": 0.15, "learning_rate": 4.752667093660437e-05, "loss": 0.0133, "step": 7648 }, { "epoch": 0.15, "learning_rate": 4.752602397634713e-05, "loss": 0.0266, "step": 7650 }, { "epoch": 0.15, "learning_rate": 4.752537701608991e-05, "loss": 0.0295, "step": 7652 }, { "epoch": 0.15, "learning_rate": 4.752473005583267e-05, "loss": 0.017, "step": 7654 }, { "epoch": 0.15, "learning_rate": 4.752408309557544e-05, "loss": 0.0191, "step": 7656 }, { "epoch": 0.15, "learning_rate": 4.7523436135318214e-05, "loss": 0.0187, "step": 7658 }, { "epoch": 0.15, "learning_rate": 4.7522789175060977e-05, "loss": 0.0266, "step": 7660 }, { "epoch": 0.15, "learning_rate": 4.7522142214803746e-05, "loss": 0.0105, "step": 7662 }, { "epoch": 0.15, "learning_rate": 4.7521495254546515e-05, "loss": 0.0102, "step": 7664 }, { "epoch": 0.15, "learning_rate": 4.7520848294289284e-05, "loss": 0.0113, "step": 7666 }, { "epoch": 0.15, "learning_rate": 4.752020133403205e-05, "loss": 0.0393, "step": 7668 }, { "epoch": 0.15, "learning_rate": 4.751955437377482e-05, "loss": 0.0567, "step": 7670 }, { "epoch": 0.15, "learning_rate": 4.751890741351759e-05, "loss": 0.0116, "step": 7672 }, { "epoch": 0.15, "learning_rate": 4.751826045326036e-05, "loss": 0.0119, "step": 7674 }, { "epoch": 0.15, "learning_rate": 4.751761349300313e-05, "loss": 0.0284, "step": 7676 }, { "epoch": 0.15, "learning_rate": 4.751696653274589e-05, "loss": 0.0205, "step": 7678 }, { "epoch": 0.15, "learning_rate": 4.751631957248867e-05, "loss": 0.011, "step": 7680 }, { "epoch": 0.15, "learning_rate": 4.751567261223143e-05, "loss": 0.0372, "step": 7682 }, { "epoch": 0.15, "learning_rate": 4.75150256519742e-05, "loss": 0.0179, "step": 7684 }, { "epoch": 0.15, "learning_rate": 4.751437869171697e-05, "loss": 0.035, "step": 7686 }, { "epoch": 0.15, "learning_rate": 4.751373173145974e-05, "loss": 0.0069, "step": 7688 }, { "epoch": 0.15, "learning_rate": 4.751308477120251e-05, "loss": 0.0104, "step": 7690 }, { "epoch": 0.15, "learning_rate": 4.7512437810945275e-05, "loss": 0.0052, "step": 7692 }, { "epoch": 0.15, "learning_rate": 4.7511790850688044e-05, "loss": 0.0534, "step": 7694 }, { "epoch": 0.15, "learning_rate": 4.751114389043081e-05, "loss": 0.0062, "step": 7696 }, { "epoch": 0.15, "learning_rate": 4.751049693017358e-05, "loss": 0.0077, "step": 7698 }, { "epoch": 0.15, "learning_rate": 4.7509849969916345e-05, "loss": 0.0212, "step": 7700 }, { "epoch": 0.15, "learning_rate": 4.750920300965912e-05, "loss": 0.0229, "step": 7702 }, { "epoch": 0.15, "learning_rate": 4.750855604940189e-05, "loss": 0.0178, "step": 7704 }, { "epoch": 0.15, "learning_rate": 4.750790908914466e-05, "loss": 0.0201, "step": 7706 }, { "epoch": 0.15, "learning_rate": 4.750726212888743e-05, "loss": 0.0365, "step": 7708 }, { "epoch": 0.15, "learning_rate": 4.750661516863019e-05, "loss": 0.0053, "step": 7710 }, { "epoch": 0.15, "learning_rate": 4.7505968208372966e-05, "loss": 0.013, "step": 7712 }, { "epoch": 0.15, "learning_rate": 4.750532124811573e-05, "loss": 0.0173, "step": 7714 }, { "epoch": 0.15, "learning_rate": 4.75046742878585e-05, "loss": 0.0051, "step": 7716 }, { "epoch": 0.15, "learning_rate": 4.7504027327601266e-05, "loss": 0.0056, "step": 7718 }, { "epoch": 0.15, "learning_rate": 4.7503380367344036e-05, "loss": 0.0114, "step": 7720 }, { "epoch": 0.15, "learning_rate": 4.7502733407086805e-05, "loss": 0.0247, "step": 7722 }, { "epoch": 0.15, "learning_rate": 4.7502086446829574e-05, "loss": 0.0161, "step": 7724 }, { "epoch": 0.15, "learning_rate": 4.750143948657234e-05, "loss": 0.0116, "step": 7726 }, { "epoch": 0.15, "learning_rate": 4.750079252631511e-05, "loss": 0.0119, "step": 7728 }, { "epoch": 0.15, "learning_rate": 4.750014556605788e-05, "loss": 0.0195, "step": 7730 }, { "epoch": 0.15, "learning_rate": 4.749949860580064e-05, "loss": 0.0192, "step": 7732 }, { "epoch": 0.15, "learning_rate": 4.749885164554342e-05, "loss": 0.0299, "step": 7734 }, { "epoch": 0.15, "learning_rate": 4.749820468528619e-05, "loss": 0.0106, "step": 7736 }, { "epoch": 0.15, "learning_rate": 4.749755772502895e-05, "loss": 0.0131, "step": 7738 }, { "epoch": 0.15, "learning_rate": 4.7496910764771726e-05, "loss": 0.0108, "step": 7740 }, { "epoch": 0.15, "learning_rate": 4.749626380451449e-05, "loss": 0.0143, "step": 7742 }, { "epoch": 0.15, "learning_rate": 4.749561684425726e-05, "loss": 0.0145, "step": 7744 }, { "epoch": 0.15, "learning_rate": 4.749496988400003e-05, "loss": 0.0057, "step": 7746 }, { "epoch": 0.15, "learning_rate": 4.7494322923742796e-05, "loss": 0.0177, "step": 7748 }, { "epoch": 0.15, "learning_rate": 4.7493675963485565e-05, "loss": 0.0422, "step": 7750 }, { "epoch": 0.15, "learning_rate": 4.7493029003228334e-05, "loss": 0.1787, "step": 7752 }, { "epoch": 0.15, "learning_rate": 4.74923820429711e-05, "loss": 0.0226, "step": 7754 }, { "epoch": 0.15, "learning_rate": 4.749173508271387e-05, "loss": 0.0224, "step": 7756 }, { "epoch": 0.15, "learning_rate": 4.749108812245664e-05, "loss": 0.0151, "step": 7758 }, { "epoch": 0.15, "learning_rate": 4.7490441162199404e-05, "loss": 0.0187, "step": 7760 }, { "epoch": 0.15, "learning_rate": 4.748979420194218e-05, "loss": 0.0422, "step": 7762 }, { "epoch": 0.15, "learning_rate": 4.748914724168494e-05, "loss": 0.0167, "step": 7764 }, { "epoch": 0.15, "learning_rate": 4.748850028142772e-05, "loss": 0.0067, "step": 7766 }, { "epoch": 0.15, "learning_rate": 4.748785332117049e-05, "loss": 0.0108, "step": 7768 }, { "epoch": 0.15, "learning_rate": 4.748720636091325e-05, "loss": 0.0152, "step": 7770 }, { "epoch": 0.15, "learning_rate": 4.7486559400656025e-05, "loss": 0.0115, "step": 7772 }, { "epoch": 0.15, "learning_rate": 4.748591244039879e-05, "loss": 0.0148, "step": 7774 }, { "epoch": 0.15, "learning_rate": 4.7485265480141556e-05, "loss": 0.0176, "step": 7776 }, { "epoch": 0.15, "learning_rate": 4.7484618519884325e-05, "loss": 0.0182, "step": 7778 }, { "epoch": 0.15, "learning_rate": 4.7483971559627094e-05, "loss": 0.0033, "step": 7780 }, { "epoch": 0.15, "learning_rate": 4.7483324599369864e-05, "loss": 0.0223, "step": 7782 }, { "epoch": 0.15, "learning_rate": 4.748267763911263e-05, "loss": 0.0092, "step": 7784 }, { "epoch": 0.15, "learning_rate": 4.74820306788554e-05, "loss": 0.0086, "step": 7786 }, { "epoch": 0.15, "learning_rate": 4.748138371859817e-05, "loss": 0.0037, "step": 7788 }, { "epoch": 0.15, "learning_rate": 4.748073675834094e-05, "loss": 0.0185, "step": 7790 }, { "epoch": 0.15, "learning_rate": 4.74800897980837e-05, "loss": 0.0129, "step": 7792 }, { "epoch": 0.15, "learning_rate": 4.747944283782648e-05, "loss": 0.0129, "step": 7794 }, { "epoch": 0.15, "learning_rate": 4.747879587756924e-05, "loss": 0.0241, "step": 7796 }, { "epoch": 0.15, "learning_rate": 4.747814891731201e-05, "loss": 0.0099, "step": 7798 }, { "epoch": 0.15, "learning_rate": 4.747750195705478e-05, "loss": 0.0154, "step": 7800 }, { "epoch": 0.15, "learning_rate": 4.747685499679755e-05, "loss": 0.0153, "step": 7802 }, { "epoch": 0.15, "learning_rate": 4.747620803654032e-05, "loss": 0.0119, "step": 7804 }, { "epoch": 0.15, "learning_rate": 4.7475561076283086e-05, "loss": 0.0244, "step": 7806 }, { "epoch": 0.15, "learning_rate": 4.7474914116025855e-05, "loss": 0.0119, "step": 7808 }, { "epoch": 0.15, "learning_rate": 4.7474267155768624e-05, "loss": 0.0196, "step": 7810 }, { "epoch": 0.15, "learning_rate": 4.747362019551139e-05, "loss": 0.0073, "step": 7812 }, { "epoch": 0.15, "learning_rate": 4.747297323525416e-05, "loss": 0.0388, "step": 7814 }, { "epoch": 0.15, "learning_rate": 4.747232627499693e-05, "loss": 0.0144, "step": 7816 }, { "epoch": 0.15, "learning_rate": 4.74716793147397e-05, "loss": 0.0088, "step": 7818 }, { "epoch": 0.15, "learning_rate": 4.747103235448246e-05, "loss": 0.018, "step": 7820 }, { "epoch": 0.15, "learning_rate": 4.747038539422524e-05, "loss": 0.0252, "step": 7822 }, { "epoch": 0.15, "learning_rate": 4.7469738433968e-05, "loss": 0.0167, "step": 7824 }, { "epoch": 0.15, "learning_rate": 4.746909147371077e-05, "loss": 0.0261, "step": 7826 }, { "epoch": 0.15, "learning_rate": 4.746844451345354e-05, "loss": 0.0053, "step": 7828 }, { "epoch": 0.15, "learning_rate": 4.746779755319631e-05, "loss": 0.0226, "step": 7830 }, { "epoch": 0.15, "learning_rate": 4.746715059293908e-05, "loss": 0.0275, "step": 7832 }, { "epoch": 0.15, "learning_rate": 4.7466503632681846e-05, "loss": 0.01, "step": 7834 }, { "epoch": 0.15, "learning_rate": 4.7465856672424615e-05, "loss": 0.0166, "step": 7836 }, { "epoch": 0.15, "learning_rate": 4.7465209712167384e-05, "loss": 0.011, "step": 7838 }, { "epoch": 0.15, "learning_rate": 4.7464562751910153e-05, "loss": 0.009, "step": 7840 }, { "epoch": 0.15, "learning_rate": 4.7463915791652916e-05, "loss": 0.0123, "step": 7842 }, { "epoch": 0.15, "learning_rate": 4.746326883139569e-05, "loss": 0.0111, "step": 7844 }, { "epoch": 0.15, "learning_rate": 4.746262187113846e-05, "loss": 0.0078, "step": 7846 }, { "epoch": 0.15, "learning_rate": 4.746197491088123e-05, "loss": 0.0155, "step": 7848 }, { "epoch": 0.15, "learning_rate": 4.7461327950624e-05, "loss": 0.0316, "step": 7850 }, { "epoch": 0.15, "learning_rate": 4.746068099036676e-05, "loss": 0.0036, "step": 7852 }, { "epoch": 0.15, "learning_rate": 4.746003403010954e-05, "loss": 0.0069, "step": 7854 }, { "epoch": 0.15, "learning_rate": 4.74593870698523e-05, "loss": 0.0147, "step": 7856 }, { "epoch": 0.15, "learning_rate": 4.745874010959507e-05, "loss": 0.0084, "step": 7858 }, { "epoch": 0.15, "learning_rate": 4.745809314933784e-05, "loss": 0.0105, "step": 7860 }, { "epoch": 0.15, "learning_rate": 4.745744618908061e-05, "loss": 0.0239, "step": 7862 }, { "epoch": 0.15, "learning_rate": 4.7456799228823376e-05, "loss": 0.0049, "step": 7864 }, { "epoch": 0.15, "learning_rate": 4.7456152268566145e-05, "loss": 0.0136, "step": 7866 }, { "epoch": 0.15, "learning_rate": 4.7455505308308914e-05, "loss": 0.0124, "step": 7868 }, { "epoch": 0.15, "learning_rate": 4.745485834805168e-05, "loss": 0.021, "step": 7870 }, { "epoch": 0.15, "learning_rate": 4.745421138779445e-05, "loss": 0.003, "step": 7872 }, { "epoch": 0.15, "learning_rate": 4.7453564427537214e-05, "loss": 0.0062, "step": 7874 }, { "epoch": 0.15, "learning_rate": 4.745291746727999e-05, "loss": 0.0116, "step": 7876 }, { "epoch": 0.15, "learning_rate": 4.745227050702275e-05, "loss": 0.0152, "step": 7878 }, { "epoch": 0.15, "learning_rate": 4.745162354676552e-05, "loss": 0.0059, "step": 7880 }, { "epoch": 0.15, "learning_rate": 4.74509765865083e-05, "loss": 0.0069, "step": 7882 }, { "epoch": 0.15, "learning_rate": 4.745032962625106e-05, "loss": 0.0493, "step": 7884 }, { "epoch": 0.15, "learning_rate": 4.744968266599383e-05, "loss": 0.008, "step": 7886 }, { "epoch": 0.15, "learning_rate": 4.74490357057366e-05, "loss": 0.0184, "step": 7888 }, { "epoch": 0.15, "learning_rate": 4.744838874547937e-05, "loss": 0.0509, "step": 7890 }, { "epoch": 0.15, "learning_rate": 4.7447741785222136e-05, "loss": 0.0141, "step": 7892 }, { "epoch": 0.15, "learning_rate": 4.7447094824964905e-05, "loss": 0.0095, "step": 7894 }, { "epoch": 0.15, "learning_rate": 4.7446447864707674e-05, "loss": 0.0094, "step": 7896 }, { "epoch": 0.15, "learning_rate": 4.744580090445044e-05, "loss": 0.0064, "step": 7898 }, { "epoch": 0.15, "learning_rate": 4.744515394419321e-05, "loss": 0.022, "step": 7900 }, { "epoch": 0.15, "learning_rate": 4.7444506983935975e-05, "loss": 0.0233, "step": 7902 }, { "epoch": 0.15, "learning_rate": 4.744386002367875e-05, "loss": 0.0384, "step": 7904 }, { "epoch": 0.15, "learning_rate": 4.744321306342151e-05, "loss": 0.003, "step": 7906 }, { "epoch": 0.15, "learning_rate": 4.744256610316429e-05, "loss": 0.0345, "step": 7908 }, { "epoch": 0.15, "learning_rate": 4.744191914290705e-05, "loss": 0.0292, "step": 7910 }, { "epoch": 0.15, "learning_rate": 4.744127218264982e-05, "loss": 0.0673, "step": 7912 }, { "epoch": 0.15, "learning_rate": 4.7440625222392596e-05, "loss": 0.01, "step": 7914 }, { "epoch": 0.15, "learning_rate": 4.743997826213536e-05, "loss": 0.0115, "step": 7916 }, { "epoch": 0.15, "learning_rate": 4.743933130187813e-05, "loss": 0.0128, "step": 7918 }, { "epoch": 0.15, "learning_rate": 4.7438684341620897e-05, "loss": 0.0208, "step": 7920 }, { "epoch": 0.15, "learning_rate": 4.7438037381363666e-05, "loss": 0.0148, "step": 7922 }, { "epoch": 0.15, "learning_rate": 4.743739042110643e-05, "loss": 0.0114, "step": 7924 }, { "epoch": 0.15, "learning_rate": 4.7436743460849204e-05, "loss": 0.0249, "step": 7926 }, { "epoch": 0.15, "learning_rate": 4.743609650059197e-05, "loss": 0.0317, "step": 7928 }, { "epoch": 0.15, "learning_rate": 4.743544954033474e-05, "loss": 0.0163, "step": 7930 }, { "epoch": 0.15, "learning_rate": 4.743480258007751e-05, "loss": 0.0194, "step": 7932 }, { "epoch": 0.15, "learning_rate": 4.743415561982027e-05, "loss": 0.0054, "step": 7934 }, { "epoch": 0.15, "learning_rate": 4.743350865956305e-05, "loss": 0.0183, "step": 7936 }, { "epoch": 0.15, "learning_rate": 4.743286169930581e-05, "loss": 0.0172, "step": 7938 }, { "epoch": 0.15, "learning_rate": 4.743221473904858e-05, "loss": 0.0077, "step": 7940 }, { "epoch": 0.15, "learning_rate": 4.743156777879135e-05, "loss": 0.0145, "step": 7942 }, { "epoch": 0.15, "learning_rate": 4.743092081853412e-05, "loss": 0.0053, "step": 7944 }, { "epoch": 0.15, "learning_rate": 4.743027385827689e-05, "loss": 0.0209, "step": 7946 }, { "epoch": 0.15, "learning_rate": 4.742962689801966e-05, "loss": 0.0056, "step": 7948 }, { "epoch": 0.15, "learning_rate": 4.7428979937762426e-05, "loss": 0.0044, "step": 7950 }, { "epoch": 0.15, "learning_rate": 4.7428332977505195e-05, "loss": 0.0088, "step": 7952 }, { "epoch": 0.15, "learning_rate": 4.7427686017247964e-05, "loss": 0.0202, "step": 7954 }, { "epoch": 0.15, "learning_rate": 4.7427039056990726e-05, "loss": 0.0193, "step": 7956 }, { "epoch": 0.15, "learning_rate": 4.74263920967335e-05, "loss": 0.0146, "step": 7958 }, { "epoch": 0.15, "learning_rate": 4.742574513647627e-05, "loss": 0.0116, "step": 7960 }, { "epoch": 0.15, "learning_rate": 4.7425098176219034e-05, "loss": 0.0043, "step": 7962 }, { "epoch": 0.15, "learning_rate": 4.742445121596181e-05, "loss": 0.0154, "step": 7964 }, { "epoch": 0.15, "learning_rate": 4.742380425570457e-05, "loss": 0.0103, "step": 7966 }, { "epoch": 0.15, "learning_rate": 4.742315729544734e-05, "loss": 0.0065, "step": 7968 }, { "epoch": 0.15, "learning_rate": 4.742251033519011e-05, "loss": 0.007, "step": 7970 }, { "epoch": 0.15, "learning_rate": 4.742186337493288e-05, "loss": 0.0038, "step": 7972 }, { "epoch": 0.15, "learning_rate": 4.742121641467565e-05, "loss": 0.0344, "step": 7974 }, { "epoch": 0.15, "learning_rate": 4.742056945441842e-05, "loss": 0.0344, "step": 7976 }, { "epoch": 0.15, "learning_rate": 4.7419922494161186e-05, "loss": 0.0275, "step": 7978 }, { "epoch": 0.15, "learning_rate": 4.7419275533903955e-05, "loss": 0.0884, "step": 7980 }, { "epoch": 0.15, "learning_rate": 4.7418628573646725e-05, "loss": 0.0213, "step": 7982 }, { "epoch": 0.15, "learning_rate": 4.741798161338949e-05, "loss": 0.017, "step": 7984 }, { "epoch": 0.15, "learning_rate": 4.741733465313226e-05, "loss": 0.0123, "step": 7986 }, { "epoch": 0.16, "learning_rate": 4.7416687692875025e-05, "loss": 0.0252, "step": 7988 }, { "epoch": 0.16, "learning_rate": 4.74160407326178e-05, "loss": 0.0128, "step": 7990 }, { "epoch": 0.16, "learning_rate": 4.741539377236057e-05, "loss": 0.0158, "step": 7992 }, { "epoch": 0.16, "learning_rate": 4.741474681210333e-05, "loss": 0.0108, "step": 7994 }, { "epoch": 0.16, "learning_rate": 4.741409985184611e-05, "loss": 0.0205, "step": 7996 }, { "epoch": 0.16, "learning_rate": 4.741345289158887e-05, "loss": 0.0262, "step": 7998 }, { "epoch": 0.16, "learning_rate": 4.741280593133164e-05, "loss": 0.0246, "step": 8000 }, { "epoch": 0.16, "learning_rate": 4.741215897107441e-05, "loss": 0.012, "step": 8002 }, { "epoch": 0.16, "learning_rate": 4.741151201081718e-05, "loss": 0.0187, "step": 8004 }, { "epoch": 0.16, "learning_rate": 4.741086505055995e-05, "loss": 0.0095, "step": 8006 }, { "epoch": 0.16, "learning_rate": 4.7410218090302716e-05, "loss": 0.0142, "step": 8008 }, { "epoch": 0.16, "learning_rate": 4.7409571130045485e-05, "loss": 0.008, "step": 8010 }, { "epoch": 0.16, "learning_rate": 4.7408924169788254e-05, "loss": 0.0056, "step": 8012 }, { "epoch": 0.16, "learning_rate": 4.740827720953102e-05, "loss": 0.0358, "step": 8014 }, { "epoch": 0.16, "learning_rate": 4.7407630249273785e-05, "loss": 0.013, "step": 8016 }, { "epoch": 0.16, "learning_rate": 4.740698328901656e-05, "loss": 0.0092, "step": 8018 }, { "epoch": 0.16, "learning_rate": 4.7406336328759324e-05, "loss": 0.0135, "step": 8020 }, { "epoch": 0.16, "learning_rate": 4.740568936850209e-05, "loss": 0.0151, "step": 8022 }, { "epoch": 0.16, "learning_rate": 4.740504240824487e-05, "loss": 0.0054, "step": 8024 }, { "epoch": 0.16, "learning_rate": 4.740439544798763e-05, "loss": 0.0052, "step": 8026 }, { "epoch": 0.16, "learning_rate": 4.74037484877304e-05, "loss": 0.0517, "step": 8028 }, { "epoch": 0.16, "learning_rate": 4.740310152747317e-05, "loss": 0.0115, "step": 8030 }, { "epoch": 0.16, "learning_rate": 4.740245456721594e-05, "loss": 0.0115, "step": 8032 }, { "epoch": 0.16, "learning_rate": 4.740180760695871e-05, "loss": 0.0269, "step": 8034 }, { "epoch": 0.16, "learning_rate": 4.7401160646701476e-05, "loss": 0.0061, "step": 8036 }, { "epoch": 0.16, "learning_rate": 4.7400513686444245e-05, "loss": 0.0092, "step": 8038 }, { "epoch": 0.16, "learning_rate": 4.7399866726187014e-05, "loss": 0.0069, "step": 8040 }, { "epoch": 0.16, "learning_rate": 4.7399219765929784e-05, "loss": 0.0285, "step": 8042 }, { "epoch": 0.16, "learning_rate": 4.7398572805672546e-05, "loss": 0.0174, "step": 8044 }, { "epoch": 0.16, "learning_rate": 4.739792584541532e-05, "loss": 0.0131, "step": 8046 }, { "epoch": 0.16, "learning_rate": 4.7397278885158084e-05, "loss": 0.0048, "step": 8048 }, { "epoch": 0.16, "learning_rate": 4.739663192490086e-05, "loss": 0.0124, "step": 8050 }, { "epoch": 0.16, "learning_rate": 4.739598496464362e-05, "loss": 0.1072, "step": 8052 }, { "epoch": 0.16, "learning_rate": 4.739533800438639e-05, "loss": 0.0206, "step": 8054 }, { "epoch": 0.16, "learning_rate": 4.739469104412916e-05, "loss": 0.013, "step": 8056 }, { "epoch": 0.16, "learning_rate": 4.739404408387193e-05, "loss": 0.0053, "step": 8058 }, { "epoch": 0.16, "learning_rate": 4.73933971236147e-05, "loss": 0.006, "step": 8060 }, { "epoch": 0.16, "learning_rate": 4.739275016335747e-05, "loss": 0.0105, "step": 8062 }, { "epoch": 0.16, "learning_rate": 4.739210320310024e-05, "loss": 0.0143, "step": 8064 }, { "epoch": 0.16, "learning_rate": 4.7391456242843e-05, "loss": 0.0059, "step": 8066 }, { "epoch": 0.16, "learning_rate": 4.7390809282585775e-05, "loss": 0.0074, "step": 8068 }, { "epoch": 0.16, "learning_rate": 4.7390162322328544e-05, "loss": 0.0199, "step": 8070 }, { "epoch": 0.16, "learning_rate": 4.738951536207131e-05, "loss": 0.008, "step": 8072 }, { "epoch": 0.16, "learning_rate": 4.738886840181408e-05, "loss": 0.0031, "step": 8074 }, { "epoch": 0.16, "learning_rate": 4.7388221441556844e-05, "loss": 0.0216, "step": 8076 }, { "epoch": 0.16, "learning_rate": 4.738757448129962e-05, "loss": 0.0115, "step": 8078 }, { "epoch": 0.16, "learning_rate": 4.738692752104238e-05, "loss": 0.0381, "step": 8080 }, { "epoch": 0.16, "learning_rate": 4.738628056078515e-05, "loss": 0.0133, "step": 8082 }, { "epoch": 0.16, "learning_rate": 4.738563360052792e-05, "loss": 0.0104, "step": 8084 }, { "epoch": 0.16, "learning_rate": 4.738498664027069e-05, "loss": 0.0117, "step": 8086 }, { "epoch": 0.16, "learning_rate": 4.738433968001346e-05, "loss": 0.0247, "step": 8088 }, { "epoch": 0.16, "learning_rate": 4.738369271975623e-05, "loss": 0.0057, "step": 8090 }, { "epoch": 0.16, "learning_rate": 4.7383045759499e-05, "loss": 0.0147, "step": 8092 }, { "epoch": 0.16, "learning_rate": 4.7382398799241766e-05, "loss": 0.0388, "step": 8094 }, { "epoch": 0.16, "learning_rate": 4.7381751838984535e-05, "loss": 0.008, "step": 8096 }, { "epoch": 0.16, "learning_rate": 4.73811048787273e-05, "loss": 0.0235, "step": 8098 }, { "epoch": 0.16, "learning_rate": 4.7380457918470073e-05, "loss": 0.0096, "step": 8100 }, { "epoch": 0.16, "learning_rate": 4.7379810958212836e-05, "loss": 0.0049, "step": 8102 }, { "epoch": 0.16, "learning_rate": 4.7379163997955605e-05, "loss": 0.0033, "step": 8104 }, { "epoch": 0.16, "learning_rate": 4.737851703769838e-05, "loss": 0.0278, "step": 8106 }, { "epoch": 0.16, "learning_rate": 4.737787007744114e-05, "loss": 0.0056, "step": 8108 }, { "epoch": 0.16, "learning_rate": 4.737722311718391e-05, "loss": 0.0122, "step": 8110 }, { "epoch": 0.16, "learning_rate": 4.737657615692668e-05, "loss": 0.0031, "step": 8112 }, { "epoch": 0.16, "learning_rate": 4.737592919666945e-05, "loss": 0.0072, "step": 8114 }, { "epoch": 0.16, "learning_rate": 4.737528223641222e-05, "loss": 0.0321, "step": 8116 }, { "epoch": 0.16, "learning_rate": 4.737463527615499e-05, "loss": 0.0252, "step": 8118 }, { "epoch": 0.16, "learning_rate": 4.737398831589776e-05, "loss": 0.0236, "step": 8120 }, { "epoch": 0.16, "learning_rate": 4.7373341355640527e-05, "loss": 0.0273, "step": 8122 }, { "epoch": 0.16, "learning_rate": 4.7372694395383296e-05, "loss": 0.0193, "step": 8124 }, { "epoch": 0.16, "learning_rate": 4.737204743512606e-05, "loss": 0.0302, "step": 8126 }, { "epoch": 0.16, "learning_rate": 4.7371400474868834e-05, "loss": 0.0046, "step": 8128 }, { "epoch": 0.16, "learning_rate": 4.7370753514611596e-05, "loss": 0.0149, "step": 8130 }, { "epoch": 0.16, "learning_rate": 4.737010655435437e-05, "loss": 0.0088, "step": 8132 }, { "epoch": 0.16, "learning_rate": 4.7369459594097134e-05, "loss": 0.0039, "step": 8134 }, { "epoch": 0.16, "learning_rate": 4.7368812633839903e-05, "loss": 0.0033, "step": 8136 }, { "epoch": 0.16, "learning_rate": 4.736816567358268e-05, "loss": 0.0042, "step": 8138 }, { "epoch": 0.16, "learning_rate": 4.736751871332544e-05, "loss": 0.0109, "step": 8140 }, { "epoch": 0.16, "learning_rate": 4.736687175306821e-05, "loss": 0.0187, "step": 8142 }, { "epoch": 0.16, "learning_rate": 4.736622479281098e-05, "loss": 0.0051, "step": 8144 }, { "epoch": 0.16, "learning_rate": 4.736557783255375e-05, "loss": 0.0102, "step": 8146 }, { "epoch": 0.16, "learning_rate": 4.736493087229652e-05, "loss": 0.008, "step": 8148 }, { "epoch": 0.16, "learning_rate": 4.736428391203929e-05, "loss": 0.0087, "step": 8150 }, { "epoch": 0.16, "learning_rate": 4.7363636951782056e-05, "loss": 0.006, "step": 8152 }, { "epoch": 0.16, "learning_rate": 4.7362989991524825e-05, "loss": 0.0038, "step": 8154 }, { "epoch": 0.16, "learning_rate": 4.7362343031267594e-05, "loss": 0.0197, "step": 8156 }, { "epoch": 0.16, "learning_rate": 4.7361696071010357e-05, "loss": 0.0182, "step": 8158 }, { "epoch": 0.16, "learning_rate": 4.736104911075313e-05, "loss": 0.0227, "step": 8160 }, { "epoch": 0.16, "learning_rate": 4.7360402150495895e-05, "loss": 0.0032, "step": 8162 }, { "epoch": 0.16, "learning_rate": 4.7359755190238664e-05, "loss": 0.0271, "step": 8164 }, { "epoch": 0.16, "learning_rate": 4.735910822998143e-05, "loss": 0.0107, "step": 8166 }, { "epoch": 0.16, "learning_rate": 4.73584612697242e-05, "loss": 0.0276, "step": 8168 }, { "epoch": 0.16, "learning_rate": 4.735781430946697e-05, "loss": 0.0022, "step": 8170 }, { "epoch": 0.16, "learning_rate": 4.735716734920974e-05, "loss": 0.0048, "step": 8172 }, { "epoch": 0.16, "learning_rate": 4.735652038895251e-05, "loss": 0.0287, "step": 8174 }, { "epoch": 0.16, "learning_rate": 4.735587342869528e-05, "loss": 0.0215, "step": 8176 }, { "epoch": 0.16, "learning_rate": 4.735522646843805e-05, "loss": 0.0181, "step": 8178 }, { "epoch": 0.16, "learning_rate": 4.735457950818081e-05, "loss": 0.0159, "step": 8180 }, { "epoch": 0.16, "learning_rate": 4.7353932547923586e-05, "loss": 0.0221, "step": 8182 }, { "epoch": 0.16, "learning_rate": 4.7353285587666355e-05, "loss": 0.0177, "step": 8184 }, { "epoch": 0.16, "learning_rate": 4.735263862740912e-05, "loss": 0.0181, "step": 8186 }, { "epoch": 0.16, "learning_rate": 4.735199166715189e-05, "loss": 0.0109, "step": 8188 }, { "epoch": 0.16, "learning_rate": 4.7351344706894655e-05, "loss": 0.0125, "step": 8190 }, { "epoch": 0.16, "learning_rate": 4.735069774663743e-05, "loss": 0.0173, "step": 8192 }, { "epoch": 0.16, "learning_rate": 4.735005078638019e-05, "loss": 0.0021, "step": 8194 }, { "epoch": 0.16, "learning_rate": 4.734940382612296e-05, "loss": 0.0166, "step": 8196 }, { "epoch": 0.16, "learning_rate": 4.734875686586573e-05, "loss": 0.0199, "step": 8198 }, { "epoch": 0.16, "learning_rate": 4.73481099056085e-05, "loss": 0.0517, "step": 8200 }, { "epoch": 0.16, "learning_rate": 4.734746294535127e-05, "loss": 0.0296, "step": 8202 }, { "epoch": 0.16, "learning_rate": 4.734681598509404e-05, "loss": 0.0073, "step": 8204 }, { "epoch": 0.16, "learning_rate": 4.734616902483681e-05, "loss": 0.0156, "step": 8206 }, { "epoch": 0.16, "learning_rate": 4.734552206457957e-05, "loss": 0.007, "step": 8208 }, { "epoch": 0.16, "learning_rate": 4.7344875104322346e-05, "loss": 0.006, "step": 8210 }, { "epoch": 0.16, "learning_rate": 4.734422814406511e-05, "loss": 0.0078, "step": 8212 }, { "epoch": 0.16, "learning_rate": 4.7343581183807884e-05, "loss": 0.0114, "step": 8214 }, { "epoch": 0.16, "learning_rate": 4.734293422355065e-05, "loss": 0.0281, "step": 8216 }, { "epoch": 0.16, "learning_rate": 4.7342287263293416e-05, "loss": 0.0283, "step": 8218 }, { "epoch": 0.16, "learning_rate": 4.734164030303619e-05, "loss": 0.005, "step": 8220 }, { "epoch": 0.16, "learning_rate": 4.7340993342778954e-05, "loss": 0.0164, "step": 8222 }, { "epoch": 0.16, "learning_rate": 4.734034638252172e-05, "loss": 0.0293, "step": 8224 }, { "epoch": 0.16, "learning_rate": 4.733969942226449e-05, "loss": 0.0172, "step": 8226 }, { "epoch": 0.16, "learning_rate": 4.733905246200726e-05, "loss": 0.0267, "step": 8228 }, { "epoch": 0.16, "learning_rate": 4.733840550175003e-05, "loss": 0.0099, "step": 8230 }, { "epoch": 0.16, "learning_rate": 4.73377585414928e-05, "loss": 0.0165, "step": 8232 }, { "epoch": 0.16, "learning_rate": 4.733711158123557e-05, "loss": 0.0099, "step": 8234 }, { "epoch": 0.16, "learning_rate": 4.733646462097834e-05, "loss": 0.0149, "step": 8236 }, { "epoch": 0.16, "learning_rate": 4.7335817660721106e-05, "loss": 0.0116, "step": 8238 }, { "epoch": 0.16, "learning_rate": 4.733517070046387e-05, "loss": 0.0168, "step": 8240 }, { "epoch": 0.16, "learning_rate": 4.7334523740206645e-05, "loss": 0.0391, "step": 8242 }, { "epoch": 0.16, "learning_rate": 4.733387677994941e-05, "loss": 0.0179, "step": 8244 }, { "epoch": 0.16, "learning_rate": 4.7333229819692176e-05, "loss": 0.014, "step": 8246 }, { "epoch": 0.16, "learning_rate": 4.733258285943495e-05, "loss": 0.0041, "step": 8248 }, { "epoch": 0.16, "learning_rate": 4.7331935899177714e-05, "loss": 0.0269, "step": 8250 }, { "epoch": 0.16, "learning_rate": 4.733128893892048e-05, "loss": 0.0203, "step": 8252 }, { "epoch": 0.16, "learning_rate": 4.733064197866325e-05, "loss": 0.0079, "step": 8254 }, { "epoch": 0.16, "learning_rate": 4.732999501840602e-05, "loss": 0.0126, "step": 8256 }, { "epoch": 0.16, "learning_rate": 4.732934805814879e-05, "loss": 0.0504, "step": 8258 }, { "epoch": 0.16, "learning_rate": 4.732870109789156e-05, "loss": 0.0497, "step": 8260 }, { "epoch": 0.16, "learning_rate": 4.732805413763433e-05, "loss": 0.0176, "step": 8262 }, { "epoch": 0.16, "learning_rate": 4.73274071773771e-05, "loss": 0.0034, "step": 8264 }, { "epoch": 0.16, "learning_rate": 4.732676021711987e-05, "loss": 0.008, "step": 8266 }, { "epoch": 0.16, "learning_rate": 4.732611325686263e-05, "loss": 0.011, "step": 8268 }, { "epoch": 0.16, "learning_rate": 4.7325466296605405e-05, "loss": 0.0022, "step": 8270 }, { "epoch": 0.16, "learning_rate": 4.732481933634817e-05, "loss": 0.0223, "step": 8272 }, { "epoch": 0.16, "learning_rate": 4.732417237609094e-05, "loss": 0.0034, "step": 8274 }, { "epoch": 0.16, "learning_rate": 4.7323525415833705e-05, "loss": 0.0322, "step": 8276 }, { "epoch": 0.16, "learning_rate": 4.7322878455576475e-05, "loss": 0.0118, "step": 8278 }, { "epoch": 0.16, "learning_rate": 4.7322231495319244e-05, "loss": 0.0158, "step": 8280 }, { "epoch": 0.16, "learning_rate": 4.732158453506201e-05, "loss": 0.0103, "step": 8282 }, { "epoch": 0.16, "learning_rate": 4.732093757480478e-05, "loss": 0.0111, "step": 8284 }, { "epoch": 0.16, "learning_rate": 4.732029061454755e-05, "loss": 0.0156, "step": 8286 }, { "epoch": 0.16, "learning_rate": 4.731964365429032e-05, "loss": 0.0103, "step": 8288 }, { "epoch": 0.16, "learning_rate": 4.731899669403308e-05, "loss": 0.0267, "step": 8290 }, { "epoch": 0.16, "learning_rate": 4.731834973377586e-05, "loss": 0.0075, "step": 8292 }, { "epoch": 0.16, "learning_rate": 4.731770277351863e-05, "loss": 0.0912, "step": 8294 }, { "epoch": 0.16, "learning_rate": 4.7317055813261396e-05, "loss": 0.0036, "step": 8296 }, { "epoch": 0.16, "learning_rate": 4.7316408853004165e-05, "loss": 0.0159, "step": 8298 }, { "epoch": 0.16, "learning_rate": 4.731576189274693e-05, "loss": 0.0175, "step": 8300 }, { "epoch": 0.16, "learning_rate": 4.7315114932489704e-05, "loss": 0.0036, "step": 8302 }, { "epoch": 0.16, "learning_rate": 4.7314467972232466e-05, "loss": 0.0057, "step": 8304 }, { "epoch": 0.16, "learning_rate": 4.7313821011975235e-05, "loss": 0.0066, "step": 8306 }, { "epoch": 0.16, "learning_rate": 4.7313174051718004e-05, "loss": 0.0714, "step": 8308 }, { "epoch": 0.16, "learning_rate": 4.731252709146077e-05, "loss": 0.0272, "step": 8310 }, { "epoch": 0.16, "learning_rate": 4.731188013120354e-05, "loss": 0.0056, "step": 8312 }, { "epoch": 0.16, "learning_rate": 4.731123317094631e-05, "loss": 0.0216, "step": 8314 }, { "epoch": 0.16, "learning_rate": 4.731058621068908e-05, "loss": 0.0157, "step": 8316 }, { "epoch": 0.16, "learning_rate": 4.730993925043185e-05, "loss": 0.0359, "step": 8318 }, { "epoch": 0.16, "learning_rate": 4.730929229017462e-05, "loss": 0.0048, "step": 8320 }, { "epoch": 0.16, "learning_rate": 4.730864532991738e-05, "loss": 0.0217, "step": 8322 }, { "epoch": 0.16, "learning_rate": 4.730799836966016e-05, "loss": 0.0234, "step": 8324 }, { "epoch": 0.16, "learning_rate": 4.7307351409402926e-05, "loss": 0.0189, "step": 8326 }, { "epoch": 0.16, "learning_rate": 4.730670444914569e-05, "loss": 0.0186, "step": 8328 }, { "epoch": 0.16, "learning_rate": 4.7306057488888464e-05, "loss": 0.0095, "step": 8330 }, { "epoch": 0.16, "learning_rate": 4.7305410528631226e-05, "loss": 0.0157, "step": 8332 }, { "epoch": 0.16, "learning_rate": 4.7304763568374e-05, "loss": 0.0708, "step": 8334 }, { "epoch": 0.16, "learning_rate": 4.7304116608116764e-05, "loss": 0.0079, "step": 8336 }, { "epoch": 0.16, "learning_rate": 4.7303469647859533e-05, "loss": 0.0125, "step": 8338 }, { "epoch": 0.16, "learning_rate": 4.73028226876023e-05, "loss": 0.0115, "step": 8340 }, { "epoch": 0.16, "learning_rate": 4.730217572734507e-05, "loss": 0.0198, "step": 8342 }, { "epoch": 0.16, "learning_rate": 4.730152876708784e-05, "loss": 0.0067, "step": 8344 }, { "epoch": 0.16, "learning_rate": 4.730088180683061e-05, "loss": 0.0059, "step": 8346 }, { "epoch": 0.16, "learning_rate": 4.730023484657338e-05, "loss": 0.0099, "step": 8348 }, { "epoch": 0.16, "learning_rate": 4.729958788631614e-05, "loss": 0.0142, "step": 8350 }, { "epoch": 0.16, "learning_rate": 4.729894092605892e-05, "loss": 0.0226, "step": 8352 }, { "epoch": 0.16, "learning_rate": 4.729829396580168e-05, "loss": 0.0186, "step": 8354 }, { "epoch": 0.16, "learning_rate": 4.7297647005544455e-05, "loss": 0.0016, "step": 8356 }, { "epoch": 0.16, "learning_rate": 4.729700004528722e-05, "loss": 0.0058, "step": 8358 }, { "epoch": 0.16, "learning_rate": 4.729635308502999e-05, "loss": 0.0312, "step": 8360 }, { "epoch": 0.16, "learning_rate": 4.729570612477276e-05, "loss": 0.0289, "step": 8362 }, { "epoch": 0.16, "learning_rate": 4.7295059164515525e-05, "loss": 0.0113, "step": 8364 }, { "epoch": 0.16, "learning_rate": 4.7294412204258294e-05, "loss": 0.0136, "step": 8366 }, { "epoch": 0.16, "learning_rate": 4.729376524400106e-05, "loss": 0.0301, "step": 8368 }, { "epoch": 0.16, "learning_rate": 4.729311828374383e-05, "loss": 0.0301, "step": 8370 }, { "epoch": 0.16, "learning_rate": 4.72924713234866e-05, "loss": 0.0187, "step": 8372 }, { "epoch": 0.16, "learning_rate": 4.729182436322937e-05, "loss": 0.0166, "step": 8374 }, { "epoch": 0.16, "learning_rate": 4.729117740297214e-05, "loss": 0.025, "step": 8376 }, { "epoch": 0.16, "learning_rate": 4.729053044271491e-05, "loss": 0.0181, "step": 8378 }, { "epoch": 0.16, "learning_rate": 4.728988348245768e-05, "loss": 0.0206, "step": 8380 }, { "epoch": 0.16, "learning_rate": 4.728923652220044e-05, "loss": 0.0122, "step": 8382 }, { "epoch": 0.16, "learning_rate": 4.7288589561943216e-05, "loss": 0.0161, "step": 8384 }, { "epoch": 0.16, "learning_rate": 4.728794260168598e-05, "loss": 0.0137, "step": 8386 }, { "epoch": 0.16, "learning_rate": 4.728729564142875e-05, "loss": 0.0085, "step": 8388 }, { "epoch": 0.16, "learning_rate": 4.7286648681171516e-05, "loss": 0.0088, "step": 8390 }, { "epoch": 0.16, "learning_rate": 4.7286001720914285e-05, "loss": 0.0199, "step": 8392 }, { "epoch": 0.16, "learning_rate": 4.7285354760657054e-05, "loss": 0.0168, "step": 8394 }, { "epoch": 0.16, "learning_rate": 4.7284707800399823e-05, "loss": 0.0045, "step": 8396 }, { "epoch": 0.16, "learning_rate": 4.728406084014259e-05, "loss": 0.0084, "step": 8398 }, { "epoch": 0.16, "learning_rate": 4.728341387988536e-05, "loss": 0.0388, "step": 8400 }, { "epoch": 0.16, "learning_rate": 4.728276691962813e-05, "loss": 0.0074, "step": 8402 }, { "epoch": 0.16, "learning_rate": 4.728211995937089e-05, "loss": 0.0108, "step": 8404 }, { "epoch": 0.16, "learning_rate": 4.728147299911367e-05, "loss": 0.0144, "step": 8406 }, { "epoch": 0.16, "learning_rate": 4.728082603885644e-05, "loss": 0.0176, "step": 8408 }, { "epoch": 0.16, "learning_rate": 4.72801790785992e-05, "loss": 0.0059, "step": 8410 }, { "epoch": 0.16, "learning_rate": 4.7279532118341976e-05, "loss": 0.0039, "step": 8412 }, { "epoch": 0.16, "learning_rate": 4.727888515808474e-05, "loss": 0.0097, "step": 8414 }, { "epoch": 0.16, "learning_rate": 4.7278238197827514e-05, "loss": 0.026, "step": 8416 }, { "epoch": 0.16, "learning_rate": 4.7277591237570277e-05, "loss": 0.0046, "step": 8418 }, { "epoch": 0.16, "learning_rate": 4.7276944277313046e-05, "loss": 0.0053, "step": 8420 }, { "epoch": 0.16, "learning_rate": 4.7276297317055815e-05, "loss": 0.0186, "step": 8422 }, { "epoch": 0.16, "learning_rate": 4.7275650356798584e-05, "loss": 0.013, "step": 8424 }, { "epoch": 0.16, "learning_rate": 4.727500339654135e-05, "loss": 0.0095, "step": 8426 }, { "epoch": 0.16, "learning_rate": 4.727435643628412e-05, "loss": 0.0485, "step": 8428 }, { "epoch": 0.16, "learning_rate": 4.727370947602689e-05, "loss": 0.0039, "step": 8430 }, { "epoch": 0.16, "learning_rate": 4.727306251576965e-05, "loss": 0.0082, "step": 8432 }, { "epoch": 0.16, "learning_rate": 4.727241555551243e-05, "loss": 0.0227, "step": 8434 }, { "epoch": 0.16, "learning_rate": 4.727176859525519e-05, "loss": 0.0026, "step": 8436 }, { "epoch": 0.16, "learning_rate": 4.727112163499797e-05, "loss": 0.024, "step": 8438 }, { "epoch": 0.16, "learning_rate": 4.7270474674740736e-05, "loss": 0.0279, "step": 8440 }, { "epoch": 0.16, "learning_rate": 4.72698277144835e-05, "loss": 0.0167, "step": 8442 }, { "epoch": 0.16, "learning_rate": 4.7269180754226275e-05, "loss": 0.0137, "step": 8444 }, { "epoch": 0.16, "learning_rate": 4.726853379396904e-05, "loss": 0.0144, "step": 8446 }, { "epoch": 0.16, "learning_rate": 4.7267886833711806e-05, "loss": 0.0297, "step": 8448 }, { "epoch": 0.16, "learning_rate": 4.7267239873454575e-05, "loss": 0.0151, "step": 8450 }, { "epoch": 0.16, "learning_rate": 4.7266592913197344e-05, "loss": 0.0078, "step": 8452 }, { "epoch": 0.16, "learning_rate": 4.726594595294011e-05, "loss": 0.0115, "step": 8454 }, { "epoch": 0.16, "learning_rate": 4.726529899268288e-05, "loss": 0.0081, "step": 8456 }, { "epoch": 0.16, "learning_rate": 4.726465203242565e-05, "loss": 0.0155, "step": 8458 }, { "epoch": 0.16, "learning_rate": 4.726400507216842e-05, "loss": 0.0309, "step": 8460 }, { "epoch": 0.16, "learning_rate": 4.726335811191119e-05, "loss": 0.0342, "step": 8462 }, { "epoch": 0.16, "learning_rate": 4.726271115165395e-05, "loss": 0.0096, "step": 8464 }, { "epoch": 0.16, "learning_rate": 4.726206419139673e-05, "loss": 0.0324, "step": 8466 }, { "epoch": 0.16, "learning_rate": 4.726141723113949e-05, "loss": 0.0119, "step": 8468 }, { "epoch": 0.16, "learning_rate": 4.726077027088226e-05, "loss": 0.0078, "step": 8470 }, { "epoch": 0.16, "learning_rate": 4.7260123310625035e-05, "loss": 0.0173, "step": 8472 }, { "epoch": 0.16, "learning_rate": 4.72594763503678e-05, "loss": 0.0265, "step": 8474 }, { "epoch": 0.16, "learning_rate": 4.725882939011057e-05, "loss": 0.0295, "step": 8476 }, { "epoch": 0.16, "learning_rate": 4.7258182429853336e-05, "loss": 0.0077, "step": 8478 }, { "epoch": 0.16, "learning_rate": 4.7257535469596105e-05, "loss": 0.0208, "step": 8480 }, { "epoch": 0.16, "learning_rate": 4.7256888509338874e-05, "loss": 0.004, "step": 8482 }, { "epoch": 0.16, "learning_rate": 4.725624154908164e-05, "loss": 0.0265, "step": 8484 }, { "epoch": 0.16, "learning_rate": 4.725559458882441e-05, "loss": 0.0105, "step": 8486 }, { "epoch": 0.16, "learning_rate": 4.725494762856718e-05, "loss": 0.0233, "step": 8488 }, { "epoch": 0.16, "learning_rate": 4.725430066830995e-05, "loss": 0.0072, "step": 8490 }, { "epoch": 0.16, "learning_rate": 4.725365370805271e-05, "loss": 0.0143, "step": 8492 }, { "epoch": 0.16, "learning_rate": 4.725300674779549e-05, "loss": 0.0104, "step": 8494 }, { "epoch": 0.16, "learning_rate": 4.725235978753825e-05, "loss": 0.0356, "step": 8496 }, { "epoch": 0.16, "learning_rate": 4.7251712827281026e-05, "loss": 0.0062, "step": 8498 }, { "epoch": 0.16, "learning_rate": 4.725106586702379e-05, "loss": 0.0092, "step": 8500 }, { "epoch": 0.17, "learning_rate": 4.725041890676656e-05, "loss": 0.0146, "step": 8502 }, { "epoch": 0.17, "learning_rate": 4.7249771946509334e-05, "loss": 0.0074, "step": 8504 }, { "epoch": 0.17, "learning_rate": 4.7249124986252096e-05, "loss": 0.0133, "step": 8506 }, { "epoch": 0.17, "learning_rate": 4.7248478025994865e-05, "loss": 0.0047, "step": 8508 }, { "epoch": 0.17, "learning_rate": 4.7247831065737634e-05, "loss": 0.0043, "step": 8510 }, { "epoch": 0.17, "learning_rate": 4.72471841054804e-05, "loss": 0.0027, "step": 8512 }, { "epoch": 0.17, "learning_rate": 4.7246537145223165e-05, "loss": 0.0095, "step": 8514 }, { "epoch": 0.17, "learning_rate": 4.724589018496594e-05, "loss": 0.0307, "step": 8516 }, { "epoch": 0.17, "learning_rate": 4.724524322470871e-05, "loss": 0.0071, "step": 8518 }, { "epoch": 0.17, "learning_rate": 4.724459626445148e-05, "loss": 0.0044, "step": 8520 }, { "epoch": 0.17, "learning_rate": 4.724394930419425e-05, "loss": 0.0069, "step": 8522 }, { "epoch": 0.17, "learning_rate": 4.724330234393701e-05, "loss": 0.0087, "step": 8524 }, { "epoch": 0.17, "learning_rate": 4.724265538367979e-05, "loss": 0.019, "step": 8526 }, { "epoch": 0.17, "learning_rate": 4.724200842342255e-05, "loss": 0.0303, "step": 8528 }, { "epoch": 0.17, "learning_rate": 4.724136146316532e-05, "loss": 0.0077, "step": 8530 }, { "epoch": 0.17, "learning_rate": 4.724071450290809e-05, "loss": 0.0046, "step": 8532 }, { "epoch": 0.17, "learning_rate": 4.7240067542650856e-05, "loss": 0.0032, "step": 8534 }, { "epoch": 0.17, "learning_rate": 4.7239420582393625e-05, "loss": 0.0095, "step": 8536 }, { "epoch": 0.17, "learning_rate": 4.7238773622136394e-05, "loss": 0.0171, "step": 8538 }, { "epoch": 0.17, "learning_rate": 4.7238126661879164e-05, "loss": 0.0427, "step": 8540 }, { "epoch": 0.17, "learning_rate": 4.723747970162193e-05, "loss": 0.0073, "step": 8542 }, { "epoch": 0.17, "learning_rate": 4.72368327413647e-05, "loss": 0.0134, "step": 8544 }, { "epoch": 0.17, "learning_rate": 4.7236185781107464e-05, "loss": 0.0154, "step": 8546 }, { "epoch": 0.17, "learning_rate": 4.723553882085024e-05, "loss": 0.0126, "step": 8548 }, { "epoch": 0.17, "learning_rate": 4.723489186059301e-05, "loss": 0.0045, "step": 8550 }, { "epoch": 0.17, "learning_rate": 4.723424490033577e-05, "loss": 0.0056, "step": 8552 }, { "epoch": 0.17, "learning_rate": 4.723359794007855e-05, "loss": 0.0171, "step": 8554 }, { "epoch": 0.17, "learning_rate": 4.723295097982131e-05, "loss": 0.0146, "step": 8556 }, { "epoch": 0.17, "learning_rate": 4.7232304019564085e-05, "loss": 0.0112, "step": 8558 }, { "epoch": 0.17, "learning_rate": 4.723165705930685e-05, "loss": 0.0167, "step": 8560 }, { "epoch": 0.17, "learning_rate": 4.723101009904962e-05, "loss": 0.0595, "step": 8562 }, { "epoch": 0.17, "learning_rate": 4.7230363138792386e-05, "loss": 0.0126, "step": 8564 }, { "epoch": 0.17, "learning_rate": 4.7229716178535155e-05, "loss": 0.0287, "step": 8566 }, { "epoch": 0.17, "learning_rate": 4.7229069218277924e-05, "loss": 0.0153, "step": 8568 }, { "epoch": 0.17, "learning_rate": 4.722842225802069e-05, "loss": 0.0093, "step": 8570 }, { "epoch": 0.17, "learning_rate": 4.722777529776346e-05, "loss": 0.009, "step": 8572 }, { "epoch": 0.17, "learning_rate": 4.7227128337506224e-05, "loss": 0.0114, "step": 8574 }, { "epoch": 0.17, "learning_rate": 4.7226481377249e-05, "loss": 0.0113, "step": 8576 }, { "epoch": 0.17, "learning_rate": 4.722583441699176e-05, "loss": 0.0043, "step": 8578 }, { "epoch": 0.17, "learning_rate": 4.722518745673454e-05, "loss": 0.0093, "step": 8580 }, { "epoch": 0.17, "learning_rate": 4.72245404964773e-05, "loss": 0.0048, "step": 8582 }, { "epoch": 0.17, "learning_rate": 4.722389353622007e-05, "loss": 0.0148, "step": 8584 }, { "epoch": 0.17, "learning_rate": 4.7223246575962846e-05, "loss": 0.0197, "step": 8586 }, { "epoch": 0.17, "learning_rate": 4.722259961570561e-05, "loss": 0.0167, "step": 8588 }, { "epoch": 0.17, "learning_rate": 4.722195265544838e-05, "loss": 0.0165, "step": 8590 }, { "epoch": 0.17, "learning_rate": 4.7221305695191146e-05, "loss": 0.0082, "step": 8592 }, { "epoch": 0.17, "learning_rate": 4.7220658734933915e-05, "loss": 0.0102, "step": 8594 }, { "epoch": 0.17, "learning_rate": 4.7220011774676684e-05, "loss": 0.0233, "step": 8596 }, { "epoch": 0.17, "learning_rate": 4.7219364814419453e-05, "loss": 0.0091, "step": 8598 }, { "epoch": 0.17, "learning_rate": 4.721871785416222e-05, "loss": 0.0072, "step": 8600 }, { "epoch": 0.17, "learning_rate": 4.721807089390499e-05, "loss": 0.01, "step": 8602 }, { "epoch": 0.17, "learning_rate": 4.721742393364776e-05, "loss": 0.0073, "step": 8604 }, { "epoch": 0.17, "learning_rate": 4.721677697339052e-05, "loss": 0.0023, "step": 8606 }, { "epoch": 0.17, "learning_rate": 4.72161300131333e-05, "loss": 0.0021, "step": 8608 }, { "epoch": 0.17, "learning_rate": 4.721548305287606e-05, "loss": 0.0167, "step": 8610 }, { "epoch": 0.17, "learning_rate": 4.721483609261883e-05, "loss": 0.0378, "step": 8612 }, { "epoch": 0.17, "learning_rate": 4.72141891323616e-05, "loss": 0.0168, "step": 8614 }, { "epoch": 0.17, "learning_rate": 4.721354217210437e-05, "loss": 0.0075, "step": 8616 }, { "epoch": 0.17, "learning_rate": 4.7212895211847144e-05, "loss": 0.0176, "step": 8618 }, { "epoch": 0.17, "learning_rate": 4.721224825158991e-05, "loss": 0.0074, "step": 8620 }, { "epoch": 0.17, "learning_rate": 4.7211601291332676e-05, "loss": 0.0022, "step": 8622 }, { "epoch": 0.17, "learning_rate": 4.7210954331075445e-05, "loss": 0.0176, "step": 8624 }, { "epoch": 0.17, "learning_rate": 4.7210307370818214e-05, "loss": 0.0369, "step": 8626 }, { "epoch": 0.17, "learning_rate": 4.720966041056098e-05, "loss": 0.0262, "step": 8628 }, { "epoch": 0.17, "learning_rate": 4.720901345030375e-05, "loss": 0.0144, "step": 8630 }, { "epoch": 0.17, "learning_rate": 4.720836649004652e-05, "loss": 0.0211, "step": 8632 }, { "epoch": 0.17, "learning_rate": 4.7207719529789283e-05, "loss": 0.0073, "step": 8634 }, { "epoch": 0.17, "learning_rate": 4.720707256953206e-05, "loss": 0.0151, "step": 8636 }, { "epoch": 0.17, "learning_rate": 4.720642560927482e-05, "loss": 0.0134, "step": 8638 }, { "epoch": 0.17, "learning_rate": 4.72057786490176e-05, "loss": 0.0042, "step": 8640 }, { "epoch": 0.17, "learning_rate": 4.720513168876036e-05, "loss": 0.0071, "step": 8642 }, { "epoch": 0.17, "learning_rate": 4.720448472850313e-05, "loss": 0.024, "step": 8644 }, { "epoch": 0.17, "learning_rate": 4.72038377682459e-05, "loss": 0.027, "step": 8646 }, { "epoch": 0.17, "learning_rate": 4.720319080798867e-05, "loss": 0.0158, "step": 8648 }, { "epoch": 0.17, "learning_rate": 4.7202543847731436e-05, "loss": 0.0126, "step": 8650 }, { "epoch": 0.17, "learning_rate": 4.7201896887474205e-05, "loss": 0.0108, "step": 8652 }, { "epoch": 0.17, "learning_rate": 4.7201249927216974e-05, "loss": 0.0134, "step": 8654 }, { "epoch": 0.17, "learning_rate": 4.7200602966959737e-05, "loss": 0.0278, "step": 8656 }, { "epoch": 0.17, "learning_rate": 4.719995600670251e-05, "loss": 0.0229, "step": 8658 }, { "epoch": 0.17, "learning_rate": 4.7199309046445275e-05, "loss": 0.012, "step": 8660 }, { "epoch": 0.17, "learning_rate": 4.719866208618805e-05, "loss": 0.0211, "step": 8662 }, { "epoch": 0.17, "learning_rate": 4.719801512593082e-05, "loss": 0.0093, "step": 8664 }, { "epoch": 0.17, "learning_rate": 4.719736816567358e-05, "loss": 0.0142, "step": 8666 }, { "epoch": 0.17, "learning_rate": 4.719672120541636e-05, "loss": 0.0076, "step": 8668 }, { "epoch": 0.17, "learning_rate": 4.719607424515912e-05, "loss": 0.0076, "step": 8670 }, { "epoch": 0.17, "learning_rate": 4.719542728490189e-05, "loss": 0.0072, "step": 8672 }, { "epoch": 0.17, "learning_rate": 4.719478032464466e-05, "loss": 0.044, "step": 8674 }, { "epoch": 0.17, "learning_rate": 4.719413336438743e-05, "loss": 0.0073, "step": 8676 }, { "epoch": 0.17, "learning_rate": 4.7193486404130197e-05, "loss": 0.022, "step": 8678 }, { "epoch": 0.17, "learning_rate": 4.7192839443872966e-05, "loss": 0.0053, "step": 8680 }, { "epoch": 0.17, "learning_rate": 4.7192192483615735e-05, "loss": 0.0082, "step": 8682 }, { "epoch": 0.17, "learning_rate": 4.7191545523358504e-05, "loss": 0.0216, "step": 8684 }, { "epoch": 0.17, "learning_rate": 4.719089856310127e-05, "loss": 0.0111, "step": 8686 }, { "epoch": 0.17, "learning_rate": 4.7190251602844035e-05, "loss": 0.0016, "step": 8688 }, { "epoch": 0.17, "learning_rate": 4.718960464258681e-05, "loss": 0.0051, "step": 8690 }, { "epoch": 0.17, "learning_rate": 4.718895768232957e-05, "loss": 0.0055, "step": 8692 }, { "epoch": 0.17, "learning_rate": 4.718831072207234e-05, "loss": 0.0081, "step": 8694 }, { "epoch": 0.17, "learning_rate": 4.718766376181512e-05, "loss": 0.0228, "step": 8696 }, { "epoch": 0.17, "learning_rate": 4.718701680155788e-05, "loss": 0.0221, "step": 8698 }, { "epoch": 0.17, "learning_rate": 4.7186369841300656e-05, "loss": 0.0178, "step": 8700 }, { "epoch": 0.17, "learning_rate": 4.718572288104342e-05, "loss": 0.0386, "step": 8702 }, { "epoch": 0.17, "learning_rate": 4.718507592078619e-05, "loss": 0.0057, "step": 8704 }, { "epoch": 0.17, "learning_rate": 4.718442896052896e-05, "loss": 0.0352, "step": 8706 }, { "epoch": 0.17, "learning_rate": 4.7183782000271726e-05, "loss": 0.076, "step": 8708 }, { "epoch": 0.17, "learning_rate": 4.7183135040014495e-05, "loss": 0.0089, "step": 8710 }, { "epoch": 0.17, "learning_rate": 4.7182488079757264e-05, "loss": 0.0209, "step": 8712 }, { "epoch": 0.17, "learning_rate": 4.718184111950003e-05, "loss": 0.0164, "step": 8714 }, { "epoch": 0.17, "learning_rate": 4.7181194159242796e-05, "loss": 0.0073, "step": 8716 }, { "epoch": 0.17, "learning_rate": 4.718054719898557e-05, "loss": 0.0122, "step": 8718 }, { "epoch": 0.17, "learning_rate": 4.7179900238728334e-05, "loss": 0.0194, "step": 8720 }, { "epoch": 0.17, "learning_rate": 4.717925327847111e-05, "loss": 0.0067, "step": 8722 }, { "epoch": 0.17, "learning_rate": 4.717860631821387e-05, "loss": 0.0076, "step": 8724 }, { "epoch": 0.17, "learning_rate": 4.717795935795664e-05, "loss": 0.0119, "step": 8726 }, { "epoch": 0.17, "learning_rate": 4.717731239769942e-05, "loss": 0.0169, "step": 8728 }, { "epoch": 0.17, "learning_rate": 4.717666543744218e-05, "loss": 0.0378, "step": 8730 }, { "epoch": 0.17, "learning_rate": 4.717601847718495e-05, "loss": 0.0034, "step": 8732 }, { "epoch": 0.17, "learning_rate": 4.717537151692772e-05, "loss": 0.0036, "step": 8734 }, { "epoch": 0.17, "learning_rate": 4.7174724556670486e-05, "loss": 0.0318, "step": 8736 }, { "epoch": 0.17, "learning_rate": 4.717407759641325e-05, "loss": 0.0092, "step": 8738 }, { "epoch": 0.17, "learning_rate": 4.7173430636156025e-05, "loss": 0.0029, "step": 8740 }, { "epoch": 0.17, "learning_rate": 4.7172783675898794e-05, "loss": 0.008, "step": 8742 }, { "epoch": 0.17, "learning_rate": 4.717213671564156e-05, "loss": 0.0319, "step": 8744 }, { "epoch": 0.17, "learning_rate": 4.717148975538433e-05, "loss": 0.0087, "step": 8746 }, { "epoch": 0.17, "learning_rate": 4.7170842795127094e-05, "loss": 0.0122, "step": 8748 }, { "epoch": 0.17, "learning_rate": 4.717019583486987e-05, "loss": 0.0033, "step": 8750 }, { "epoch": 0.17, "learning_rate": 4.716954887461263e-05, "loss": 0.0098, "step": 8752 }, { "epoch": 0.17, "learning_rate": 4.71689019143554e-05, "loss": 0.0116, "step": 8754 }, { "epoch": 0.17, "learning_rate": 4.716825495409817e-05, "loss": 0.0239, "step": 8756 }, { "epoch": 0.17, "learning_rate": 4.716760799384094e-05, "loss": 0.0094, "step": 8758 }, { "epoch": 0.17, "learning_rate": 4.716696103358371e-05, "loss": 0.0243, "step": 8760 }, { "epoch": 0.17, "learning_rate": 4.716631407332648e-05, "loss": 0.0412, "step": 8762 }, { "epoch": 0.17, "learning_rate": 4.716566711306925e-05, "loss": 0.0078, "step": 8764 }, { "epoch": 0.17, "learning_rate": 4.7165020152812016e-05, "loss": 0.012, "step": 8766 }, { "epoch": 0.17, "learning_rate": 4.7164373192554785e-05, "loss": 0.0177, "step": 8768 }, { "epoch": 0.17, "learning_rate": 4.716372623229755e-05, "loss": 0.0275, "step": 8770 }, { "epoch": 0.17, "learning_rate": 4.716307927204032e-05, "loss": 0.0172, "step": 8772 }, { "epoch": 0.17, "learning_rate": 4.716243231178309e-05, "loss": 0.003, "step": 8774 }, { "epoch": 0.17, "learning_rate": 4.7161785351525855e-05, "loss": 0.0048, "step": 8776 }, { "epoch": 0.17, "learning_rate": 4.716113839126863e-05, "loss": 0.007, "step": 8778 }, { "epoch": 0.17, "learning_rate": 4.716049143101139e-05, "loss": 0.0142, "step": 8780 }, { "epoch": 0.17, "learning_rate": 4.715984447075417e-05, "loss": 0.026, "step": 8782 }, { "epoch": 0.17, "learning_rate": 4.715919751049693e-05, "loss": 0.0066, "step": 8784 }, { "epoch": 0.17, "learning_rate": 4.71585505502397e-05, "loss": 0.0281, "step": 8786 }, { "epoch": 0.17, "learning_rate": 4.715790358998247e-05, "loss": 0.0034, "step": 8788 }, { "epoch": 0.17, "learning_rate": 4.715725662972524e-05, "loss": 0.0026, "step": 8790 }, { "epoch": 0.17, "learning_rate": 4.715660966946801e-05, "loss": 0.0168, "step": 8792 }, { "epoch": 0.17, "learning_rate": 4.7155962709210776e-05, "loss": 0.0086, "step": 8794 }, { "epoch": 0.17, "learning_rate": 4.7155315748953545e-05, "loss": 0.0266, "step": 8796 }, { "epoch": 0.17, "learning_rate": 4.715466878869631e-05, "loss": 0.0167, "step": 8798 }, { "epoch": 0.17, "learning_rate": 4.7154021828439084e-05, "loss": 0.0233, "step": 8800 }, { "epoch": 0.17, "learning_rate": 4.7153374868181846e-05, "loss": 0.0226, "step": 8802 }, { "epoch": 0.17, "learning_rate": 4.715272790792462e-05, "loss": 0.0029, "step": 8804 }, { "epoch": 0.17, "learning_rate": 4.715208094766739e-05, "loss": 0.0141, "step": 8806 }, { "epoch": 0.17, "learning_rate": 4.715143398741015e-05, "loss": 0.005, "step": 8808 }, { "epoch": 0.17, "learning_rate": 4.715078702715293e-05, "loss": 0.0053, "step": 8810 }, { "epoch": 0.17, "learning_rate": 4.715014006689569e-05, "loss": 0.0081, "step": 8812 }, { "epoch": 0.17, "learning_rate": 4.714949310663846e-05, "loss": 0.0072, "step": 8814 }, { "epoch": 0.17, "learning_rate": 4.714884614638123e-05, "loss": 0.0228, "step": 8816 }, { "epoch": 0.17, "learning_rate": 4.7148199186124e-05, "loss": 0.0083, "step": 8818 }, { "epoch": 0.17, "learning_rate": 4.714755222586677e-05, "loss": 0.0231, "step": 8820 }, { "epoch": 0.17, "learning_rate": 4.714690526560954e-05, "loss": 0.011, "step": 8822 }, { "epoch": 0.17, "learning_rate": 4.7146258305352306e-05, "loss": 0.005, "step": 8824 }, { "epoch": 0.17, "learning_rate": 4.7145611345095075e-05, "loss": 0.0211, "step": 8826 }, { "epoch": 0.17, "learning_rate": 4.7144964384837844e-05, "loss": 0.0399, "step": 8828 }, { "epoch": 0.17, "learning_rate": 4.7144317424580606e-05, "loss": 0.0322, "step": 8830 }, { "epoch": 0.17, "learning_rate": 4.714367046432338e-05, "loss": 0.0508, "step": 8832 }, { "epoch": 0.17, "learning_rate": 4.7143023504066144e-05, "loss": 0.0081, "step": 8834 }, { "epoch": 0.17, "learning_rate": 4.7142376543808914e-05, "loss": 0.0066, "step": 8836 }, { "epoch": 0.17, "learning_rate": 4.714172958355168e-05, "loss": 0.0167, "step": 8838 }, { "epoch": 0.17, "learning_rate": 4.714108262329445e-05, "loss": 0.0589, "step": 8840 }, { "epoch": 0.17, "learning_rate": 4.714043566303723e-05, "loss": 0.0141, "step": 8842 }, { "epoch": 0.17, "learning_rate": 4.713978870277999e-05, "loss": 0.0452, "step": 8844 }, { "epoch": 0.17, "learning_rate": 4.713914174252276e-05, "loss": 0.0211, "step": 8846 }, { "epoch": 0.17, "learning_rate": 4.713849478226553e-05, "loss": 0.0028, "step": 8848 }, { "epoch": 0.17, "learning_rate": 4.71378478220083e-05, "loss": 0.0047, "step": 8850 }, { "epoch": 0.17, "learning_rate": 4.7137200861751066e-05, "loss": 0.0173, "step": 8852 }, { "epoch": 0.17, "learning_rate": 4.7136553901493835e-05, "loss": 0.0142, "step": 8854 }, { "epoch": 0.17, "learning_rate": 4.7135906941236604e-05, "loss": 0.0124, "step": 8856 }, { "epoch": 0.17, "learning_rate": 4.713525998097937e-05, "loss": 0.0206, "step": 8858 }, { "epoch": 0.17, "learning_rate": 4.713461302072214e-05, "loss": 0.0243, "step": 8860 }, { "epoch": 0.17, "learning_rate": 4.7133966060464905e-05, "loss": 0.028, "step": 8862 }, { "epoch": 0.17, "learning_rate": 4.713331910020768e-05, "loss": 0.0349, "step": 8864 }, { "epoch": 0.17, "learning_rate": 4.713267213995044e-05, "loss": 0.0056, "step": 8866 }, { "epoch": 0.17, "learning_rate": 4.713202517969321e-05, "loss": 0.0284, "step": 8868 }, { "epoch": 0.17, "learning_rate": 4.713137821943598e-05, "loss": 0.0102, "step": 8870 }, { "epoch": 0.17, "learning_rate": 4.713073125917875e-05, "loss": 0.0074, "step": 8872 }, { "epoch": 0.17, "learning_rate": 4.713008429892152e-05, "loss": 0.006, "step": 8874 }, { "epoch": 0.17, "learning_rate": 4.712943733866429e-05, "loss": 0.0182, "step": 8876 }, { "epoch": 0.17, "learning_rate": 4.712879037840706e-05, "loss": 0.023, "step": 8878 }, { "epoch": 0.17, "learning_rate": 4.712814341814982e-05, "loss": 0.0254, "step": 8880 }, { "epoch": 0.17, "learning_rate": 4.7127496457892596e-05, "loss": 0.029, "step": 8882 }, { "epoch": 0.17, "learning_rate": 4.712684949763536e-05, "loss": 0.0157, "step": 8884 }, { "epoch": 0.17, "learning_rate": 4.7126202537378134e-05, "loss": 0.0222, "step": 8886 }, { "epoch": 0.17, "learning_rate": 4.71255555771209e-05, "loss": 0.0049, "step": 8888 }, { "epoch": 0.17, "learning_rate": 4.7124908616863665e-05, "loss": 0.0148, "step": 8890 }, { "epoch": 0.17, "learning_rate": 4.712426165660644e-05, "loss": 0.0348, "step": 8892 }, { "epoch": 0.17, "learning_rate": 4.7123614696349203e-05, "loss": 0.0035, "step": 8894 }, { "epoch": 0.17, "learning_rate": 4.712296773609197e-05, "loss": 0.0112, "step": 8896 }, { "epoch": 0.17, "learning_rate": 4.712232077583474e-05, "loss": 0.0039, "step": 8898 }, { "epoch": 0.17, "learning_rate": 4.712167381557751e-05, "loss": 0.0252, "step": 8900 }, { "epoch": 0.17, "learning_rate": 4.712102685532028e-05, "loss": 0.011, "step": 8902 }, { "epoch": 0.17, "learning_rate": 4.712037989506305e-05, "loss": 0.0151, "step": 8904 }, { "epoch": 0.17, "learning_rate": 4.711973293480582e-05, "loss": 0.0096, "step": 8906 }, { "epoch": 0.17, "learning_rate": 4.711908597454859e-05, "loss": 0.0506, "step": 8908 }, { "epoch": 0.17, "learning_rate": 4.7118439014291356e-05, "loss": 0.0474, "step": 8910 }, { "epoch": 0.17, "learning_rate": 4.711779205403412e-05, "loss": 0.0055, "step": 8912 }, { "epoch": 0.17, "learning_rate": 4.7117145093776894e-05, "loss": 0.0198, "step": 8914 }, { "epoch": 0.17, "learning_rate": 4.7116498133519657e-05, "loss": 0.0237, "step": 8916 }, { "epoch": 0.17, "learning_rate": 4.7115851173262426e-05, "loss": 0.0177, "step": 8918 }, { "epoch": 0.17, "learning_rate": 4.71152042130052e-05, "loss": 0.0095, "step": 8920 }, { "epoch": 0.17, "learning_rate": 4.7114557252747964e-05, "loss": 0.0339, "step": 8922 }, { "epoch": 0.17, "learning_rate": 4.711391029249074e-05, "loss": 0.0268, "step": 8924 }, { "epoch": 0.17, "learning_rate": 4.71132633322335e-05, "loss": 0.0216, "step": 8926 }, { "epoch": 0.17, "learning_rate": 4.711261637197627e-05, "loss": 0.0151, "step": 8928 }, { "epoch": 0.17, "learning_rate": 4.711196941171904e-05, "loss": 0.0228, "step": 8930 }, { "epoch": 0.17, "learning_rate": 4.711132245146181e-05, "loss": 0.0056, "step": 8932 }, { "epoch": 0.17, "learning_rate": 4.711067549120458e-05, "loss": 0.0168, "step": 8934 }, { "epoch": 0.17, "learning_rate": 4.711002853094735e-05, "loss": 0.0045, "step": 8936 }, { "epoch": 0.17, "learning_rate": 4.7109381570690117e-05, "loss": 0.0219, "step": 8938 }, { "epoch": 0.17, "learning_rate": 4.710873461043288e-05, "loss": 0.0138, "step": 8940 }, { "epoch": 0.17, "learning_rate": 4.7108087650175655e-05, "loss": 0.0238, "step": 8942 }, { "epoch": 0.17, "learning_rate": 4.710744068991842e-05, "loss": 0.0132, "step": 8944 }, { "epoch": 0.17, "learning_rate": 4.710679372966119e-05, "loss": 0.0575, "step": 8946 }, { "epoch": 0.17, "learning_rate": 4.7106146769403955e-05, "loss": 0.0121, "step": 8948 }, { "epoch": 0.17, "learning_rate": 4.7105499809146724e-05, "loss": 0.0531, "step": 8950 }, { "epoch": 0.17, "learning_rate": 4.71048528488895e-05, "loss": 0.0008, "step": 8952 }, { "epoch": 0.17, "learning_rate": 4.710420588863226e-05, "loss": 0.0172, "step": 8954 }, { "epoch": 0.17, "learning_rate": 4.710355892837503e-05, "loss": 0.017, "step": 8956 }, { "epoch": 0.17, "learning_rate": 4.71029119681178e-05, "loss": 0.0086, "step": 8958 }, { "epoch": 0.17, "learning_rate": 4.710226500786057e-05, "loss": 0.0189, "step": 8960 }, { "epoch": 0.17, "learning_rate": 4.710161804760333e-05, "loss": 0.0246, "step": 8962 }, { "epoch": 0.17, "learning_rate": 4.710097108734611e-05, "loss": 0.0021, "step": 8964 }, { "epoch": 0.17, "learning_rate": 4.710032412708888e-05, "loss": 0.0172, "step": 8966 }, { "epoch": 0.17, "learning_rate": 4.7099677166831646e-05, "loss": 0.0065, "step": 8968 }, { "epoch": 0.17, "learning_rate": 4.7099030206574415e-05, "loss": 0.005, "step": 8970 }, { "epoch": 0.17, "learning_rate": 4.709838324631718e-05, "loss": 0.0246, "step": 8972 }, { "epoch": 0.17, "learning_rate": 4.709773628605995e-05, "loss": 0.0147, "step": 8974 }, { "epoch": 0.17, "learning_rate": 4.7097089325802716e-05, "loss": 0.0092, "step": 8976 }, { "epoch": 0.17, "learning_rate": 4.7096442365545485e-05, "loss": 0.0223, "step": 8978 }, { "epoch": 0.17, "learning_rate": 4.7095795405288254e-05, "loss": 0.0086, "step": 8980 }, { "epoch": 0.17, "learning_rate": 4.709514844503102e-05, "loss": 0.0141, "step": 8982 }, { "epoch": 0.17, "learning_rate": 4.70945014847738e-05, "loss": 0.0145, "step": 8984 }, { "epoch": 0.17, "learning_rate": 4.709385452451656e-05, "loss": 0.002, "step": 8986 }, { "epoch": 0.17, "learning_rate": 4.709320756425933e-05, "loss": 0.0068, "step": 8988 }, { "epoch": 0.17, "learning_rate": 4.70925606040021e-05, "loss": 0.0303, "step": 8990 }, { "epoch": 0.17, "learning_rate": 4.709191364374487e-05, "loss": 0.0245, "step": 8992 }, { "epoch": 0.17, "learning_rate": 4.709126668348763e-05, "loss": 0.0155, "step": 8994 }, { "epoch": 0.17, "learning_rate": 4.7090619723230406e-05, "loss": 0.0034, "step": 8996 }, { "epoch": 0.17, "learning_rate": 4.7089972762973175e-05, "loss": 0.0381, "step": 8998 }, { "epoch": 0.17, "learning_rate": 4.708932580271594e-05, "loss": 0.0016, "step": 9000 }, { "epoch": 0.17, "learning_rate": 4.7088678842458714e-05, "loss": 0.0083, "step": 9002 }, { "epoch": 0.17, "learning_rate": 4.7088031882201476e-05, "loss": 0.0027, "step": 9004 }, { "epoch": 0.17, "learning_rate": 4.708738492194425e-05, "loss": 0.0064, "step": 9006 }, { "epoch": 0.17, "learning_rate": 4.7086737961687014e-05, "loss": 0.0261, "step": 9008 }, { "epoch": 0.17, "learning_rate": 4.708609100142978e-05, "loss": 0.0132, "step": 9010 }, { "epoch": 0.17, "learning_rate": 4.708544404117255e-05, "loss": 0.024, "step": 9012 }, { "epoch": 0.17, "learning_rate": 4.708479708091532e-05, "loss": 0.0058, "step": 9014 }, { "epoch": 0.17, "learning_rate": 4.708415012065809e-05, "loss": 0.0034, "step": 9016 }, { "epoch": 0.18, "learning_rate": 4.708350316040086e-05, "loss": 0.0149, "step": 9018 }, { "epoch": 0.18, "learning_rate": 4.708285620014363e-05, "loss": 0.0108, "step": 9020 }, { "epoch": 0.18, "learning_rate": 4.708220923988639e-05, "loss": 0.0273, "step": 9022 }, { "epoch": 0.18, "learning_rate": 4.708156227962917e-05, "loss": 0.0124, "step": 9024 }, { "epoch": 0.18, "learning_rate": 4.708091531937193e-05, "loss": 0.0025, "step": 9026 }, { "epoch": 0.18, "learning_rate": 4.7080268359114705e-05, "loss": 0.016, "step": 9028 }, { "epoch": 0.18, "learning_rate": 4.7079621398857474e-05, "loss": 0.0184, "step": 9030 }, { "epoch": 0.18, "learning_rate": 4.7078974438600236e-05, "loss": 0.0022, "step": 9032 }, { "epoch": 0.18, "learning_rate": 4.707832747834301e-05, "loss": 0.0076, "step": 9034 }, { "epoch": 0.18, "learning_rate": 4.7077680518085775e-05, "loss": 0.0051, "step": 9036 }, { "epoch": 0.18, "learning_rate": 4.7077033557828544e-05, "loss": 0.0172, "step": 9038 }, { "epoch": 0.18, "learning_rate": 4.707638659757131e-05, "loss": 0.0067, "step": 9040 }, { "epoch": 0.18, "learning_rate": 4.707573963731408e-05, "loss": 0.0113, "step": 9042 }, { "epoch": 0.18, "learning_rate": 4.707509267705685e-05, "loss": 0.0015, "step": 9044 }, { "epoch": 0.18, "learning_rate": 4.707444571679962e-05, "loss": 0.0543, "step": 9046 }, { "epoch": 0.18, "learning_rate": 4.707379875654239e-05, "loss": 0.0065, "step": 9048 }, { "epoch": 0.18, "learning_rate": 4.707315179628516e-05, "loss": 0.0031, "step": 9050 }, { "epoch": 0.18, "learning_rate": 4.707250483602793e-05, "loss": 0.0026, "step": 9052 }, { "epoch": 0.18, "learning_rate": 4.707185787577069e-05, "loss": 0.0191, "step": 9054 }, { "epoch": 0.18, "learning_rate": 4.7071210915513465e-05, "loss": 0.0081, "step": 9056 }, { "epoch": 0.18, "learning_rate": 4.707056395525623e-05, "loss": 0.0044, "step": 9058 }, { "epoch": 0.18, "learning_rate": 4.7069916994999e-05, "loss": 0.0165, "step": 9060 }, { "epoch": 0.18, "learning_rate": 4.7069270034741766e-05, "loss": 0.0232, "step": 9062 }, { "epoch": 0.18, "learning_rate": 4.7068623074484535e-05, "loss": 0.0049, "step": 9064 }, { "epoch": 0.18, "learning_rate": 4.706797611422731e-05, "loss": 0.0081, "step": 9066 }, { "epoch": 0.18, "learning_rate": 4.706732915397007e-05, "loss": 0.0155, "step": 9068 }, { "epoch": 0.18, "learning_rate": 4.706668219371284e-05, "loss": 0.0067, "step": 9070 }, { "epoch": 0.18, "learning_rate": 4.706603523345561e-05, "loss": 0.0031, "step": 9072 }, { "epoch": 0.18, "learning_rate": 4.706538827319838e-05, "loss": 0.0052, "step": 9074 }, { "epoch": 0.18, "learning_rate": 4.706474131294115e-05, "loss": 0.0091, "step": 9076 }, { "epoch": 0.18, "learning_rate": 4.706409435268392e-05, "loss": 0.0117, "step": 9078 }, { "epoch": 0.18, "learning_rate": 4.706344739242669e-05, "loss": 0.0141, "step": 9080 }, { "epoch": 0.18, "learning_rate": 4.706280043216945e-05, "loss": 0.043, "step": 9082 }, { "epoch": 0.18, "learning_rate": 4.7062153471912226e-05, "loss": 0.0069, "step": 9084 }, { "epoch": 0.18, "learning_rate": 4.706150651165499e-05, "loss": 0.0142, "step": 9086 }, { "epoch": 0.18, "learning_rate": 4.7060859551397764e-05, "loss": 0.0196, "step": 9088 }, { "epoch": 0.18, "learning_rate": 4.7060212591140526e-05, "loss": 0.1341, "step": 9090 }, { "epoch": 0.18, "learning_rate": 4.7059565630883295e-05, "loss": 0.0066, "step": 9092 }, { "epoch": 0.18, "learning_rate": 4.7058918670626064e-05, "loss": 0.0098, "step": 9094 }, { "epoch": 0.18, "learning_rate": 4.7058271710368834e-05, "loss": 0.0059, "step": 9096 }, { "epoch": 0.18, "learning_rate": 4.70576247501116e-05, "loss": 0.0096, "step": 9098 }, { "epoch": 0.18, "learning_rate": 4.705697778985437e-05, "loss": 0.0068, "step": 9100 }, { "epoch": 0.18, "learning_rate": 4.705633082959714e-05, "loss": 0.0147, "step": 9102 }, { "epoch": 0.18, "learning_rate": 4.70556838693399e-05, "loss": 0.0121, "step": 9104 }, { "epoch": 0.18, "learning_rate": 4.705503690908268e-05, "loss": 0.0074, "step": 9106 }, { "epoch": 0.18, "learning_rate": 4.705438994882545e-05, "loss": 0.0191, "step": 9108 }, { "epoch": 0.18, "learning_rate": 4.705374298856822e-05, "loss": 0.0063, "step": 9110 }, { "epoch": 0.18, "learning_rate": 4.7053096028310986e-05, "loss": 0.0121, "step": 9112 }, { "epoch": 0.18, "learning_rate": 4.705244906805375e-05, "loss": 0.0207, "step": 9114 }, { "epoch": 0.18, "learning_rate": 4.7051802107796524e-05, "loss": 0.0077, "step": 9116 }, { "epoch": 0.18, "learning_rate": 4.705115514753929e-05, "loss": 0.0054, "step": 9118 }, { "epoch": 0.18, "learning_rate": 4.7050508187282056e-05, "loss": 0.0086, "step": 9120 }, { "epoch": 0.18, "learning_rate": 4.7049861227024825e-05, "loss": 0.0106, "step": 9122 }, { "epoch": 0.18, "learning_rate": 4.7049214266767594e-05, "loss": 0.0146, "step": 9124 }, { "epoch": 0.18, "learning_rate": 4.704856730651036e-05, "loss": 0.0092, "step": 9126 }, { "epoch": 0.18, "learning_rate": 4.704792034625313e-05, "loss": 0.0036, "step": 9128 }, { "epoch": 0.18, "learning_rate": 4.70472733859959e-05, "loss": 0.0706, "step": 9130 }, { "epoch": 0.18, "learning_rate": 4.704662642573867e-05, "loss": 0.0142, "step": 9132 }, { "epoch": 0.18, "learning_rate": 4.704597946548144e-05, "loss": 0.0086, "step": 9134 }, { "epoch": 0.18, "learning_rate": 4.70453325052242e-05, "loss": 0.0179, "step": 9136 }, { "epoch": 0.18, "learning_rate": 4.704468554496698e-05, "loss": 0.0103, "step": 9138 }, { "epoch": 0.18, "learning_rate": 4.704403858470974e-05, "loss": 0.0027, "step": 9140 }, { "epoch": 0.18, "learning_rate": 4.704339162445251e-05, "loss": 0.0024, "step": 9142 }, { "epoch": 0.18, "learning_rate": 4.7042744664195285e-05, "loss": 0.0048, "step": 9144 }, { "epoch": 0.18, "learning_rate": 4.704209770393805e-05, "loss": 0.0247, "step": 9146 }, { "epoch": 0.18, "learning_rate": 4.704145074368082e-05, "loss": 0.015, "step": 9148 }, { "epoch": 0.18, "learning_rate": 4.7040803783423585e-05, "loss": 0.0179, "step": 9150 }, { "epoch": 0.18, "learning_rate": 4.7040156823166354e-05, "loss": 0.015, "step": 9152 }, { "epoch": 0.18, "learning_rate": 4.7039509862909123e-05, "loss": 0.0124, "step": 9154 }, { "epoch": 0.18, "learning_rate": 4.703886290265189e-05, "loss": 0.0119, "step": 9156 }, { "epoch": 0.18, "learning_rate": 4.703821594239466e-05, "loss": 0.0243, "step": 9158 }, { "epoch": 0.18, "learning_rate": 4.703756898213743e-05, "loss": 0.0128, "step": 9160 }, { "epoch": 0.18, "learning_rate": 4.70369220218802e-05, "loss": 0.018, "step": 9162 }, { "epoch": 0.18, "learning_rate": 4.703627506162296e-05, "loss": 0.0104, "step": 9164 }, { "epoch": 0.18, "learning_rate": 4.703562810136574e-05, "loss": 0.0265, "step": 9166 }, { "epoch": 0.18, "learning_rate": 4.70349811411085e-05, "loss": 0.0058, "step": 9168 }, { "epoch": 0.18, "learning_rate": 4.7034334180851276e-05, "loss": 0.0088, "step": 9170 }, { "epoch": 0.18, "learning_rate": 4.703368722059404e-05, "loss": 0.0051, "step": 9172 }, { "epoch": 0.18, "learning_rate": 4.703304026033681e-05, "loss": 0.0257, "step": 9174 }, { "epoch": 0.18, "learning_rate": 4.703239330007958e-05, "loss": 0.0051, "step": 9176 }, { "epoch": 0.18, "learning_rate": 4.7031746339822346e-05, "loss": 0.0213, "step": 9178 }, { "epoch": 0.18, "learning_rate": 4.7031099379565115e-05, "loss": 0.0898, "step": 9180 }, { "epoch": 0.18, "learning_rate": 4.7030452419307884e-05, "loss": 0.0168, "step": 9182 }, { "epoch": 0.18, "learning_rate": 4.702980545905065e-05, "loss": 0.0056, "step": 9184 }, { "epoch": 0.18, "learning_rate": 4.702915849879342e-05, "loss": 0.0016, "step": 9186 }, { "epoch": 0.18, "learning_rate": 4.702851153853619e-05, "loss": 0.0359, "step": 9188 }, { "epoch": 0.18, "learning_rate": 4.702786457827896e-05, "loss": 0.0104, "step": 9190 }, { "epoch": 0.18, "learning_rate": 4.702721761802173e-05, "loss": 0.0049, "step": 9192 }, { "epoch": 0.18, "learning_rate": 4.70265706577645e-05, "loss": 0.029, "step": 9194 }, { "epoch": 0.18, "learning_rate": 4.702592369750726e-05, "loss": 0.0085, "step": 9196 }, { "epoch": 0.18, "learning_rate": 4.7025276737250036e-05, "loss": 0.0189, "step": 9198 }, { "epoch": 0.18, "learning_rate": 4.70246297769928e-05, "loss": 0.0116, "step": 9200 }, { "epoch": 0.18, "learning_rate": 4.702398281673557e-05, "loss": 0.0214, "step": 9202 }, { "epoch": 0.18, "learning_rate": 4.702333585647834e-05, "loss": 0.0272, "step": 9204 }, { "epoch": 0.18, "learning_rate": 4.7022688896221106e-05, "loss": 0.0066, "step": 9206 }, { "epoch": 0.18, "learning_rate": 4.702204193596388e-05, "loss": 0.0105, "step": 9208 }, { "epoch": 0.18, "learning_rate": 4.7021394975706644e-05, "loss": 0.0263, "step": 9210 }, { "epoch": 0.18, "learning_rate": 4.702074801544941e-05, "loss": 0.0073, "step": 9212 }, { "epoch": 0.18, "learning_rate": 4.702010105519218e-05, "loss": 0.014, "step": 9214 }, { "epoch": 0.18, "learning_rate": 4.701945409493495e-05, "loss": 0.0082, "step": 9216 }, { "epoch": 0.18, "learning_rate": 4.7018807134677714e-05, "loss": 0.0187, "step": 9218 }, { "epoch": 0.18, "learning_rate": 4.701816017442049e-05, "loss": 0.005, "step": 9220 }, { "epoch": 0.18, "learning_rate": 4.701751321416326e-05, "loss": 0.0363, "step": 9222 }, { "epoch": 0.18, "learning_rate": 4.701686625390602e-05, "loss": 0.0277, "step": 9224 }, { "epoch": 0.18, "learning_rate": 4.70162192936488e-05, "loss": 0.0068, "step": 9226 }, { "epoch": 0.18, "learning_rate": 4.701557233339156e-05, "loss": 0.0129, "step": 9228 }, { "epoch": 0.18, "learning_rate": 4.7014925373134335e-05, "loss": 0.0093, "step": 9230 }, { "epoch": 0.18, "learning_rate": 4.70142784128771e-05, "loss": 0.0054, "step": 9232 }, { "epoch": 0.18, "learning_rate": 4.7013631452619866e-05, "loss": 0.0204, "step": 9234 }, { "epoch": 0.18, "learning_rate": 4.7012984492362636e-05, "loss": 0.0143, "step": 9236 }, { "epoch": 0.18, "learning_rate": 4.7012337532105405e-05, "loss": 0.0172, "step": 9238 }, { "epoch": 0.18, "learning_rate": 4.7011690571848174e-05, "loss": 0.0077, "step": 9240 }, { "epoch": 0.18, "learning_rate": 4.701104361159094e-05, "loss": 0.0131, "step": 9242 }, { "epoch": 0.18, "learning_rate": 4.701039665133371e-05, "loss": 0.0384, "step": 9244 }, { "epoch": 0.18, "learning_rate": 4.7009749691076474e-05, "loss": 0.0104, "step": 9246 }, { "epoch": 0.18, "learning_rate": 4.700910273081925e-05, "loss": 0.0232, "step": 9248 }, { "epoch": 0.18, "learning_rate": 4.700845577056201e-05, "loss": 0.0201, "step": 9250 }, { "epoch": 0.18, "learning_rate": 4.700780881030479e-05, "loss": 0.0069, "step": 9252 }, { "epoch": 0.18, "learning_rate": 4.700716185004756e-05, "loss": 0.0053, "step": 9254 }, { "epoch": 0.18, "learning_rate": 4.700651488979032e-05, "loss": 0.0027, "step": 9256 }, { "epoch": 0.18, "learning_rate": 4.7005867929533095e-05, "loss": 0.0018, "step": 9258 }, { "epoch": 0.18, "learning_rate": 4.700522096927586e-05, "loss": 0.0058, "step": 9260 }, { "epoch": 0.18, "learning_rate": 4.700457400901863e-05, "loss": 0.0038, "step": 9262 }, { "epoch": 0.18, "learning_rate": 4.7003927048761396e-05, "loss": 0.005, "step": 9264 }, { "epoch": 0.18, "learning_rate": 4.7003280088504165e-05, "loss": 0.0098, "step": 9266 }, { "epoch": 0.18, "learning_rate": 4.7002633128246934e-05, "loss": 0.0027, "step": 9268 }, { "epoch": 0.18, "learning_rate": 4.70019861679897e-05, "loss": 0.0227, "step": 9270 }, { "epoch": 0.18, "learning_rate": 4.700133920773247e-05, "loss": 0.0396, "step": 9272 }, { "epoch": 0.18, "learning_rate": 4.700069224747524e-05, "loss": 0.0067, "step": 9274 }, { "epoch": 0.18, "learning_rate": 4.700004528721801e-05, "loss": 0.0167, "step": 9276 }, { "epoch": 0.18, "learning_rate": 4.699939832696077e-05, "loss": 0.0297, "step": 9278 }, { "epoch": 0.18, "learning_rate": 4.699875136670355e-05, "loss": 0.0291, "step": 9280 }, { "epoch": 0.18, "learning_rate": 4.699810440644631e-05, "loss": 0.0032, "step": 9282 }, { "epoch": 0.18, "learning_rate": 4.699745744618908e-05, "loss": 0.0123, "step": 9284 }, { "epoch": 0.18, "learning_rate": 4.6996810485931856e-05, "loss": 0.0169, "step": 9286 }, { "epoch": 0.18, "learning_rate": 4.699616352567462e-05, "loss": 0.0139, "step": 9288 }, { "epoch": 0.18, "learning_rate": 4.6995516565417394e-05, "loss": 0.0047, "step": 9290 }, { "epoch": 0.18, "learning_rate": 4.6994869605160156e-05, "loss": 0.0091, "step": 9292 }, { "epoch": 0.18, "learning_rate": 4.6994222644902925e-05, "loss": 0.0112, "step": 9294 }, { "epoch": 0.18, "learning_rate": 4.6993575684645695e-05, "loss": 0.0192, "step": 9296 }, { "epoch": 0.18, "learning_rate": 4.6992928724388464e-05, "loss": 0.0062, "step": 9298 }, { "epoch": 0.18, "learning_rate": 4.699228176413123e-05, "loss": 0.012, "step": 9300 }, { "epoch": 0.18, "learning_rate": 4.6991634803874e-05, "loss": 0.0093, "step": 9302 }, { "epoch": 0.18, "learning_rate": 4.699098784361677e-05, "loss": 0.0093, "step": 9304 }, { "epoch": 0.18, "learning_rate": 4.699034088335953e-05, "loss": 0.0055, "step": 9306 }, { "epoch": 0.18, "learning_rate": 4.698969392310231e-05, "loss": 0.0153, "step": 9308 }, { "epoch": 0.18, "learning_rate": 4.698904696284507e-05, "loss": 0.0276, "step": 9310 }, { "epoch": 0.18, "learning_rate": 4.698840000258785e-05, "loss": 0.0311, "step": 9312 }, { "epoch": 0.18, "learning_rate": 4.698775304233061e-05, "loss": 0.0069, "step": 9314 }, { "epoch": 0.18, "learning_rate": 4.698710608207338e-05, "loss": 0.0407, "step": 9316 }, { "epoch": 0.18, "learning_rate": 4.698645912181615e-05, "loss": 0.0096, "step": 9318 }, { "epoch": 0.18, "learning_rate": 4.698581216155892e-05, "loss": 0.008, "step": 9320 }, { "epoch": 0.18, "learning_rate": 4.6985165201301686e-05, "loss": 0.029, "step": 9322 }, { "epoch": 0.18, "learning_rate": 4.6984518241044455e-05, "loss": 0.0084, "step": 9324 }, { "epoch": 0.18, "learning_rate": 4.6983871280787224e-05, "loss": 0.0302, "step": 9326 }, { "epoch": 0.18, "learning_rate": 4.6983224320529986e-05, "loss": 0.0146, "step": 9328 }, { "epoch": 0.18, "learning_rate": 4.698257736027276e-05, "loss": 0.0174, "step": 9330 }, { "epoch": 0.18, "learning_rate": 4.698193040001553e-05, "loss": 0.0051, "step": 9332 }, { "epoch": 0.18, "learning_rate": 4.69812834397583e-05, "loss": 0.0102, "step": 9334 }, { "epoch": 0.18, "learning_rate": 4.698063647950107e-05, "loss": 0.0207, "step": 9336 }, { "epoch": 0.18, "learning_rate": 4.697998951924383e-05, "loss": 0.0393, "step": 9338 }, { "epoch": 0.18, "learning_rate": 4.697934255898661e-05, "loss": 0.0256, "step": 9340 }, { "epoch": 0.18, "learning_rate": 4.697869559872937e-05, "loss": 0.006, "step": 9342 }, { "epoch": 0.18, "learning_rate": 4.697804863847214e-05, "loss": 0.0038, "step": 9344 }, { "epoch": 0.18, "learning_rate": 4.697740167821491e-05, "loss": 0.0097, "step": 9346 }, { "epoch": 0.18, "learning_rate": 4.697675471795768e-05, "loss": 0.0191, "step": 9348 }, { "epoch": 0.18, "learning_rate": 4.6976107757700446e-05, "loss": 0.0111, "step": 9350 }, { "epoch": 0.18, "learning_rate": 4.6975460797443215e-05, "loss": 0.0135, "step": 9352 }, { "epoch": 0.18, "learning_rate": 4.6974813837185984e-05, "loss": 0.0016, "step": 9354 }, { "epoch": 0.18, "learning_rate": 4.6974166876928753e-05, "loss": 0.0046, "step": 9356 }, { "epoch": 0.18, "learning_rate": 4.697351991667152e-05, "loss": 0.0119, "step": 9358 }, { "epoch": 0.18, "learning_rate": 4.6972872956414285e-05, "loss": 0.0017, "step": 9360 }, { "epoch": 0.18, "learning_rate": 4.697222599615706e-05, "loss": 0.0357, "step": 9362 }, { "epoch": 0.18, "learning_rate": 4.697157903589983e-05, "loss": 0.0093, "step": 9364 }, { "epoch": 0.18, "learning_rate": 4.697093207564259e-05, "loss": 0.0168, "step": 9366 }, { "epoch": 0.18, "learning_rate": 4.697028511538537e-05, "loss": 0.0082, "step": 9368 }, { "epoch": 0.18, "learning_rate": 4.696963815512813e-05, "loss": 0.0308, "step": 9370 }, { "epoch": 0.18, "learning_rate": 4.6968991194870906e-05, "loss": 0.0061, "step": 9372 }, { "epoch": 0.18, "learning_rate": 4.696834423461367e-05, "loss": 0.012, "step": 9374 }, { "epoch": 0.18, "learning_rate": 4.696769727435644e-05, "loss": 0.0327, "step": 9376 }, { "epoch": 0.18, "learning_rate": 4.696705031409921e-05, "loss": 0.0138, "step": 9378 }, { "epoch": 0.18, "learning_rate": 4.6966403353841976e-05, "loss": 0.0084, "step": 9380 }, { "epoch": 0.18, "learning_rate": 4.6965756393584745e-05, "loss": 0.0036, "step": 9382 }, { "epoch": 0.18, "learning_rate": 4.6965109433327514e-05, "loss": 0.0119, "step": 9384 }, { "epoch": 0.18, "learning_rate": 4.696446247307028e-05, "loss": 0.0038, "step": 9386 }, { "epoch": 0.18, "learning_rate": 4.6963815512813045e-05, "loss": 0.0123, "step": 9388 }, { "epoch": 0.18, "learning_rate": 4.696316855255582e-05, "loss": 0.0038, "step": 9390 }, { "epoch": 0.18, "learning_rate": 4.6962521592298583e-05, "loss": 0.0152, "step": 9392 }, { "epoch": 0.18, "learning_rate": 4.696187463204136e-05, "loss": 0.008, "step": 9394 }, { "epoch": 0.18, "learning_rate": 4.696122767178412e-05, "loss": 0.0099, "step": 9396 }, { "epoch": 0.18, "learning_rate": 4.696058071152689e-05, "loss": 0.0141, "step": 9398 }, { "epoch": 0.18, "learning_rate": 4.6959933751269667e-05, "loss": 0.0017, "step": 9400 }, { "epoch": 0.18, "learning_rate": 4.695928679101243e-05, "loss": 0.0147, "step": 9402 }, { "epoch": 0.18, "learning_rate": 4.69586398307552e-05, "loss": 0.0058, "step": 9404 }, { "epoch": 0.18, "learning_rate": 4.695799287049797e-05, "loss": 0.0058, "step": 9406 }, { "epoch": 0.18, "learning_rate": 4.6957345910240736e-05, "loss": 0.0084, "step": 9408 }, { "epoch": 0.18, "learning_rate": 4.6956698949983505e-05, "loss": 0.0091, "step": 9410 }, { "epoch": 0.18, "learning_rate": 4.6956051989726274e-05, "loss": 0.004, "step": 9412 }, { "epoch": 0.18, "learning_rate": 4.6955405029469043e-05, "loss": 0.0037, "step": 9414 }, { "epoch": 0.18, "learning_rate": 4.695475806921181e-05, "loss": 0.0208, "step": 9416 }, { "epoch": 0.18, "learning_rate": 4.695411110895458e-05, "loss": 0.0044, "step": 9418 }, { "epoch": 0.18, "learning_rate": 4.6953464148697344e-05, "loss": 0.0151, "step": 9420 }, { "epoch": 0.18, "learning_rate": 4.695281718844012e-05, "loss": 0.0028, "step": 9422 }, { "epoch": 0.18, "learning_rate": 4.695217022818288e-05, "loss": 0.0048, "step": 9424 }, { "epoch": 0.18, "learning_rate": 4.695152326792565e-05, "loss": 0.0091, "step": 9426 }, { "epoch": 0.18, "learning_rate": 4.695087630766842e-05, "loss": 0.0194, "step": 9428 }, { "epoch": 0.18, "learning_rate": 4.695022934741119e-05, "loss": 0.0102, "step": 9430 }, { "epoch": 0.18, "learning_rate": 4.6949582387153965e-05, "loss": 0.0105, "step": 9432 }, { "epoch": 0.18, "learning_rate": 4.694893542689673e-05, "loss": 0.0075, "step": 9434 }, { "epoch": 0.18, "learning_rate": 4.6948288466639497e-05, "loss": 0.0103, "step": 9436 }, { "epoch": 0.18, "learning_rate": 4.6947641506382266e-05, "loss": 0.0138, "step": 9438 }, { "epoch": 0.18, "learning_rate": 4.6946994546125035e-05, "loss": 0.0027, "step": 9440 }, { "epoch": 0.18, "learning_rate": 4.69463475858678e-05, "loss": 0.0226, "step": 9442 }, { "epoch": 0.18, "learning_rate": 4.694570062561057e-05, "loss": 0.002, "step": 9444 }, { "epoch": 0.18, "learning_rate": 4.694505366535334e-05, "loss": 0.005, "step": 9446 }, { "epoch": 0.18, "learning_rate": 4.6944406705096104e-05, "loss": 0.0031, "step": 9448 }, { "epoch": 0.18, "learning_rate": 4.694375974483888e-05, "loss": 0.0033, "step": 9450 }, { "epoch": 0.18, "learning_rate": 4.694311278458164e-05, "loss": 0.0135, "step": 9452 }, { "epoch": 0.18, "learning_rate": 4.694246582432442e-05, "loss": 0.0251, "step": 9454 }, { "epoch": 0.18, "learning_rate": 4.694181886406718e-05, "loss": 0.0052, "step": 9456 }, { "epoch": 0.18, "learning_rate": 4.694117190380995e-05, "loss": 0.003, "step": 9458 }, { "epoch": 0.18, "learning_rate": 4.694052494355272e-05, "loss": 0.0175, "step": 9460 }, { "epoch": 0.18, "learning_rate": 4.693987798329549e-05, "loss": 0.0038, "step": 9462 }, { "epoch": 0.18, "learning_rate": 4.693923102303826e-05, "loss": 0.0103, "step": 9464 }, { "epoch": 0.18, "learning_rate": 4.6938584062781026e-05, "loss": 0.0552, "step": 9466 }, { "epoch": 0.18, "learning_rate": 4.6937937102523795e-05, "loss": 0.0299, "step": 9468 }, { "epoch": 0.18, "learning_rate": 4.693729014226656e-05, "loss": 0.005, "step": 9470 }, { "epoch": 0.18, "learning_rate": 4.693664318200933e-05, "loss": 0.0163, "step": 9472 }, { "epoch": 0.18, "learning_rate": 4.6935996221752096e-05, "loss": 0.0103, "step": 9474 }, { "epoch": 0.18, "learning_rate": 4.693534926149487e-05, "loss": 0.0135, "step": 9476 }, { "epoch": 0.18, "learning_rate": 4.693470230123764e-05, "loss": 0.0075, "step": 9478 }, { "epoch": 0.18, "learning_rate": 4.69340553409804e-05, "loss": 0.0038, "step": 9480 }, { "epoch": 0.18, "learning_rate": 4.693340838072318e-05, "loss": 0.0142, "step": 9482 }, { "epoch": 0.18, "learning_rate": 4.693276142046594e-05, "loss": 0.0201, "step": 9484 }, { "epoch": 0.18, "learning_rate": 4.693211446020871e-05, "loss": 0.0014, "step": 9486 }, { "epoch": 0.18, "learning_rate": 4.693146749995148e-05, "loss": 0.0052, "step": 9488 }, { "epoch": 0.18, "learning_rate": 4.693082053969425e-05, "loss": 0.0084, "step": 9490 }, { "epoch": 0.18, "learning_rate": 4.693017357943702e-05, "loss": 0.0023, "step": 9492 }, { "epoch": 0.18, "learning_rate": 4.6929526619179786e-05, "loss": 0.0127, "step": 9494 }, { "epoch": 0.18, "learning_rate": 4.6928879658922556e-05, "loss": 0.006, "step": 9496 }, { "epoch": 0.18, "learning_rate": 4.6928232698665325e-05, "loss": 0.0026, "step": 9498 }, { "epoch": 0.18, "learning_rate": 4.6927585738408094e-05, "loss": 0.0198, "step": 9500 }, { "epoch": 0.18, "learning_rate": 4.6926938778150856e-05, "loss": 0.0083, "step": 9502 }, { "epoch": 0.18, "learning_rate": 4.692629181789363e-05, "loss": 0.0023, "step": 9504 }, { "epoch": 0.18, "learning_rate": 4.6925644857636394e-05, "loss": 0.0065, "step": 9506 }, { "epoch": 0.18, "learning_rate": 4.692499789737916e-05, "loss": 0.0116, "step": 9508 }, { "epoch": 0.18, "learning_rate": 4.692435093712194e-05, "loss": 0.0121, "step": 9510 }, { "epoch": 0.18, "learning_rate": 4.69237039768647e-05, "loss": 0.0092, "step": 9512 }, { "epoch": 0.18, "learning_rate": 4.692305701660748e-05, "loss": 0.0041, "step": 9514 }, { "epoch": 0.18, "learning_rate": 4.692241005635024e-05, "loss": 0.0241, "step": 9516 }, { "epoch": 0.18, "learning_rate": 4.692176309609301e-05, "loss": 0.0019, "step": 9518 }, { "epoch": 0.18, "learning_rate": 4.692111613583578e-05, "loss": 0.0142, "step": 9520 }, { "epoch": 0.18, "learning_rate": 4.692046917557855e-05, "loss": 0.0108, "step": 9522 }, { "epoch": 0.18, "learning_rate": 4.6919822215321316e-05, "loss": 0.0033, "step": 9524 }, { "epoch": 0.18, "learning_rate": 4.6919175255064085e-05, "loss": 0.0223, "step": 9526 }, { "epoch": 0.18, "learning_rate": 4.6918528294806854e-05, "loss": 0.0048, "step": 9528 }, { "epoch": 0.18, "learning_rate": 4.6917881334549616e-05, "loss": 0.0052, "step": 9530 }, { "epoch": 0.19, "learning_rate": 4.691723437429239e-05, "loss": 0.0098, "step": 9532 }, { "epoch": 0.19, "learning_rate": 4.6916587414035155e-05, "loss": 0.0568, "step": 9534 }, { "epoch": 0.19, "learning_rate": 4.691594045377793e-05, "loss": 0.0021, "step": 9536 }, { "epoch": 0.19, "learning_rate": 4.691529349352069e-05, "loss": 0.0041, "step": 9538 }, { "epoch": 0.19, "learning_rate": 4.691464653326346e-05, "loss": 0.0045, "step": 9540 }, { "epoch": 0.19, "learning_rate": 4.691399957300623e-05, "loss": 0.0231, "step": 9542 }, { "epoch": 0.19, "learning_rate": 4.6913352612749e-05, "loss": 0.0136, "step": 9544 }, { "epoch": 0.19, "learning_rate": 4.691270565249177e-05, "loss": 0.0036, "step": 9546 }, { "epoch": 0.19, "learning_rate": 4.691205869223454e-05, "loss": 0.0051, "step": 9548 }, { "epoch": 0.19, "learning_rate": 4.691141173197731e-05, "loss": 0.0194, "step": 9550 }, { "epoch": 0.19, "learning_rate": 4.6910764771720076e-05, "loss": 0.013, "step": 9552 }, { "epoch": 0.19, "learning_rate": 4.6910117811462845e-05, "loss": 0.0058, "step": 9554 }, { "epoch": 0.19, "learning_rate": 4.6909470851205614e-05, "loss": 0.0315, "step": 9556 }, { "epoch": 0.19, "learning_rate": 4.6908823890948384e-05, "loss": 0.0081, "step": 9558 }, { "epoch": 0.19, "learning_rate": 4.690817693069115e-05, "loss": 0.009, "step": 9560 }, { "epoch": 0.19, "learning_rate": 4.6907529970433915e-05, "loss": 0.006, "step": 9562 }, { "epoch": 0.19, "learning_rate": 4.690688301017669e-05, "loss": 0.0075, "step": 9564 }, { "epoch": 0.19, "learning_rate": 4.690623604991945e-05, "loss": 0.0162, "step": 9566 }, { "epoch": 0.19, "learning_rate": 4.690558908966222e-05, "loss": 0.0123, "step": 9568 }, { "epoch": 0.19, "learning_rate": 4.690494212940499e-05, "loss": 0.0083, "step": 9570 }, { "epoch": 0.19, "learning_rate": 4.690429516914776e-05, "loss": 0.0051, "step": 9572 }, { "epoch": 0.19, "learning_rate": 4.690364820889053e-05, "loss": 0.002, "step": 9574 }, { "epoch": 0.19, "learning_rate": 4.69030012486333e-05, "loss": 0.012, "step": 9576 }, { "epoch": 0.19, "learning_rate": 4.690235428837607e-05, "loss": 0.0038, "step": 9578 }, { "epoch": 0.19, "learning_rate": 4.690170732811884e-05, "loss": 0.0042, "step": 9580 }, { "epoch": 0.19, "learning_rate": 4.6901060367861606e-05, "loss": 0.0128, "step": 9582 }, { "epoch": 0.19, "learning_rate": 4.690041340760437e-05, "loss": 0.0081, "step": 9584 }, { "epoch": 0.19, "learning_rate": 4.6899766447347144e-05, "loss": 0.0169, "step": 9586 }, { "epoch": 0.19, "learning_rate": 4.689911948708991e-05, "loss": 0.0072, "step": 9588 }, { "epoch": 0.19, "learning_rate": 4.6898472526832675e-05, "loss": 0.011, "step": 9590 }, { "epoch": 0.19, "learning_rate": 4.689782556657545e-05, "loss": 0.0122, "step": 9592 }, { "epoch": 0.19, "learning_rate": 4.6897178606318214e-05, "loss": 0.0183, "step": 9594 }, { "epoch": 0.19, "learning_rate": 4.689653164606099e-05, "loss": 0.0185, "step": 9596 }, { "epoch": 0.19, "learning_rate": 4.689588468580375e-05, "loss": 0.0029, "step": 9598 }, { "epoch": 0.19, "learning_rate": 4.689523772554652e-05, "loss": 0.0025, "step": 9600 }, { "epoch": 0.19, "learning_rate": 4.689459076528929e-05, "loss": 0.0086, "step": 9602 }, { "epoch": 0.19, "learning_rate": 4.689394380503206e-05, "loss": 0.0128, "step": 9604 }, { "epoch": 0.19, "learning_rate": 4.689329684477483e-05, "loss": 0.0281, "step": 9606 }, { "epoch": 0.19, "learning_rate": 4.68926498845176e-05, "loss": 0.0031, "step": 9608 }, { "epoch": 0.19, "learning_rate": 4.6892002924260366e-05, "loss": 0.0014, "step": 9610 }, { "epoch": 0.19, "learning_rate": 4.689135596400313e-05, "loss": 0.0039, "step": 9612 }, { "epoch": 0.19, "learning_rate": 4.6890709003745904e-05, "loss": 0.006, "step": 9614 }, { "epoch": 0.19, "learning_rate": 4.689006204348867e-05, "loss": 0.0134, "step": 9616 }, { "epoch": 0.19, "learning_rate": 4.688941508323144e-05, "loss": 0.0019, "step": 9618 }, { "epoch": 0.19, "learning_rate": 4.6888768122974205e-05, "loss": 0.0105, "step": 9620 }, { "epoch": 0.19, "learning_rate": 4.6888121162716974e-05, "loss": 0.0092, "step": 9622 }, { "epoch": 0.19, "learning_rate": 4.688747420245975e-05, "loss": 0.0009, "step": 9624 }, { "epoch": 0.19, "learning_rate": 4.688682724220251e-05, "loss": 0.0015, "step": 9626 }, { "epoch": 0.19, "learning_rate": 4.688618028194528e-05, "loss": 0.0028, "step": 9628 }, { "epoch": 0.19, "learning_rate": 4.688553332168805e-05, "loss": 0.0171, "step": 9630 }, { "epoch": 0.19, "learning_rate": 4.688488636143082e-05, "loss": 0.0046, "step": 9632 }, { "epoch": 0.19, "learning_rate": 4.688423940117359e-05, "loss": 0.009, "step": 9634 }, { "epoch": 0.19, "learning_rate": 4.688359244091636e-05, "loss": 0.0146, "step": 9636 }, { "epoch": 0.19, "learning_rate": 4.688294548065913e-05, "loss": 0.0008, "step": 9638 }, { "epoch": 0.19, "learning_rate": 4.6882298520401896e-05, "loss": 0.0105, "step": 9640 }, { "epoch": 0.19, "learning_rate": 4.6881651560144665e-05, "loss": 0.0048, "step": 9642 }, { "epoch": 0.19, "learning_rate": 4.688100459988743e-05, "loss": 0.0107, "step": 9644 }, { "epoch": 0.19, "learning_rate": 4.68803576396302e-05, "loss": 0.0229, "step": 9646 }, { "epoch": 0.19, "learning_rate": 4.6879710679372965e-05, "loss": 0.0164, "step": 9648 }, { "epoch": 0.19, "learning_rate": 4.6879063719115734e-05, "loss": 0.0388, "step": 9650 }, { "epoch": 0.19, "learning_rate": 4.6878416758858503e-05, "loss": 0.0126, "step": 9652 }, { "epoch": 0.19, "learning_rate": 4.687776979860127e-05, "loss": 0.0262, "step": 9654 }, { "epoch": 0.19, "learning_rate": 4.687712283834405e-05, "loss": 0.0139, "step": 9656 }, { "epoch": 0.19, "learning_rate": 4.687647587808681e-05, "loss": 0.0202, "step": 9658 }, { "epoch": 0.19, "learning_rate": 4.687582891782958e-05, "loss": 0.0142, "step": 9660 }, { "epoch": 0.19, "learning_rate": 4.687518195757235e-05, "loss": 0.0047, "step": 9662 }, { "epoch": 0.19, "learning_rate": 4.687453499731512e-05, "loss": 0.0139, "step": 9664 }, { "epoch": 0.19, "learning_rate": 4.687388803705789e-05, "loss": 0.0288, "step": 9666 }, { "epoch": 0.19, "learning_rate": 4.6873241076800656e-05, "loss": 0.015, "step": 9668 }, { "epoch": 0.19, "learning_rate": 4.6872594116543425e-05, "loss": 0.0188, "step": 9670 }, { "epoch": 0.19, "learning_rate": 4.687194715628619e-05, "loss": 0.0345, "step": 9672 }, { "epoch": 0.19, "learning_rate": 4.687130019602896e-05, "loss": 0.0077, "step": 9674 }, { "epoch": 0.19, "learning_rate": 4.6870653235771726e-05, "loss": 0.0026, "step": 9676 }, { "epoch": 0.19, "learning_rate": 4.68700062755145e-05, "loss": 0.0102, "step": 9678 }, { "epoch": 0.19, "learning_rate": 4.6869359315257264e-05, "loss": 0.0306, "step": 9680 }, { "epoch": 0.19, "learning_rate": 4.686871235500003e-05, "loss": 0.0068, "step": 9682 }, { "epoch": 0.19, "learning_rate": 4.68680653947428e-05, "loss": 0.0061, "step": 9684 }, { "epoch": 0.19, "learning_rate": 4.686741843448557e-05, "loss": 0.0037, "step": 9686 }, { "epoch": 0.19, "learning_rate": 4.686677147422834e-05, "loss": 0.0033, "step": 9688 }, { "epoch": 0.19, "learning_rate": 4.686612451397111e-05, "loss": 0.0077, "step": 9690 }, { "epoch": 0.19, "learning_rate": 4.686547755371388e-05, "loss": 0.0211, "step": 9692 }, { "epoch": 0.19, "learning_rate": 4.686483059345665e-05, "loss": 0.0113, "step": 9694 }, { "epoch": 0.19, "learning_rate": 4.6864183633199417e-05, "loss": 0.0034, "step": 9696 }, { "epoch": 0.19, "learning_rate": 4.686353667294218e-05, "loss": 0.0034, "step": 9698 }, { "epoch": 0.19, "learning_rate": 4.6862889712684955e-05, "loss": 0.0035, "step": 9700 }, { "epoch": 0.19, "learning_rate": 4.6862242752427724e-05, "loss": 0.0083, "step": 9702 }, { "epoch": 0.19, "learning_rate": 4.6861595792170486e-05, "loss": 0.0034, "step": 9704 }, { "epoch": 0.19, "learning_rate": 4.686094883191326e-05, "loss": 0.0037, "step": 9706 }, { "epoch": 0.19, "learning_rate": 4.6860301871656024e-05, "loss": 0.0018, "step": 9708 }, { "epoch": 0.19, "learning_rate": 4.685965491139879e-05, "loss": 0.0106, "step": 9710 }, { "epoch": 0.19, "learning_rate": 4.685900795114156e-05, "loss": 0.0038, "step": 9712 }, { "epoch": 0.19, "learning_rate": 4.685836099088433e-05, "loss": 0.0214, "step": 9714 }, { "epoch": 0.19, "learning_rate": 4.68577140306271e-05, "loss": 0.0049, "step": 9716 }, { "epoch": 0.19, "learning_rate": 4.685706707036987e-05, "loss": 0.0283, "step": 9718 }, { "epoch": 0.19, "learning_rate": 4.685642011011264e-05, "loss": 0.0038, "step": 9720 }, { "epoch": 0.19, "learning_rate": 4.685577314985541e-05, "loss": 0.004, "step": 9722 }, { "epoch": 0.19, "learning_rate": 4.685512618959818e-05, "loss": 0.0029, "step": 9724 }, { "epoch": 0.19, "learning_rate": 4.685447922934094e-05, "loss": 0.0136, "step": 9726 }, { "epoch": 0.19, "learning_rate": 4.6853832269083715e-05, "loss": 0.0322, "step": 9728 }, { "epoch": 0.19, "learning_rate": 4.685318530882648e-05, "loss": 0.0043, "step": 9730 }, { "epoch": 0.19, "learning_rate": 4.6852538348569246e-05, "loss": 0.0158, "step": 9732 }, { "epoch": 0.19, "learning_rate": 4.685189138831202e-05, "loss": 0.0241, "step": 9734 }, { "epoch": 0.19, "learning_rate": 4.6851244428054785e-05, "loss": 0.0208, "step": 9736 }, { "epoch": 0.19, "learning_rate": 4.685059746779756e-05, "loss": 0.0034, "step": 9738 }, { "epoch": 0.19, "learning_rate": 4.684995050754032e-05, "loss": 0.0303, "step": 9740 }, { "epoch": 0.19, "learning_rate": 4.684930354728309e-05, "loss": 0.0066, "step": 9742 }, { "epoch": 0.19, "learning_rate": 4.684865658702586e-05, "loss": 0.0221, "step": 9744 }, { "epoch": 0.19, "learning_rate": 4.684800962676863e-05, "loss": 0.0064, "step": 9746 }, { "epoch": 0.19, "learning_rate": 4.68473626665114e-05, "loss": 0.0035, "step": 9748 }, { "epoch": 0.19, "learning_rate": 4.684671570625417e-05, "loss": 0.0116, "step": 9750 }, { "epoch": 0.19, "learning_rate": 4.684606874599694e-05, "loss": 0.003, "step": 9752 }, { "epoch": 0.19, "learning_rate": 4.68454217857397e-05, "loss": 0.0152, "step": 9754 }, { "epoch": 0.19, "learning_rate": 4.6844774825482476e-05, "loss": 0.0102, "step": 9756 }, { "epoch": 0.19, "learning_rate": 4.684412786522524e-05, "loss": 0.0122, "step": 9758 }, { "epoch": 0.19, "learning_rate": 4.6843480904968014e-05, "loss": 0.0165, "step": 9760 }, { "epoch": 0.19, "learning_rate": 4.6842833944710776e-05, "loss": 0.0136, "step": 9762 }, { "epoch": 0.19, "learning_rate": 4.6842186984453545e-05, "loss": 0.0127, "step": 9764 }, { "epoch": 0.19, "learning_rate": 4.684154002419632e-05, "loss": 0.0045, "step": 9766 }, { "epoch": 0.19, "learning_rate": 4.684089306393908e-05, "loss": 0.0104, "step": 9768 }, { "epoch": 0.19, "learning_rate": 4.684024610368185e-05, "loss": 0.0514, "step": 9770 }, { "epoch": 0.19, "learning_rate": 4.683959914342462e-05, "loss": 0.0036, "step": 9772 }, { "epoch": 0.19, "learning_rate": 4.683895218316739e-05, "loss": 0.0128, "step": 9774 }, { "epoch": 0.19, "learning_rate": 4.683830522291016e-05, "loss": 0.0428, "step": 9776 }, { "epoch": 0.19, "learning_rate": 4.683765826265293e-05, "loss": 0.0092, "step": 9778 }, { "epoch": 0.19, "learning_rate": 4.68370113023957e-05, "loss": 0.0181, "step": 9780 }, { "epoch": 0.19, "learning_rate": 4.683636434213847e-05, "loss": 0.0107, "step": 9782 }, { "epoch": 0.19, "learning_rate": 4.6835717381881236e-05, "loss": 0.0202, "step": 9784 }, { "epoch": 0.19, "learning_rate": 4.6835070421624e-05, "loss": 0.0049, "step": 9786 }, { "epoch": 0.19, "learning_rate": 4.6834423461366774e-05, "loss": 0.03, "step": 9788 }, { "epoch": 0.19, "learning_rate": 4.6833776501109536e-05, "loss": 0.0218, "step": 9790 }, { "epoch": 0.19, "learning_rate": 4.6833129540852305e-05, "loss": 0.0269, "step": 9792 }, { "epoch": 0.19, "learning_rate": 4.6832482580595075e-05, "loss": 0.0097, "step": 9794 }, { "epoch": 0.19, "learning_rate": 4.6831835620337844e-05, "loss": 0.004, "step": 9796 }, { "epoch": 0.19, "learning_rate": 4.683118866008061e-05, "loss": 0.0024, "step": 9798 }, { "epoch": 0.19, "learning_rate": 4.683054169982338e-05, "loss": 0.0022, "step": 9800 }, { "epoch": 0.19, "learning_rate": 4.682989473956615e-05, "loss": 0.0089, "step": 9802 }, { "epoch": 0.19, "learning_rate": 4.682924777930892e-05, "loss": 0.0048, "step": 9804 }, { "epoch": 0.19, "learning_rate": 4.682860081905169e-05, "loss": 0.03, "step": 9806 }, { "epoch": 0.19, "learning_rate": 4.682795385879445e-05, "loss": 0.0044, "step": 9808 }, { "epoch": 0.19, "learning_rate": 4.682730689853723e-05, "loss": 0.006, "step": 9810 }, { "epoch": 0.19, "learning_rate": 4.6826659938279996e-05, "loss": 0.0167, "step": 9812 }, { "epoch": 0.19, "learning_rate": 4.682601297802276e-05, "loss": 0.0172, "step": 9814 }, { "epoch": 0.19, "learning_rate": 4.6825366017765534e-05, "loss": 0.0204, "step": 9816 }, { "epoch": 0.19, "learning_rate": 4.68247190575083e-05, "loss": 0.0043, "step": 9818 }, { "epoch": 0.19, "learning_rate": 4.682407209725107e-05, "loss": 0.048, "step": 9820 }, { "epoch": 0.19, "learning_rate": 4.6823425136993835e-05, "loss": 0.0048, "step": 9822 }, { "epoch": 0.19, "learning_rate": 4.6822778176736604e-05, "loss": 0.0123, "step": 9824 }, { "epoch": 0.19, "learning_rate": 4.682213121647937e-05, "loss": 0.0202, "step": 9826 }, { "epoch": 0.19, "learning_rate": 4.682148425622214e-05, "loss": 0.0015, "step": 9828 }, { "epoch": 0.19, "learning_rate": 4.682083729596491e-05, "loss": 0.0073, "step": 9830 }, { "epoch": 0.19, "learning_rate": 4.682019033570768e-05, "loss": 0.009, "step": 9832 }, { "epoch": 0.19, "learning_rate": 4.681954337545045e-05, "loss": 0.0275, "step": 9834 }, { "epoch": 0.19, "learning_rate": 4.681889641519322e-05, "loss": 0.0059, "step": 9836 }, { "epoch": 0.19, "learning_rate": 4.681824945493599e-05, "loss": 0.0201, "step": 9838 }, { "epoch": 0.19, "learning_rate": 4.681760249467875e-05, "loss": 0.0383, "step": 9840 }, { "epoch": 0.19, "learning_rate": 4.6816955534421526e-05, "loss": 0.007, "step": 9842 }, { "epoch": 0.19, "learning_rate": 4.6816308574164295e-05, "loss": 0.0045, "step": 9844 }, { "epoch": 0.19, "learning_rate": 4.681566161390706e-05, "loss": 0.0197, "step": 9846 }, { "epoch": 0.19, "learning_rate": 4.681501465364983e-05, "loss": 0.0055, "step": 9848 }, { "epoch": 0.19, "learning_rate": 4.6814367693392595e-05, "loss": 0.0112, "step": 9850 }, { "epoch": 0.19, "learning_rate": 4.6813720733135364e-05, "loss": 0.0173, "step": 9852 }, { "epoch": 0.19, "learning_rate": 4.6813073772878134e-05, "loss": 0.0132, "step": 9854 }, { "epoch": 0.19, "learning_rate": 4.68124268126209e-05, "loss": 0.0238, "step": 9856 }, { "epoch": 0.19, "learning_rate": 4.681177985236367e-05, "loss": 0.007, "step": 9858 }, { "epoch": 0.19, "learning_rate": 4.681113289210644e-05, "loss": 0.012, "step": 9860 }, { "epoch": 0.19, "learning_rate": 4.681048593184921e-05, "loss": 0.0158, "step": 9862 }, { "epoch": 0.19, "learning_rate": 4.680983897159198e-05, "loss": 0.0103, "step": 9864 }, { "epoch": 0.19, "learning_rate": 4.680919201133475e-05, "loss": 0.0053, "step": 9866 }, { "epoch": 0.19, "learning_rate": 4.680854505107751e-05, "loss": 0.0127, "step": 9868 }, { "epoch": 0.19, "learning_rate": 4.6807898090820286e-05, "loss": 0.0095, "step": 9870 }, { "epoch": 0.19, "learning_rate": 4.680725113056305e-05, "loss": 0.0084, "step": 9872 }, { "epoch": 0.19, "learning_rate": 4.680660417030582e-05, "loss": 0.0029, "step": 9874 }, { "epoch": 0.19, "learning_rate": 4.680595721004859e-05, "loss": 0.0245, "step": 9876 }, { "epoch": 0.19, "learning_rate": 4.6805310249791356e-05, "loss": 0.0322, "step": 9878 }, { "epoch": 0.19, "learning_rate": 4.680466328953413e-05, "loss": 0.0023, "step": 9880 }, { "epoch": 0.19, "learning_rate": 4.6804016329276894e-05, "loss": 0.0317, "step": 9882 }, { "epoch": 0.19, "learning_rate": 4.680336936901966e-05, "loss": 0.0032, "step": 9884 }, { "epoch": 0.19, "learning_rate": 4.680272240876243e-05, "loss": 0.0193, "step": 9886 }, { "epoch": 0.19, "learning_rate": 4.68020754485052e-05, "loss": 0.0147, "step": 9888 }, { "epoch": 0.19, "learning_rate": 4.680142848824797e-05, "loss": 0.0121, "step": 9890 }, { "epoch": 0.19, "learning_rate": 4.680078152799074e-05, "loss": 0.0146, "step": 9892 }, { "epoch": 0.19, "learning_rate": 4.680013456773351e-05, "loss": 0.0229, "step": 9894 }, { "epoch": 0.19, "learning_rate": 4.679948760747627e-05, "loss": 0.0298, "step": 9896 }, { "epoch": 0.19, "learning_rate": 4.6798840647219047e-05, "loss": 0.0314, "step": 9898 }, { "epoch": 0.19, "learning_rate": 4.679819368696181e-05, "loss": 0.0142, "step": 9900 }, { "epoch": 0.19, "learning_rate": 4.6797546726704585e-05, "loss": 0.0037, "step": 9902 }, { "epoch": 0.19, "learning_rate": 4.679689976644735e-05, "loss": 0.0136, "step": 9904 }, { "epoch": 0.19, "learning_rate": 4.6796252806190116e-05, "loss": 0.0067, "step": 9906 }, { "epoch": 0.19, "learning_rate": 4.6795605845932885e-05, "loss": 0.0126, "step": 9908 }, { "epoch": 0.19, "learning_rate": 4.6794958885675654e-05, "loss": 0.0135, "step": 9910 }, { "epoch": 0.19, "learning_rate": 4.6794311925418423e-05, "loss": 0.0089, "step": 9912 }, { "epoch": 0.19, "learning_rate": 4.679366496516119e-05, "loss": 0.0146, "step": 9914 }, { "epoch": 0.19, "learning_rate": 4.679301800490396e-05, "loss": 0.0086, "step": 9916 }, { "epoch": 0.19, "learning_rate": 4.679237104464673e-05, "loss": 0.0215, "step": 9918 }, { "epoch": 0.19, "learning_rate": 4.67917240843895e-05, "loss": 0.0335, "step": 9920 }, { "epoch": 0.19, "learning_rate": 4.679107712413226e-05, "loss": 0.0095, "step": 9922 }, { "epoch": 0.19, "learning_rate": 4.679043016387504e-05, "loss": 0.0243, "step": 9924 }, { "epoch": 0.19, "learning_rate": 4.678978320361781e-05, "loss": 0.0075, "step": 9926 }, { "epoch": 0.19, "learning_rate": 4.678913624336057e-05, "loss": 0.004, "step": 9928 }, { "epoch": 0.19, "learning_rate": 4.6788489283103345e-05, "loss": 0.0023, "step": 9930 }, { "epoch": 0.19, "learning_rate": 4.678784232284611e-05, "loss": 0.0164, "step": 9932 }, { "epoch": 0.19, "learning_rate": 4.6787195362588877e-05, "loss": 0.0365, "step": 9934 }, { "epoch": 0.19, "learning_rate": 4.6786548402331646e-05, "loss": 0.002, "step": 9936 }, { "epoch": 0.19, "learning_rate": 4.6785901442074415e-05, "loss": 0.0047, "step": 9938 }, { "epoch": 0.19, "learning_rate": 4.6785254481817184e-05, "loss": 0.0184, "step": 9940 }, { "epoch": 0.19, "learning_rate": 4.678460752155995e-05, "loss": 0.0202, "step": 9942 }, { "epoch": 0.19, "learning_rate": 4.678396056130272e-05, "loss": 0.02, "step": 9944 }, { "epoch": 0.19, "learning_rate": 4.678331360104549e-05, "loss": 0.0268, "step": 9946 }, { "epoch": 0.19, "learning_rate": 4.678266664078826e-05, "loss": 0.0052, "step": 9948 }, { "epoch": 0.19, "learning_rate": 4.678201968053102e-05, "loss": 0.0017, "step": 9950 }, { "epoch": 0.19, "learning_rate": 4.67813727202738e-05, "loss": 0.0151, "step": 9952 }, { "epoch": 0.19, "learning_rate": 4.678072576001656e-05, "loss": 0.0058, "step": 9954 }, { "epoch": 0.19, "learning_rate": 4.678007879975933e-05, "loss": 0.0083, "step": 9956 }, { "epoch": 0.19, "learning_rate": 4.6779431839502106e-05, "loss": 0.0032, "step": 9958 }, { "epoch": 0.19, "learning_rate": 4.677878487924487e-05, "loss": 0.0022, "step": 9960 }, { "epoch": 0.19, "learning_rate": 4.6778137918987644e-05, "loss": 0.0183, "step": 9962 }, { "epoch": 0.19, "learning_rate": 4.6777490958730406e-05, "loss": 0.0044, "step": 9964 }, { "epoch": 0.19, "learning_rate": 4.6776843998473175e-05, "loss": 0.0075, "step": 9966 }, { "epoch": 0.19, "learning_rate": 4.6776197038215944e-05, "loss": 0.0065, "step": 9968 }, { "epoch": 0.19, "learning_rate": 4.677555007795871e-05, "loss": 0.0081, "step": 9970 }, { "epoch": 0.19, "learning_rate": 4.677490311770148e-05, "loss": 0.0014, "step": 9972 }, { "epoch": 0.19, "learning_rate": 4.677425615744425e-05, "loss": 0.0046, "step": 9974 }, { "epoch": 0.19, "learning_rate": 4.677360919718702e-05, "loss": 0.0035, "step": 9976 }, { "epoch": 0.19, "learning_rate": 4.677296223692979e-05, "loss": 0.0169, "step": 9978 }, { "epoch": 0.19, "learning_rate": 4.677231527667256e-05, "loss": 0.0082, "step": 9980 }, { "epoch": 0.19, "learning_rate": 4.677166831641532e-05, "loss": 0.004, "step": 9982 }, { "epoch": 0.19, "learning_rate": 4.67710213561581e-05, "loss": 0.0108, "step": 9984 }, { "epoch": 0.19, "learning_rate": 4.677037439590086e-05, "loss": 0.0045, "step": 9986 }, { "epoch": 0.19, "learning_rate": 4.676972743564363e-05, "loss": 0.0042, "step": 9988 }, { "epoch": 0.19, "learning_rate": 4.6769080475386404e-05, "loss": 0.0138, "step": 9990 }, { "epoch": 0.19, "learning_rate": 4.6768433515129166e-05, "loss": 0.0109, "step": 9992 }, { "epoch": 0.19, "learning_rate": 4.6767786554871936e-05, "loss": 0.0044, "step": 9994 }, { "epoch": 0.19, "learning_rate": 4.6767139594614705e-05, "loss": 0.0041, "step": 9996 }, { "epoch": 0.19, "learning_rate": 4.6766492634357474e-05, "loss": 0.0073, "step": 9998 }, { "epoch": 0.19, "learning_rate": 4.676584567410024e-05, "loss": 0.0042, "step": 10000 }, { "epoch": 0.19, "learning_rate": 4.676519871384301e-05, "loss": 0.0074, "step": 10002 }, { "epoch": 0.19, "learning_rate": 4.676455175358578e-05, "loss": 0.0046, "step": 10004 }, { "epoch": 0.19, "learning_rate": 4.676390479332855e-05, "loss": 0.0022, "step": 10006 }, { "epoch": 0.19, "learning_rate": 4.676325783307132e-05, "loss": 0.0082, "step": 10008 }, { "epoch": 0.19, "learning_rate": 4.676261087281408e-05, "loss": 0.0128, "step": 10010 }, { "epoch": 0.19, "learning_rate": 4.676196391255686e-05, "loss": 0.0074, "step": 10012 }, { "epoch": 0.19, "learning_rate": 4.676131695229962e-05, "loss": 0.049, "step": 10014 }, { "epoch": 0.19, "learning_rate": 4.676066999204239e-05, "loss": 0.0031, "step": 10016 }, { "epoch": 0.19, "learning_rate": 4.676002303178516e-05, "loss": 0.003, "step": 10018 }, { "epoch": 0.19, "learning_rate": 4.675937607152793e-05, "loss": 0.0039, "step": 10020 }, { "epoch": 0.19, "learning_rate": 4.6758729111270696e-05, "loss": 0.004, "step": 10022 }, { "epoch": 0.19, "learning_rate": 4.6758082151013465e-05, "loss": 0.0053, "step": 10024 }, { "epoch": 0.19, "learning_rate": 4.6757435190756234e-05, "loss": 0.0086, "step": 10026 }, { "epoch": 0.19, "learning_rate": 4.6756788230499e-05, "loss": 0.0165, "step": 10028 }, { "epoch": 0.19, "learning_rate": 4.675614127024177e-05, "loss": 0.0245, "step": 10030 }, { "epoch": 0.19, "learning_rate": 4.6755494309984535e-05, "loss": 0.0044, "step": 10032 }, { "epoch": 0.19, "learning_rate": 4.675484734972731e-05, "loss": 0.0085, "step": 10034 }, { "epoch": 0.19, "learning_rate": 4.675420038947008e-05, "loss": 0.0049, "step": 10036 }, { "epoch": 0.19, "learning_rate": 4.675355342921284e-05, "loss": 0.0654, "step": 10038 }, { "epoch": 0.19, "learning_rate": 4.675290646895562e-05, "loss": 0.0113, "step": 10040 }, { "epoch": 0.19, "learning_rate": 4.675225950869838e-05, "loss": 0.0076, "step": 10042 }, { "epoch": 0.19, "learning_rate": 4.6751612548441156e-05, "loss": 0.009, "step": 10044 }, { "epoch": 0.19, "learning_rate": 4.675096558818392e-05, "loss": 0.024, "step": 10046 }, { "epoch": 0.2, "learning_rate": 4.675031862792669e-05, "loss": 0.0461, "step": 10048 }, { "epoch": 0.2, "learning_rate": 4.6749671667669456e-05, "loss": 0.0231, "step": 10050 }, { "epoch": 0.2, "learning_rate": 4.6749024707412225e-05, "loss": 0.0033, "step": 10052 }, { "epoch": 0.2, "learning_rate": 4.6748377747154995e-05, "loss": 0.0064, "step": 10054 }, { "epoch": 0.2, "learning_rate": 4.6747730786897764e-05, "loss": 0.0027, "step": 10056 }, { "epoch": 0.2, "learning_rate": 4.674708382664053e-05, "loss": 0.0427, "step": 10058 }, { "epoch": 0.2, "learning_rate": 4.67464368663833e-05, "loss": 0.0055, "step": 10060 }, { "epoch": 0.2, "learning_rate": 4.674578990612607e-05, "loss": 0.0309, "step": 10062 }, { "epoch": 0.2, "learning_rate": 4.674514294586883e-05, "loss": 0.0219, "step": 10064 }, { "epoch": 0.2, "learning_rate": 4.674449598561161e-05, "loss": 0.0052, "step": 10066 }, { "epoch": 0.2, "learning_rate": 4.674384902535438e-05, "loss": 0.0015, "step": 10068 }, { "epoch": 0.2, "learning_rate": 4.674320206509714e-05, "loss": 0.0184, "step": 10070 }, { "epoch": 0.2, "learning_rate": 4.6742555104839916e-05, "loss": 0.0209, "step": 10072 }, { "epoch": 0.2, "learning_rate": 4.674190814458268e-05, "loss": 0.0102, "step": 10074 }, { "epoch": 0.2, "learning_rate": 4.674126118432545e-05, "loss": 0.0082, "step": 10076 }, { "epoch": 0.2, "learning_rate": 4.674061422406822e-05, "loss": 0.0017, "step": 10078 }, { "epoch": 0.2, "learning_rate": 4.6739967263810986e-05, "loss": 0.0024, "step": 10080 }, { "epoch": 0.2, "learning_rate": 4.6739320303553755e-05, "loss": 0.0043, "step": 10082 }, { "epoch": 0.2, "learning_rate": 4.6738673343296524e-05, "loss": 0.001, "step": 10084 }, { "epoch": 0.2, "learning_rate": 4.673802638303929e-05, "loss": 0.0285, "step": 10086 }, { "epoch": 0.2, "learning_rate": 4.673737942278206e-05, "loss": 0.0031, "step": 10088 }, { "epoch": 0.2, "learning_rate": 4.673673246252483e-05, "loss": 0.002, "step": 10090 }, { "epoch": 0.2, "learning_rate": 4.6736085502267594e-05, "loss": 0.005, "step": 10092 }, { "epoch": 0.2, "learning_rate": 4.673543854201037e-05, "loss": 0.0118, "step": 10094 }, { "epoch": 0.2, "learning_rate": 4.673479158175313e-05, "loss": 0.0045, "step": 10096 }, { "epoch": 0.2, "learning_rate": 4.67341446214959e-05, "loss": 0.0044, "step": 10098 }, { "epoch": 0.2, "learning_rate": 4.673349766123867e-05, "loss": 0.0138, "step": 10100 }, { "epoch": 0.2, "learning_rate": 4.673285070098144e-05, "loss": 0.0074, "step": 10102 }, { "epoch": 0.2, "learning_rate": 4.6732203740724215e-05, "loss": 0.0332, "step": 10104 }, { "epoch": 0.2, "learning_rate": 4.673155678046698e-05, "loss": 0.0098, "step": 10106 }, { "epoch": 0.2, "learning_rate": 4.6730909820209746e-05, "loss": 0.0046, "step": 10108 }, { "epoch": 0.2, "learning_rate": 4.6730262859952515e-05, "loss": 0.0166, "step": 10110 }, { "epoch": 0.2, "learning_rate": 4.6729615899695284e-05, "loss": 0.0124, "step": 10112 }, { "epoch": 0.2, "learning_rate": 4.6728968939438054e-05, "loss": 0.001, "step": 10114 }, { "epoch": 0.2, "learning_rate": 4.672832197918082e-05, "loss": 0.0155, "step": 10116 }, { "epoch": 0.2, "learning_rate": 4.672767501892359e-05, "loss": 0.0041, "step": 10118 }, { "epoch": 0.2, "learning_rate": 4.672702805866636e-05, "loss": 0.0391, "step": 10120 }, { "epoch": 0.2, "learning_rate": 4.672638109840913e-05, "loss": 0.0191, "step": 10122 }, { "epoch": 0.2, "learning_rate": 4.672573413815189e-05, "loss": 0.0095, "step": 10124 }, { "epoch": 0.2, "learning_rate": 4.672508717789467e-05, "loss": 0.0019, "step": 10126 }, { "epoch": 0.2, "learning_rate": 4.672444021763743e-05, "loss": 0.0026, "step": 10128 }, { "epoch": 0.2, "learning_rate": 4.67237932573802e-05, "loss": 0.0017, "step": 10130 }, { "epoch": 0.2, "learning_rate": 4.672314629712297e-05, "loss": 0.0085, "step": 10132 }, { "epoch": 0.2, "learning_rate": 4.672249933686574e-05, "loss": 0.0069, "step": 10134 }, { "epoch": 0.2, "learning_rate": 4.672185237660851e-05, "loss": 0.0097, "step": 10136 }, { "epoch": 0.2, "learning_rate": 4.6721205416351276e-05, "loss": 0.0074, "step": 10138 }, { "epoch": 0.2, "learning_rate": 4.6720558456094045e-05, "loss": 0.0109, "step": 10140 }, { "epoch": 0.2, "learning_rate": 4.6719911495836814e-05, "loss": 0.0171, "step": 10142 }, { "epoch": 0.2, "learning_rate": 4.671926453557958e-05, "loss": 0.0025, "step": 10144 }, { "epoch": 0.2, "learning_rate": 4.671861757532235e-05, "loss": 0.0124, "step": 10146 }, { "epoch": 0.2, "learning_rate": 4.671797061506512e-05, "loss": 0.0042, "step": 10148 }, { "epoch": 0.2, "learning_rate": 4.671732365480789e-05, "loss": 0.0058, "step": 10150 }, { "epoch": 0.2, "learning_rate": 4.671667669455065e-05, "loss": 0.0122, "step": 10152 }, { "epoch": 0.2, "learning_rate": 4.671602973429343e-05, "loss": 0.027, "step": 10154 }, { "epoch": 0.2, "learning_rate": 4.671538277403619e-05, "loss": 0.0054, "step": 10156 }, { "epoch": 0.2, "learning_rate": 4.671473581377896e-05, "loss": 0.0029, "step": 10158 }, { "epoch": 0.2, "learning_rate": 4.671408885352173e-05, "loss": 0.0166, "step": 10160 }, { "epoch": 0.2, "learning_rate": 4.67134418932645e-05, "loss": 0.0274, "step": 10162 }, { "epoch": 0.2, "learning_rate": 4.671279493300727e-05, "loss": 0.0128, "step": 10164 }, { "epoch": 0.2, "learning_rate": 4.6712147972750036e-05, "loss": 0.0088, "step": 10166 }, { "epoch": 0.2, "learning_rate": 4.6711501012492805e-05, "loss": 0.0137, "step": 10168 }, { "epoch": 0.2, "learning_rate": 4.6710854052235574e-05, "loss": 0.0269, "step": 10170 }, { "epoch": 0.2, "learning_rate": 4.6710207091978343e-05, "loss": 0.0132, "step": 10172 }, { "epoch": 0.2, "learning_rate": 4.6709560131721106e-05, "loss": 0.0232, "step": 10174 }, { "epoch": 0.2, "learning_rate": 4.670891317146388e-05, "loss": 0.0107, "step": 10176 }, { "epoch": 0.2, "learning_rate": 4.6708266211206644e-05, "loss": 0.0208, "step": 10178 }, { "epoch": 0.2, "learning_rate": 4.670761925094941e-05, "loss": 0.0229, "step": 10180 }, { "epoch": 0.2, "learning_rate": 4.670697229069219e-05, "loss": 0.0282, "step": 10182 }, { "epoch": 0.2, "learning_rate": 4.670632533043495e-05, "loss": 0.0035, "step": 10184 }, { "epoch": 0.2, "learning_rate": 4.670567837017773e-05, "loss": 0.006, "step": 10186 }, { "epoch": 0.2, "learning_rate": 4.670503140992049e-05, "loss": 0.0035, "step": 10188 }, { "epoch": 0.2, "learning_rate": 4.670438444966326e-05, "loss": 0.013, "step": 10190 }, { "epoch": 0.2, "learning_rate": 4.670373748940603e-05, "loss": 0.0105, "step": 10192 }, { "epoch": 0.2, "learning_rate": 4.6703090529148797e-05, "loss": 0.0036, "step": 10194 }, { "epoch": 0.2, "learning_rate": 4.6702443568891566e-05, "loss": 0.0115, "step": 10196 }, { "epoch": 0.2, "learning_rate": 4.6701796608634335e-05, "loss": 0.0126, "step": 10198 }, { "epoch": 0.2, "learning_rate": 4.6701149648377104e-05, "loss": 0.0499, "step": 10200 }, { "epoch": 0.2, "learning_rate": 4.670050268811987e-05, "loss": 0.0049, "step": 10202 }, { "epoch": 0.2, "learning_rate": 4.669985572786264e-05, "loss": 0.0045, "step": 10204 }, { "epoch": 0.2, "learning_rate": 4.6699208767605404e-05, "loss": 0.0086, "step": 10206 }, { "epoch": 0.2, "learning_rate": 4.669856180734818e-05, "loss": 0.0145, "step": 10208 }, { "epoch": 0.2, "learning_rate": 4.669791484709094e-05, "loss": 0.0007, "step": 10210 }, { "epoch": 0.2, "learning_rate": 4.669726788683371e-05, "loss": 0.0078, "step": 10212 }, { "epoch": 0.2, "learning_rate": 4.669662092657649e-05, "loss": 0.0101, "step": 10214 }, { "epoch": 0.2, "learning_rate": 4.669597396631925e-05, "loss": 0.0053, "step": 10216 }, { "epoch": 0.2, "learning_rate": 4.669532700606202e-05, "loss": 0.0277, "step": 10218 }, { "epoch": 0.2, "learning_rate": 4.669468004580479e-05, "loss": 0.0111, "step": 10220 }, { "epoch": 0.2, "learning_rate": 4.669403308554756e-05, "loss": 0.0031, "step": 10222 }, { "epoch": 0.2, "learning_rate": 4.6693386125290326e-05, "loss": 0.0394, "step": 10224 }, { "epoch": 0.2, "learning_rate": 4.6692739165033095e-05, "loss": 0.0063, "step": 10226 }, { "epoch": 0.2, "learning_rate": 4.6692092204775864e-05, "loss": 0.0133, "step": 10228 }, { "epoch": 0.2, "learning_rate": 4.669144524451863e-05, "loss": 0.0119, "step": 10230 }, { "epoch": 0.2, "learning_rate": 4.66907982842614e-05, "loss": 0.0302, "step": 10232 }, { "epoch": 0.2, "learning_rate": 4.6690151324004165e-05, "loss": 0.0039, "step": 10234 }, { "epoch": 0.2, "learning_rate": 4.668950436374694e-05, "loss": 0.0035, "step": 10236 }, { "epoch": 0.2, "learning_rate": 4.66888574034897e-05, "loss": 0.0268, "step": 10238 }, { "epoch": 0.2, "learning_rate": 4.668821044323247e-05, "loss": 0.005, "step": 10240 }, { "epoch": 0.2, "learning_rate": 4.668756348297524e-05, "loss": 0.0072, "step": 10242 }, { "epoch": 0.2, "learning_rate": 4.668691652271801e-05, "loss": 0.0023, "step": 10244 }, { "epoch": 0.2, "learning_rate": 4.6686269562460786e-05, "loss": 0.0039, "step": 10246 }, { "epoch": 0.2, "learning_rate": 4.668562260220355e-05, "loss": 0.0042, "step": 10248 }, { "epoch": 0.2, "learning_rate": 4.668497564194632e-05, "loss": 0.0013, "step": 10250 }, { "epoch": 0.2, "learning_rate": 4.6684328681689086e-05, "loss": 0.0018, "step": 10252 }, { "epoch": 0.2, "learning_rate": 4.6683681721431856e-05, "loss": 0.0013, "step": 10254 }, { "epoch": 0.2, "learning_rate": 4.668303476117462e-05, "loss": 0.0021, "step": 10256 }, { "epoch": 0.2, "learning_rate": 4.6682387800917394e-05, "loss": 0.0088, "step": 10258 }, { "epoch": 0.2, "learning_rate": 4.668174084066016e-05, "loss": 0.0313, "step": 10260 }, { "epoch": 0.2, "learning_rate": 4.668109388040293e-05, "loss": 0.0024, "step": 10262 }, { "epoch": 0.2, "learning_rate": 4.66804469201457e-05, "loss": 0.0049, "step": 10264 }, { "epoch": 0.2, "learning_rate": 4.667979995988846e-05, "loss": 0.0261, "step": 10266 }, { "epoch": 0.2, "learning_rate": 4.667915299963124e-05, "loss": 0.0044, "step": 10268 }, { "epoch": 0.2, "learning_rate": 4.6678506039374e-05, "loss": 0.0018, "step": 10270 }, { "epoch": 0.2, "learning_rate": 4.667785907911677e-05, "loss": 0.0046, "step": 10272 }, { "epoch": 0.2, "learning_rate": 4.667721211885954e-05, "loss": 0.0072, "step": 10274 }, { "epoch": 0.2, "learning_rate": 4.667656515860231e-05, "loss": 0.0052, "step": 10276 }, { "epoch": 0.2, "learning_rate": 4.667591819834508e-05, "loss": 0.004, "step": 10278 }, { "epoch": 0.2, "learning_rate": 4.667527123808785e-05, "loss": 0.0148, "step": 10280 }, { "epoch": 0.2, "learning_rate": 4.6674624277830616e-05, "loss": 0.0218, "step": 10282 }, { "epoch": 0.2, "learning_rate": 4.6673977317573385e-05, "loss": 0.0076, "step": 10284 }, { "epoch": 0.2, "learning_rate": 4.6673330357316154e-05, "loss": 0.0105, "step": 10286 }, { "epoch": 0.2, "learning_rate": 4.6672683397058916e-05, "loss": 0.0014, "step": 10288 }, { "epoch": 0.2, "learning_rate": 4.667203643680169e-05, "loss": 0.012, "step": 10290 }, { "epoch": 0.2, "learning_rate": 4.667138947654446e-05, "loss": 0.01, "step": 10292 }, { "epoch": 0.2, "learning_rate": 4.6670742516287224e-05, "loss": 0.0021, "step": 10294 }, { "epoch": 0.2, "learning_rate": 4.667009555603e-05, "loss": 0.001, "step": 10296 }, { "epoch": 0.2, "learning_rate": 4.666944859577276e-05, "loss": 0.004, "step": 10298 }, { "epoch": 0.2, "learning_rate": 4.666880163551553e-05, "loss": 0.0292, "step": 10300 }, { "epoch": 0.2, "learning_rate": 4.66681546752583e-05, "loss": 0.0037, "step": 10302 }, { "epoch": 0.2, "learning_rate": 4.666750771500107e-05, "loss": 0.0039, "step": 10304 }, { "epoch": 0.2, "learning_rate": 4.666686075474384e-05, "loss": 0.0118, "step": 10306 }, { "epoch": 0.2, "learning_rate": 4.666621379448661e-05, "loss": 0.0027, "step": 10308 }, { "epoch": 0.2, "learning_rate": 4.6665566834229376e-05, "loss": 0.0025, "step": 10310 }, { "epoch": 0.2, "learning_rate": 4.6664919873972145e-05, "loss": 0.004, "step": 10312 }, { "epoch": 0.2, "learning_rate": 4.6664272913714915e-05, "loss": 0.0145, "step": 10314 }, { "epoch": 0.2, "learning_rate": 4.666362595345768e-05, "loss": 0.0106, "step": 10316 }, { "epoch": 0.2, "learning_rate": 4.666297899320045e-05, "loss": 0.0145, "step": 10318 }, { "epoch": 0.2, "learning_rate": 4.6662332032943215e-05, "loss": 0.0192, "step": 10320 }, { "epoch": 0.2, "learning_rate": 4.6661685072685984e-05, "loss": 0.0028, "step": 10322 }, { "epoch": 0.2, "learning_rate": 4.666103811242876e-05, "loss": 0.0008, "step": 10324 }, { "epoch": 0.2, "learning_rate": 4.666039115217152e-05, "loss": 0.0102, "step": 10326 }, { "epoch": 0.2, "learning_rate": 4.66597441919143e-05, "loss": 0.042, "step": 10328 }, { "epoch": 0.2, "learning_rate": 4.665909723165706e-05, "loss": 0.0021, "step": 10330 }, { "epoch": 0.2, "learning_rate": 4.665845027139983e-05, "loss": 0.0084, "step": 10332 }, { "epoch": 0.2, "learning_rate": 4.66578033111426e-05, "loss": 0.0251, "step": 10334 }, { "epoch": 0.2, "learning_rate": 4.665715635088537e-05, "loss": 0.016, "step": 10336 }, { "epoch": 0.2, "learning_rate": 4.665650939062814e-05, "loss": 0.0073, "step": 10338 }, { "epoch": 0.2, "learning_rate": 4.6655862430370906e-05, "loss": 0.0185, "step": 10340 }, { "epoch": 0.2, "learning_rate": 4.6655215470113675e-05, "loss": 0.0236, "step": 10342 }, { "epoch": 0.2, "learning_rate": 4.6654568509856444e-05, "loss": 0.0336, "step": 10344 }, { "epoch": 0.2, "learning_rate": 4.665392154959921e-05, "loss": 0.036, "step": 10346 }, { "epoch": 0.2, "learning_rate": 4.6653274589341975e-05, "loss": 0.0163, "step": 10348 }, { "epoch": 0.2, "learning_rate": 4.665262762908475e-05, "loss": 0.0068, "step": 10350 }, { "epoch": 0.2, "learning_rate": 4.6651980668827514e-05, "loss": 0.0031, "step": 10352 }, { "epoch": 0.2, "learning_rate": 4.665133370857028e-05, "loss": 0.0192, "step": 10354 }, { "epoch": 0.2, "learning_rate": 4.665068674831305e-05, "loss": 0.0145, "step": 10356 }, { "epoch": 0.2, "learning_rate": 4.665003978805582e-05, "loss": 0.0086, "step": 10358 }, { "epoch": 0.2, "learning_rate": 4.664939282779859e-05, "loss": 0.0021, "step": 10360 }, { "epoch": 0.2, "learning_rate": 4.664874586754136e-05, "loss": 0.0146, "step": 10362 }, { "epoch": 0.2, "learning_rate": 4.664809890728413e-05, "loss": 0.0014, "step": 10364 }, { "epoch": 0.2, "learning_rate": 4.66474519470269e-05, "loss": 0.0079, "step": 10366 }, { "epoch": 0.2, "learning_rate": 4.6646804986769666e-05, "loss": 0.0033, "step": 10368 }, { "epoch": 0.2, "learning_rate": 4.6646158026512435e-05, "loss": 0.0049, "step": 10370 }, { "epoch": 0.2, "learning_rate": 4.6645511066255204e-05, "loss": 0.0101, "step": 10372 }, { "epoch": 0.2, "learning_rate": 4.6644864105997973e-05, "loss": 0.0056, "step": 10374 }, { "epoch": 0.2, "learning_rate": 4.6644217145740736e-05, "loss": 0.0162, "step": 10376 }, { "epoch": 0.2, "learning_rate": 4.664357018548351e-05, "loss": 0.0023, "step": 10378 }, { "epoch": 0.2, "learning_rate": 4.6642923225226274e-05, "loss": 0.019, "step": 10380 }, { "epoch": 0.2, "learning_rate": 4.664227626496904e-05, "loss": 0.0038, "step": 10382 }, { "epoch": 0.2, "learning_rate": 4.664162930471181e-05, "loss": 0.0027, "step": 10384 }, { "epoch": 0.2, "learning_rate": 4.664098234445458e-05, "loss": 0.0015, "step": 10386 }, { "epoch": 0.2, "learning_rate": 4.664033538419735e-05, "loss": 0.0032, "step": 10388 }, { "epoch": 0.2, "learning_rate": 4.663968842394012e-05, "loss": 0.0034, "step": 10390 }, { "epoch": 0.2, "learning_rate": 4.663904146368289e-05, "loss": 0.0037, "step": 10392 }, { "epoch": 0.2, "learning_rate": 4.663839450342566e-05, "loss": 0.0103, "step": 10394 }, { "epoch": 0.2, "learning_rate": 4.663774754316843e-05, "loss": 0.0028, "step": 10396 }, { "epoch": 0.2, "learning_rate": 4.663710058291119e-05, "loss": 0.0008, "step": 10398 }, { "epoch": 0.2, "learning_rate": 4.6636453622653965e-05, "loss": 0.0077, "step": 10400 }, { "epoch": 0.2, "learning_rate": 4.663580666239673e-05, "loss": 0.0197, "step": 10402 }, { "epoch": 0.2, "learning_rate": 4.6635159702139496e-05, "loss": 0.0072, "step": 10404 }, { "epoch": 0.2, "learning_rate": 4.663451274188227e-05, "loss": 0.0035, "step": 10406 }, { "epoch": 0.2, "learning_rate": 4.6633865781625034e-05, "loss": 0.1665, "step": 10408 }, { "epoch": 0.2, "learning_rate": 4.663321882136781e-05, "loss": 0.0087, "step": 10410 }, { "epoch": 0.2, "learning_rate": 4.663257186111057e-05, "loss": 0.0062, "step": 10412 }, { "epoch": 0.2, "learning_rate": 4.663192490085334e-05, "loss": 0.0056, "step": 10414 }, { "epoch": 0.2, "learning_rate": 4.663127794059611e-05, "loss": 0.0034, "step": 10416 }, { "epoch": 0.2, "learning_rate": 4.663063098033888e-05, "loss": 0.0096, "step": 10418 }, { "epoch": 0.2, "learning_rate": 4.662998402008165e-05, "loss": 0.0043, "step": 10420 }, { "epoch": 0.2, "learning_rate": 4.662933705982442e-05, "loss": 0.0056, "step": 10422 }, { "epoch": 0.2, "learning_rate": 4.662869009956719e-05, "loss": 0.0103, "step": 10424 }, { "epoch": 0.2, "learning_rate": 4.6628043139309956e-05, "loss": 0.0085, "step": 10426 }, { "epoch": 0.2, "learning_rate": 4.6627396179052725e-05, "loss": 0.016, "step": 10428 }, { "epoch": 0.2, "learning_rate": 4.662674921879549e-05, "loss": 0.0085, "step": 10430 }, { "epoch": 0.2, "learning_rate": 4.662610225853826e-05, "loss": 0.0011, "step": 10432 }, { "epoch": 0.2, "learning_rate": 4.6625455298281026e-05, "loss": 0.0017, "step": 10434 }, { "epoch": 0.2, "learning_rate": 4.6624808338023795e-05, "loss": 0.0117, "step": 10436 }, { "epoch": 0.2, "learning_rate": 4.662416137776657e-05, "loss": 0.0059, "step": 10438 }, { "epoch": 0.2, "learning_rate": 4.662351441750933e-05, "loss": 0.0156, "step": 10440 }, { "epoch": 0.2, "learning_rate": 4.66228674572521e-05, "loss": 0.0151, "step": 10442 }, { "epoch": 0.2, "learning_rate": 4.662222049699487e-05, "loss": 0.0118, "step": 10444 }, { "epoch": 0.2, "learning_rate": 4.662157353673764e-05, "loss": 0.005, "step": 10446 }, { "epoch": 0.2, "learning_rate": 4.662092657648041e-05, "loss": 0.0099, "step": 10448 }, { "epoch": 0.2, "learning_rate": 4.662027961622318e-05, "loss": 0.0062, "step": 10450 }, { "epoch": 0.2, "learning_rate": 4.661963265596595e-05, "loss": 0.0141, "step": 10452 }, { "epoch": 0.2, "learning_rate": 4.6618985695708717e-05, "loss": 0.009, "step": 10454 }, { "epoch": 0.2, "learning_rate": 4.6618338735451486e-05, "loss": 0.0085, "step": 10456 }, { "epoch": 0.2, "learning_rate": 4.661769177519425e-05, "loss": 0.0059, "step": 10458 }, { "epoch": 0.2, "learning_rate": 4.6617044814937024e-05, "loss": 0.0078, "step": 10460 }, { "epoch": 0.2, "learning_rate": 4.6616397854679786e-05, "loss": 0.0036, "step": 10462 }, { "epoch": 0.2, "learning_rate": 4.6615750894422555e-05, "loss": 0.0058, "step": 10464 }, { "epoch": 0.2, "learning_rate": 4.6615103934165324e-05, "loss": 0.003, "step": 10466 }, { "epoch": 0.2, "learning_rate": 4.661445697390809e-05, "loss": 0.0291, "step": 10468 }, { "epoch": 0.2, "learning_rate": 4.661381001365087e-05, "loss": 0.0126, "step": 10470 }, { "epoch": 0.2, "learning_rate": 4.661316305339363e-05, "loss": 0.0042, "step": 10472 }, { "epoch": 0.2, "learning_rate": 4.66125160931364e-05, "loss": 0.0148, "step": 10474 }, { "epoch": 0.2, "learning_rate": 4.661186913287917e-05, "loss": 0.0061, "step": 10476 }, { "epoch": 0.2, "learning_rate": 4.661122217262194e-05, "loss": 0.0179, "step": 10478 }, { "epoch": 0.2, "learning_rate": 4.66105752123647e-05, "loss": 0.008, "step": 10480 }, { "epoch": 0.2, "learning_rate": 4.660992825210748e-05, "loss": 0.0088, "step": 10482 }, { "epoch": 0.2, "learning_rate": 4.6609281291850246e-05, "loss": 0.0077, "step": 10484 }, { "epoch": 0.2, "learning_rate": 4.6608634331593015e-05, "loss": 0.0073, "step": 10486 }, { "epoch": 0.2, "learning_rate": 4.6607987371335784e-05, "loss": 0.0124, "step": 10488 }, { "epoch": 0.2, "learning_rate": 4.6607340411078547e-05, "loss": 0.0033, "step": 10490 }, { "epoch": 0.2, "learning_rate": 4.660669345082132e-05, "loss": 0.0112, "step": 10492 }, { "epoch": 0.2, "learning_rate": 4.6606046490564085e-05, "loss": 0.0024, "step": 10494 }, { "epoch": 0.2, "learning_rate": 4.6605399530306854e-05, "loss": 0.0293, "step": 10496 }, { "epoch": 0.2, "learning_rate": 4.660475257004962e-05, "loss": 0.0028, "step": 10498 }, { "epoch": 0.2, "learning_rate": 4.660410560979239e-05, "loss": 0.0536, "step": 10500 }, { "epoch": 0.2, "learning_rate": 4.660345864953516e-05, "loss": 0.0022, "step": 10502 }, { "epoch": 0.2, "learning_rate": 4.660281168927793e-05, "loss": 0.0166, "step": 10504 }, { "epoch": 0.2, "learning_rate": 4.66021647290207e-05, "loss": 0.0171, "step": 10506 }, { "epoch": 0.2, "learning_rate": 4.660151776876347e-05, "loss": 0.0211, "step": 10508 }, { "epoch": 0.2, "learning_rate": 4.660087080850624e-05, "loss": 0.0018, "step": 10510 }, { "epoch": 0.2, "learning_rate": 4.6600223848249e-05, "loss": 0.0015, "step": 10512 }, { "epoch": 0.2, "learning_rate": 4.6599576887991776e-05, "loss": 0.0071, "step": 10514 }, { "epoch": 0.2, "learning_rate": 4.6598929927734545e-05, "loss": 0.0125, "step": 10516 }, { "epoch": 0.2, "learning_rate": 4.659828296747731e-05, "loss": 0.0049, "step": 10518 }, { "epoch": 0.2, "learning_rate": 4.659763600722008e-05, "loss": 0.0081, "step": 10520 }, { "epoch": 0.2, "learning_rate": 4.6596989046962845e-05, "loss": 0.0037, "step": 10522 }, { "epoch": 0.2, "learning_rate": 4.6596342086705614e-05, "loss": 0.0032, "step": 10524 }, { "epoch": 0.2, "learning_rate": 4.659569512644838e-05, "loss": 0.0063, "step": 10526 }, { "epoch": 0.2, "learning_rate": 4.659504816619115e-05, "loss": 0.0277, "step": 10528 }, { "epoch": 0.2, "learning_rate": 4.659440120593392e-05, "loss": 0.0016, "step": 10530 }, { "epoch": 0.2, "learning_rate": 4.659375424567669e-05, "loss": 0.0046, "step": 10532 }, { "epoch": 0.2, "learning_rate": 4.659310728541946e-05, "loss": 0.0185, "step": 10534 }, { "epoch": 0.2, "learning_rate": 4.659246032516223e-05, "loss": 0.0033, "step": 10536 }, { "epoch": 0.2, "learning_rate": 4.6591813364905e-05, "loss": 0.0301, "step": 10538 }, { "epoch": 0.2, "learning_rate": 4.659116640464776e-05, "loss": 0.0047, "step": 10540 }, { "epoch": 0.2, "learning_rate": 4.6590519444390536e-05, "loss": 0.0298, "step": 10542 }, { "epoch": 0.2, "learning_rate": 4.65898724841333e-05, "loss": 0.0187, "step": 10544 }, { "epoch": 0.2, "learning_rate": 4.658922552387607e-05, "loss": 0.008, "step": 10546 }, { "epoch": 0.2, "learning_rate": 4.658857856361884e-05, "loss": 0.0136, "step": 10548 }, { "epoch": 0.2, "learning_rate": 4.6587931603361605e-05, "loss": 0.0086, "step": 10550 }, { "epoch": 0.2, "learning_rate": 4.658728464310438e-05, "loss": 0.0404, "step": 10552 }, { "epoch": 0.2, "learning_rate": 4.6586637682847144e-05, "loss": 0.007, "step": 10554 }, { "epoch": 0.2, "learning_rate": 4.658599072258991e-05, "loss": 0.0021, "step": 10556 }, { "epoch": 0.2, "learning_rate": 4.658534376233268e-05, "loss": 0.0197, "step": 10558 }, { "epoch": 0.2, "learning_rate": 4.658469680207545e-05, "loss": 0.0068, "step": 10560 }, { "epoch": 0.2, "learning_rate": 4.658404984181822e-05, "loss": 0.0103, "step": 10562 }, { "epoch": 0.21, "learning_rate": 4.658340288156099e-05, "loss": 0.0107, "step": 10564 }, { "epoch": 0.21, "learning_rate": 4.658275592130376e-05, "loss": 0.0117, "step": 10566 }, { "epoch": 0.21, "learning_rate": 4.658210896104653e-05, "loss": 0.0063, "step": 10568 }, { "epoch": 0.21, "learning_rate": 4.6581462000789296e-05, "loss": 0.0213, "step": 10570 }, { "epoch": 0.21, "learning_rate": 4.658081504053206e-05, "loss": 0.0043, "step": 10572 }, { "epoch": 0.21, "learning_rate": 4.6580168080274834e-05, "loss": 0.0055, "step": 10574 }, { "epoch": 0.21, "learning_rate": 4.65795211200176e-05, "loss": 0.0181, "step": 10576 }, { "epoch": 0.21, "learning_rate": 4.6578874159760366e-05, "loss": 0.0093, "step": 10578 }, { "epoch": 0.21, "learning_rate": 4.6578227199503135e-05, "loss": 0.0023, "step": 10580 }, { "epoch": 0.21, "learning_rate": 4.6577580239245904e-05, "loss": 0.0378, "step": 10582 }, { "epoch": 0.21, "learning_rate": 4.657693327898867e-05, "loss": 0.0203, "step": 10584 }, { "epoch": 0.21, "learning_rate": 4.657628631873144e-05, "loss": 0.0045, "step": 10586 }, { "epoch": 0.21, "learning_rate": 4.657563935847421e-05, "loss": 0.0109, "step": 10588 }, { "epoch": 0.21, "learning_rate": 4.657499239821698e-05, "loss": 0.0266, "step": 10590 }, { "epoch": 0.21, "learning_rate": 4.657434543795975e-05, "loss": 0.059, "step": 10592 }, { "epoch": 0.21, "learning_rate": 4.657369847770252e-05, "loss": 0.0077, "step": 10594 }, { "epoch": 0.21, "learning_rate": 4.657305151744529e-05, "loss": 0.0024, "step": 10596 }, { "epoch": 0.21, "learning_rate": 4.657240455718806e-05, "loss": 0.0032, "step": 10598 }, { "epoch": 0.21, "learning_rate": 4.657175759693082e-05, "loss": 0.0043, "step": 10600 }, { "epoch": 0.21, "learning_rate": 4.6571110636673595e-05, "loss": 0.0474, "step": 10602 }, { "epoch": 0.21, "learning_rate": 4.657046367641636e-05, "loss": 0.0133, "step": 10604 }, { "epoch": 0.21, "learning_rate": 4.6569816716159126e-05, "loss": 0.0056, "step": 10606 }, { "epoch": 0.21, "learning_rate": 4.6569169755901895e-05, "loss": 0.0017, "step": 10608 }, { "epoch": 0.21, "learning_rate": 4.6568522795644664e-05, "loss": 0.0024, "step": 10610 }, { "epoch": 0.21, "learning_rate": 4.6567875835387434e-05, "loss": 0.0165, "step": 10612 }, { "epoch": 0.21, "learning_rate": 4.65672288751302e-05, "loss": 0.0106, "step": 10614 }, { "epoch": 0.21, "learning_rate": 4.656658191487297e-05, "loss": 0.0052, "step": 10616 }, { "epoch": 0.21, "learning_rate": 4.656593495461574e-05, "loss": 0.0017, "step": 10618 }, { "epoch": 0.21, "learning_rate": 4.656528799435851e-05, "loss": 0.0137, "step": 10620 }, { "epoch": 0.21, "learning_rate": 4.656464103410127e-05, "loss": 0.0142, "step": 10622 }, { "epoch": 0.21, "learning_rate": 4.656399407384405e-05, "loss": 0.0039, "step": 10624 }, { "epoch": 0.21, "learning_rate": 4.656334711358682e-05, "loss": 0.0066, "step": 10626 }, { "epoch": 0.21, "learning_rate": 4.6562700153329586e-05, "loss": 0.0074, "step": 10628 }, { "epoch": 0.21, "learning_rate": 4.6562053193072355e-05, "loss": 0.008, "step": 10630 }, { "epoch": 0.21, "learning_rate": 4.656140623281512e-05, "loss": 0.0026, "step": 10632 }, { "epoch": 0.21, "learning_rate": 4.6560759272557893e-05, "loss": 0.0165, "step": 10634 }, { "epoch": 0.21, "learning_rate": 4.6560112312300656e-05, "loss": 0.0461, "step": 10636 }, { "epoch": 0.21, "learning_rate": 4.6559465352043425e-05, "loss": 0.0115, "step": 10638 }, { "epoch": 0.21, "learning_rate": 4.6558818391786194e-05, "loss": 0.0068, "step": 10640 }, { "epoch": 0.21, "learning_rate": 4.655817143152896e-05, "loss": 0.0051, "step": 10642 }, { "epoch": 0.21, "learning_rate": 4.655752447127173e-05, "loss": 0.0097, "step": 10644 }, { "epoch": 0.21, "learning_rate": 4.65568775110145e-05, "loss": 0.0055, "step": 10646 }, { "epoch": 0.21, "learning_rate": 4.655623055075727e-05, "loss": 0.0108, "step": 10648 }, { "epoch": 0.21, "learning_rate": 4.655558359050004e-05, "loss": 0.014, "step": 10650 }, { "epoch": 0.21, "learning_rate": 4.655493663024281e-05, "loss": 0.005, "step": 10652 }, { "epoch": 0.21, "learning_rate": 4.655428966998557e-05, "loss": 0.0136, "step": 10654 }, { "epoch": 0.21, "learning_rate": 4.655364270972835e-05, "loss": 0.0073, "step": 10656 }, { "epoch": 0.21, "learning_rate": 4.655299574947111e-05, "loss": 0.0161, "step": 10658 }, { "epoch": 0.21, "learning_rate": 4.655234878921388e-05, "loss": 0.0025, "step": 10660 }, { "epoch": 0.21, "learning_rate": 4.6551701828956654e-05, "loss": 0.0217, "step": 10662 }, { "epoch": 0.21, "learning_rate": 4.6551054868699416e-05, "loss": 0.0206, "step": 10664 }, { "epoch": 0.21, "learning_rate": 4.6550407908442185e-05, "loss": 0.0012, "step": 10666 }, { "epoch": 0.21, "learning_rate": 4.6549760948184954e-05, "loss": 0.0043, "step": 10668 }, { "epoch": 0.21, "learning_rate": 4.6549113987927723e-05, "loss": 0.0026, "step": 10670 }, { "epoch": 0.21, "learning_rate": 4.654846702767049e-05, "loss": 0.0057, "step": 10672 }, { "epoch": 0.21, "learning_rate": 4.654782006741326e-05, "loss": 0.0086, "step": 10674 }, { "epoch": 0.21, "learning_rate": 4.654717310715603e-05, "loss": 0.0165, "step": 10676 }, { "epoch": 0.21, "learning_rate": 4.65465261468988e-05, "loss": 0.0023, "step": 10678 }, { "epoch": 0.21, "learning_rate": 4.654587918664157e-05, "loss": 0.0086, "step": 10680 }, { "epoch": 0.21, "learning_rate": 4.654523222638433e-05, "loss": 0.0068, "step": 10682 }, { "epoch": 0.21, "learning_rate": 4.654458526612711e-05, "loss": 0.0036, "step": 10684 }, { "epoch": 0.21, "learning_rate": 4.654393830586987e-05, "loss": 0.0051, "step": 10686 }, { "epoch": 0.21, "learning_rate": 4.654329134561264e-05, "loss": 0.0022, "step": 10688 }, { "epoch": 0.21, "learning_rate": 4.654264438535541e-05, "loss": 0.0142, "step": 10690 }, { "epoch": 0.21, "learning_rate": 4.6541997425098177e-05, "loss": 0.0047, "step": 10692 }, { "epoch": 0.21, "learning_rate": 4.654135046484095e-05, "loss": 0.0143, "step": 10694 }, { "epoch": 0.21, "learning_rate": 4.6540703504583715e-05, "loss": 0.0192, "step": 10696 }, { "epoch": 0.21, "learning_rate": 4.6540056544326484e-05, "loss": 0.0177, "step": 10698 }, { "epoch": 0.21, "learning_rate": 4.653940958406925e-05, "loss": 0.0018, "step": 10700 }, { "epoch": 0.21, "learning_rate": 4.653876262381202e-05, "loss": 0.0066, "step": 10702 }, { "epoch": 0.21, "learning_rate": 4.6538115663554784e-05, "loss": 0.0202, "step": 10704 }, { "epoch": 0.21, "learning_rate": 4.653746870329756e-05, "loss": 0.0071, "step": 10706 }, { "epoch": 0.21, "learning_rate": 4.653682174304033e-05, "loss": 0.0027, "step": 10708 }, { "epoch": 0.21, "learning_rate": 4.65361747827831e-05, "loss": 0.0168, "step": 10710 }, { "epoch": 0.21, "learning_rate": 4.653552782252587e-05, "loss": 0.0114, "step": 10712 }, { "epoch": 0.21, "learning_rate": 4.653488086226863e-05, "loss": 0.0116, "step": 10714 }, { "epoch": 0.21, "learning_rate": 4.6534233902011406e-05, "loss": 0.0476, "step": 10716 }, { "epoch": 0.21, "learning_rate": 4.653358694175417e-05, "loss": 0.0046, "step": 10718 }, { "epoch": 0.21, "learning_rate": 4.653293998149694e-05, "loss": 0.0145, "step": 10720 }, { "epoch": 0.21, "learning_rate": 4.6532293021239706e-05, "loss": 0.0021, "step": 10722 }, { "epoch": 0.21, "learning_rate": 4.6531646060982475e-05, "loss": 0.001, "step": 10724 }, { "epoch": 0.21, "learning_rate": 4.6530999100725244e-05, "loss": 0.001, "step": 10726 }, { "epoch": 0.21, "learning_rate": 4.653035214046801e-05, "loss": 0.0202, "step": 10728 }, { "epoch": 0.21, "learning_rate": 4.652970518021078e-05, "loss": 0.0044, "step": 10730 }, { "epoch": 0.21, "learning_rate": 4.652905821995355e-05, "loss": 0.0179, "step": 10732 }, { "epoch": 0.21, "learning_rate": 4.652841125969632e-05, "loss": 0.0252, "step": 10734 }, { "epoch": 0.21, "learning_rate": 4.652776429943908e-05, "loss": 0.0025, "step": 10736 }, { "epoch": 0.21, "learning_rate": 4.652711733918186e-05, "loss": 0.0045, "step": 10738 }, { "epoch": 0.21, "learning_rate": 4.652647037892463e-05, "loss": 0.0172, "step": 10740 }, { "epoch": 0.21, "learning_rate": 4.652582341866739e-05, "loss": 0.0057, "step": 10742 }, { "epoch": 0.21, "learning_rate": 4.6525176458410166e-05, "loss": 0.0112, "step": 10744 }, { "epoch": 0.21, "learning_rate": 4.652452949815293e-05, "loss": 0.0083, "step": 10746 }, { "epoch": 0.21, "learning_rate": 4.65238825378957e-05, "loss": 0.0022, "step": 10748 }, { "epoch": 0.21, "learning_rate": 4.6523235577638466e-05, "loss": 0.0172, "step": 10750 }, { "epoch": 0.21, "learning_rate": 4.6522588617381236e-05, "loss": 0.0129, "step": 10752 }, { "epoch": 0.21, "learning_rate": 4.6521941657124005e-05, "loss": 0.0165, "step": 10754 }, { "epoch": 0.21, "learning_rate": 4.6521294696866774e-05, "loss": 0.0104, "step": 10756 }, { "epoch": 0.21, "learning_rate": 4.652064773660954e-05, "loss": 0.0042, "step": 10758 }, { "epoch": 0.21, "learning_rate": 4.652000077635231e-05, "loss": 0.005, "step": 10760 }, { "epoch": 0.21, "learning_rate": 4.651935381609508e-05, "loss": 0.0292, "step": 10762 }, { "epoch": 0.21, "learning_rate": 4.651870685583784e-05, "loss": 0.0196, "step": 10764 }, { "epoch": 0.21, "learning_rate": 4.651805989558062e-05, "loss": 0.0055, "step": 10766 }, { "epoch": 0.21, "learning_rate": 4.651741293532338e-05, "loss": 0.0114, "step": 10768 }, { "epoch": 0.21, "learning_rate": 4.651676597506616e-05, "loss": 0.0064, "step": 10770 }, { "epoch": 0.21, "learning_rate": 4.6516119014808926e-05, "loss": 0.0201, "step": 10772 }, { "epoch": 0.21, "learning_rate": 4.651547205455169e-05, "loss": 0.0021, "step": 10774 }, { "epoch": 0.21, "learning_rate": 4.6514825094294465e-05, "loss": 0.0136, "step": 10776 }, { "epoch": 0.21, "learning_rate": 4.651417813403723e-05, "loss": 0.0065, "step": 10778 }, { "epoch": 0.21, "learning_rate": 4.6513531173779996e-05, "loss": 0.0032, "step": 10780 }, { "epoch": 0.21, "learning_rate": 4.6512884213522765e-05, "loss": 0.0029, "step": 10782 }, { "epoch": 0.21, "learning_rate": 4.6512237253265534e-05, "loss": 0.0094, "step": 10784 }, { "epoch": 0.21, "learning_rate": 4.65115902930083e-05, "loss": 0.0163, "step": 10786 }, { "epoch": 0.21, "learning_rate": 4.651094333275107e-05, "loss": 0.0248, "step": 10788 }, { "epoch": 0.21, "learning_rate": 4.651029637249384e-05, "loss": 0.0026, "step": 10790 }, { "epoch": 0.21, "learning_rate": 4.650964941223661e-05, "loss": 0.0181, "step": 10792 }, { "epoch": 0.21, "learning_rate": 4.650900245197938e-05, "loss": 0.0061, "step": 10794 }, { "epoch": 0.21, "learning_rate": 4.650835549172214e-05, "loss": 0.011, "step": 10796 }, { "epoch": 0.21, "learning_rate": 4.650770853146492e-05, "loss": 0.031, "step": 10798 }, { "epoch": 0.21, "learning_rate": 4.650706157120768e-05, "loss": 0.0079, "step": 10800 }, { "epoch": 0.21, "learning_rate": 4.650641461095045e-05, "loss": 0.0051, "step": 10802 }, { "epoch": 0.21, "learning_rate": 4.6505767650693225e-05, "loss": 0.023, "step": 10804 }, { "epoch": 0.21, "learning_rate": 4.650512069043599e-05, "loss": 0.0139, "step": 10806 }, { "epoch": 0.21, "learning_rate": 4.6504473730178756e-05, "loss": 0.0063, "step": 10808 }, { "epoch": 0.21, "learning_rate": 4.6503826769921525e-05, "loss": 0.0007, "step": 10810 }, { "epoch": 0.21, "learning_rate": 4.6503179809664295e-05, "loss": 0.01, "step": 10812 }, { "epoch": 0.21, "learning_rate": 4.6502532849407064e-05, "loss": 0.002, "step": 10814 }, { "epoch": 0.21, "learning_rate": 4.650188588914983e-05, "loss": 0.01, "step": 10816 }, { "epoch": 0.21, "learning_rate": 4.65012389288926e-05, "loss": 0.0032, "step": 10818 }, { "epoch": 0.21, "learning_rate": 4.650059196863537e-05, "loss": 0.0074, "step": 10820 }, { "epoch": 0.21, "learning_rate": 4.649994500837814e-05, "loss": 0.0198, "step": 10822 }, { "epoch": 0.21, "learning_rate": 4.64992980481209e-05, "loss": 0.0175, "step": 10824 }, { "epoch": 0.21, "learning_rate": 4.649865108786368e-05, "loss": 0.0016, "step": 10826 }, { "epoch": 0.21, "learning_rate": 4.649800412760644e-05, "loss": 0.0013, "step": 10828 }, { "epoch": 0.21, "learning_rate": 4.649735716734921e-05, "loss": 0.0047, "step": 10830 }, { "epoch": 0.21, "learning_rate": 4.649671020709198e-05, "loss": 0.0069, "step": 10832 }, { "epoch": 0.21, "learning_rate": 4.649606324683475e-05, "loss": 0.0106, "step": 10834 }, { "epoch": 0.21, "learning_rate": 4.649541628657752e-05, "loss": 0.0106, "step": 10836 }, { "epoch": 0.21, "learning_rate": 4.6494769326320286e-05, "loss": 0.0229, "step": 10838 }, { "epoch": 0.21, "learning_rate": 4.6494122366063055e-05, "loss": 0.0042, "step": 10840 }, { "epoch": 0.21, "learning_rate": 4.6493475405805824e-05, "loss": 0.0022, "step": 10842 }, { "epoch": 0.21, "learning_rate": 4.649282844554859e-05, "loss": 0.0134, "step": 10844 }, { "epoch": 0.21, "learning_rate": 4.6492181485291355e-05, "loss": 0.0016, "step": 10846 }, { "epoch": 0.21, "learning_rate": 4.649153452503413e-05, "loss": 0.0033, "step": 10848 }, { "epoch": 0.21, "learning_rate": 4.64908875647769e-05, "loss": 0.0127, "step": 10850 }, { "epoch": 0.21, "learning_rate": 4.649024060451967e-05, "loss": 0.0089, "step": 10852 }, { "epoch": 0.21, "learning_rate": 4.648959364426244e-05, "loss": 0.005, "step": 10854 }, { "epoch": 0.21, "learning_rate": 4.64889466840052e-05, "loss": 0.0074, "step": 10856 }, { "epoch": 0.21, "learning_rate": 4.648829972374798e-05, "loss": 0.0025, "step": 10858 }, { "epoch": 0.21, "learning_rate": 4.648765276349074e-05, "loss": 0.0035, "step": 10860 }, { "epoch": 0.21, "learning_rate": 4.648700580323351e-05, "loss": 0.0404, "step": 10862 }, { "epoch": 0.21, "learning_rate": 4.648635884297628e-05, "loss": 0.0075, "step": 10864 }, { "epoch": 0.21, "learning_rate": 4.6485711882719046e-05, "loss": 0.0024, "step": 10866 }, { "epoch": 0.21, "learning_rate": 4.6485064922461815e-05, "loss": 0.0162, "step": 10868 }, { "epoch": 0.21, "learning_rate": 4.6484417962204584e-05, "loss": 0.0012, "step": 10870 }, { "epoch": 0.21, "learning_rate": 4.6483771001947354e-05, "loss": 0.0188, "step": 10872 }, { "epoch": 0.21, "learning_rate": 4.648312404169012e-05, "loss": 0.0067, "step": 10874 }, { "epoch": 0.21, "learning_rate": 4.648247708143289e-05, "loss": 0.008, "step": 10876 }, { "epoch": 0.21, "learning_rate": 4.6481830121175654e-05, "loss": 0.0268, "step": 10878 }, { "epoch": 0.21, "learning_rate": 4.648118316091843e-05, "loss": 0.0093, "step": 10880 }, { "epoch": 0.21, "learning_rate": 4.648053620066119e-05, "loss": 0.0064, "step": 10882 }, { "epoch": 0.21, "learning_rate": 4.647988924040396e-05, "loss": 0.0215, "step": 10884 }, { "epoch": 0.21, "learning_rate": 4.647924228014674e-05, "loss": 0.0069, "step": 10886 }, { "epoch": 0.21, "learning_rate": 4.64785953198895e-05, "loss": 0.0046, "step": 10888 }, { "epoch": 0.21, "learning_rate": 4.647794835963227e-05, "loss": 0.0015, "step": 10890 }, { "epoch": 0.21, "learning_rate": 4.647730139937504e-05, "loss": 0.0033, "step": 10892 }, { "epoch": 0.21, "learning_rate": 4.647665443911781e-05, "loss": 0.004, "step": 10894 }, { "epoch": 0.21, "learning_rate": 4.6476007478860576e-05, "loss": 0.008, "step": 10896 }, { "epoch": 0.21, "learning_rate": 4.6475360518603345e-05, "loss": 0.0136, "step": 10898 }, { "epoch": 0.21, "learning_rate": 4.6474713558346114e-05, "loss": 0.01, "step": 10900 }, { "epoch": 0.21, "learning_rate": 4.647406659808888e-05, "loss": 0.0087, "step": 10902 }, { "epoch": 0.21, "learning_rate": 4.647341963783165e-05, "loss": 0.0179, "step": 10904 }, { "epoch": 0.21, "learning_rate": 4.6472772677574414e-05, "loss": 0.0075, "step": 10906 }, { "epoch": 0.21, "learning_rate": 4.647212571731719e-05, "loss": 0.0025, "step": 10908 }, { "epoch": 0.21, "learning_rate": 4.647147875705995e-05, "loss": 0.0049, "step": 10910 }, { "epoch": 0.21, "learning_rate": 4.647083179680273e-05, "loss": 0.0144, "step": 10912 }, { "epoch": 0.21, "learning_rate": 4.647018483654549e-05, "loss": 0.0088, "step": 10914 }, { "epoch": 0.21, "learning_rate": 4.646953787628826e-05, "loss": 0.0186, "step": 10916 }, { "epoch": 0.21, "learning_rate": 4.6468890916031036e-05, "loss": 0.0071, "step": 10918 }, { "epoch": 0.21, "learning_rate": 4.64682439557738e-05, "loss": 0.0254, "step": 10920 }, { "epoch": 0.21, "learning_rate": 4.646759699551657e-05, "loss": 0.0013, "step": 10922 }, { "epoch": 0.21, "learning_rate": 4.6466950035259336e-05, "loss": 0.0159, "step": 10924 }, { "epoch": 0.21, "learning_rate": 4.6466303075002105e-05, "loss": 0.0193, "step": 10926 }, { "epoch": 0.21, "learning_rate": 4.6465656114744874e-05, "loss": 0.0294, "step": 10928 }, { "epoch": 0.21, "learning_rate": 4.6465009154487643e-05, "loss": 0.0119, "step": 10930 }, { "epoch": 0.21, "learning_rate": 4.646436219423041e-05, "loss": 0.0215, "step": 10932 }, { "epoch": 0.21, "learning_rate": 4.646371523397318e-05, "loss": 0.0031, "step": 10934 }, { "epoch": 0.21, "learning_rate": 4.646306827371595e-05, "loss": 0.0106, "step": 10936 }, { "epoch": 0.21, "learning_rate": 4.646242131345871e-05, "loss": 0.0576, "step": 10938 }, { "epoch": 0.21, "learning_rate": 4.646177435320149e-05, "loss": 0.0026, "step": 10940 }, { "epoch": 0.21, "learning_rate": 4.646112739294425e-05, "loss": 0.0074, "step": 10942 }, { "epoch": 0.21, "learning_rate": 4.646048043268702e-05, "loss": 0.0082, "step": 10944 }, { "epoch": 0.21, "learning_rate": 4.645983347242979e-05, "loss": 0.0067, "step": 10946 }, { "epoch": 0.21, "learning_rate": 4.645918651217256e-05, "loss": 0.009, "step": 10948 }, { "epoch": 0.21, "learning_rate": 4.645853955191533e-05, "loss": 0.0019, "step": 10950 }, { "epoch": 0.21, "learning_rate": 4.6457892591658097e-05, "loss": 0.0058, "step": 10952 }, { "epoch": 0.21, "learning_rate": 4.6457245631400866e-05, "loss": 0.0067, "step": 10954 }, { "epoch": 0.21, "learning_rate": 4.6456598671143635e-05, "loss": 0.0015, "step": 10956 }, { "epoch": 0.21, "learning_rate": 4.6455951710886404e-05, "loss": 0.0023, "step": 10958 }, { "epoch": 0.21, "learning_rate": 4.6455304750629166e-05, "loss": 0.0116, "step": 10960 }, { "epoch": 0.21, "learning_rate": 4.645465779037194e-05, "loss": 0.0046, "step": 10962 }, { "epoch": 0.21, "learning_rate": 4.645401083011471e-05, "loss": 0.0103, "step": 10964 }, { "epoch": 0.21, "learning_rate": 4.6453363869857473e-05, "loss": 0.0015, "step": 10966 }, { "epoch": 0.21, "learning_rate": 4.645271690960025e-05, "loss": 0.0166, "step": 10968 }, { "epoch": 0.21, "learning_rate": 4.645206994934301e-05, "loss": 0.0123, "step": 10970 }, { "epoch": 0.21, "learning_rate": 4.645142298908578e-05, "loss": 0.0076, "step": 10972 }, { "epoch": 0.21, "learning_rate": 4.645077602882855e-05, "loss": 0.0129, "step": 10974 }, { "epoch": 0.21, "learning_rate": 4.645012906857132e-05, "loss": 0.0091, "step": 10976 }, { "epoch": 0.21, "learning_rate": 4.644948210831409e-05, "loss": 0.0144, "step": 10978 }, { "epoch": 0.21, "learning_rate": 4.644883514805686e-05, "loss": 0.0188, "step": 10980 }, { "epoch": 0.21, "learning_rate": 4.6448188187799626e-05, "loss": 0.0124, "step": 10982 }, { "epoch": 0.21, "learning_rate": 4.6447541227542395e-05, "loss": 0.0095, "step": 10984 }, { "epoch": 0.21, "learning_rate": 4.6446894267285164e-05, "loss": 0.0037, "step": 10986 }, { "epoch": 0.21, "learning_rate": 4.6446247307027927e-05, "loss": 0.0039, "step": 10988 }, { "epoch": 0.21, "learning_rate": 4.64456003467707e-05, "loss": 0.0057, "step": 10990 }, { "epoch": 0.21, "learning_rate": 4.6444953386513465e-05, "loss": 0.0019, "step": 10992 }, { "epoch": 0.21, "learning_rate": 4.644430642625624e-05, "loss": 0.0119, "step": 10994 }, { "epoch": 0.21, "learning_rate": 4.644365946599901e-05, "loss": 0.0004, "step": 10996 }, { "epoch": 0.21, "learning_rate": 4.644301250574177e-05, "loss": 0.0333, "step": 10998 }, { "epoch": 0.21, "learning_rate": 4.644236554548455e-05, "loss": 0.0011, "step": 11000 }, { "epoch": 0.21, "learning_rate": 4.644171858522731e-05, "loss": 0.0096, "step": 11002 }, { "epoch": 0.21, "learning_rate": 4.644107162497008e-05, "loss": 0.0012, "step": 11004 }, { "epoch": 0.21, "learning_rate": 4.644042466471285e-05, "loss": 0.0041, "step": 11006 }, { "epoch": 0.21, "learning_rate": 4.643977770445562e-05, "loss": 0.0106, "step": 11008 }, { "epoch": 0.21, "learning_rate": 4.6439130744198386e-05, "loss": 0.0051, "step": 11010 }, { "epoch": 0.21, "learning_rate": 4.6438483783941156e-05, "loss": 0.0194, "step": 11012 }, { "epoch": 0.21, "learning_rate": 4.6437836823683925e-05, "loss": 0.0027, "step": 11014 }, { "epoch": 0.21, "learning_rate": 4.6437189863426694e-05, "loss": 0.0229, "step": 11016 }, { "epoch": 0.21, "learning_rate": 4.643654290316946e-05, "loss": 0.0032, "step": 11018 }, { "epoch": 0.21, "learning_rate": 4.6435895942912225e-05, "loss": 0.0164, "step": 11020 }, { "epoch": 0.21, "learning_rate": 4.6435248982655e-05, "loss": 0.0054, "step": 11022 }, { "epoch": 0.21, "learning_rate": 4.643460202239776e-05, "loss": 0.0089, "step": 11024 }, { "epoch": 0.21, "learning_rate": 4.643395506214053e-05, "loss": 0.01, "step": 11026 }, { "epoch": 0.21, "learning_rate": 4.643330810188331e-05, "loss": 0.0015, "step": 11028 }, { "epoch": 0.21, "learning_rate": 4.643266114162607e-05, "loss": 0.0071, "step": 11030 }, { "epoch": 0.21, "learning_rate": 4.643201418136884e-05, "loss": 0.0019, "step": 11032 }, { "epoch": 0.21, "learning_rate": 4.643136722111161e-05, "loss": 0.0082, "step": 11034 }, { "epoch": 0.21, "learning_rate": 4.643072026085438e-05, "loss": 0.011, "step": 11036 }, { "epoch": 0.21, "learning_rate": 4.643007330059715e-05, "loss": 0.003, "step": 11038 }, { "epoch": 0.21, "learning_rate": 4.6429426340339916e-05, "loss": 0.0033, "step": 11040 }, { "epoch": 0.21, "learning_rate": 4.6428779380082685e-05, "loss": 0.0127, "step": 11042 }, { "epoch": 0.21, "learning_rate": 4.6428132419825454e-05, "loss": 0.0134, "step": 11044 }, { "epoch": 0.21, "learning_rate": 4.642748545956822e-05, "loss": 0.0024, "step": 11046 }, { "epoch": 0.21, "learning_rate": 4.6426838499310986e-05, "loss": 0.0038, "step": 11048 }, { "epoch": 0.21, "learning_rate": 4.642619153905376e-05, "loss": 0.0149, "step": 11050 }, { "epoch": 0.21, "learning_rate": 4.6425544578796524e-05, "loss": 0.004, "step": 11052 }, { "epoch": 0.21, "learning_rate": 4.64248976185393e-05, "loss": 0.007, "step": 11054 }, { "epoch": 0.21, "learning_rate": 4.642425065828206e-05, "loss": 0.0175, "step": 11056 }, { "epoch": 0.21, "learning_rate": 4.642360369802483e-05, "loss": 0.0252, "step": 11058 }, { "epoch": 0.21, "learning_rate": 4.64229567377676e-05, "loss": 0.0034, "step": 11060 }, { "epoch": 0.21, "learning_rate": 4.642230977751037e-05, "loss": 0.0051, "step": 11062 }, { "epoch": 0.21, "learning_rate": 4.642166281725314e-05, "loss": 0.0166, "step": 11064 }, { "epoch": 0.21, "learning_rate": 4.642101585699591e-05, "loss": 0.0146, "step": 11066 }, { "epoch": 0.21, "learning_rate": 4.6420368896738676e-05, "loss": 0.0094, "step": 11068 }, { "epoch": 0.21, "learning_rate": 4.641972193648144e-05, "loss": 0.0043, "step": 11070 }, { "epoch": 0.21, "learning_rate": 4.6419074976224215e-05, "loss": 0.0028, "step": 11072 }, { "epoch": 0.21, "learning_rate": 4.6418428015966984e-05, "loss": 0.0117, "step": 11074 }, { "epoch": 0.21, "learning_rate": 4.641778105570975e-05, "loss": 0.0079, "step": 11076 }, { "epoch": 0.22, "learning_rate": 4.641713409545252e-05, "loss": 0.001, "step": 11078 }, { "epoch": 0.22, "learning_rate": 4.6416487135195284e-05, "loss": 0.0025, "step": 11080 }, { "epoch": 0.22, "learning_rate": 4.641584017493806e-05, "loss": 0.0028, "step": 11082 }, { "epoch": 0.22, "learning_rate": 4.641519321468082e-05, "loss": 0.0036, "step": 11084 }, { "epoch": 0.22, "learning_rate": 4.641454625442359e-05, "loss": 0.0017, "step": 11086 }, { "epoch": 0.22, "learning_rate": 4.641389929416636e-05, "loss": 0.0067, "step": 11088 }, { "epoch": 0.22, "learning_rate": 4.641325233390913e-05, "loss": 0.0054, "step": 11090 }, { "epoch": 0.22, "learning_rate": 4.64126053736519e-05, "loss": 0.0057, "step": 11092 }, { "epoch": 0.22, "learning_rate": 4.641195841339467e-05, "loss": 0.0052, "step": 11094 }, { "epoch": 0.22, "learning_rate": 4.641131145313744e-05, "loss": 0.0028, "step": 11096 }, { "epoch": 0.22, "learning_rate": 4.6410664492880206e-05, "loss": 0.0451, "step": 11098 }, { "epoch": 0.22, "learning_rate": 4.6410017532622975e-05, "loss": 0.0022, "step": 11100 }, { "epoch": 0.22, "learning_rate": 4.640937057236574e-05, "loss": 0.003, "step": 11102 }, { "epoch": 0.22, "learning_rate": 4.640872361210851e-05, "loss": 0.0021, "step": 11104 }, { "epoch": 0.22, "learning_rate": 4.640807665185128e-05, "loss": 0.0078, "step": 11106 }, { "epoch": 0.22, "learning_rate": 4.6407429691594044e-05, "loss": 0.0037, "step": 11108 }, { "epoch": 0.22, "learning_rate": 4.640678273133682e-05, "loss": 0.0057, "step": 11110 }, { "epoch": 0.22, "learning_rate": 4.640613577107958e-05, "loss": 0.0063, "step": 11112 }, { "epoch": 0.22, "learning_rate": 4.640548881082235e-05, "loss": 0.0047, "step": 11114 }, { "epoch": 0.22, "learning_rate": 4.640484185056512e-05, "loss": 0.0067, "step": 11116 }, { "epoch": 0.22, "learning_rate": 4.640419489030789e-05, "loss": 0.0013, "step": 11118 }, { "epoch": 0.22, "learning_rate": 4.640354793005066e-05, "loss": 0.0006, "step": 11120 }, { "epoch": 0.22, "learning_rate": 4.640290096979343e-05, "loss": 0.0013, "step": 11122 }, { "epoch": 0.22, "learning_rate": 4.64022540095362e-05, "loss": 0.0058, "step": 11124 }, { "epoch": 0.22, "learning_rate": 4.6401607049278966e-05, "loss": 0.0026, "step": 11126 }, { "epoch": 0.22, "learning_rate": 4.6400960089021735e-05, "loss": 0.0041, "step": 11128 }, { "epoch": 0.22, "learning_rate": 4.64003131287645e-05, "loss": 0.0032, "step": 11130 }, { "epoch": 0.22, "learning_rate": 4.6399666168507274e-05, "loss": 0.0067, "step": 11132 }, { "epoch": 0.22, "learning_rate": 4.6399019208250036e-05, "loss": 0.0013, "step": 11134 }, { "epoch": 0.22, "learning_rate": 4.639837224799281e-05, "loss": 0.001, "step": 11136 }, { "epoch": 0.22, "learning_rate": 4.6397725287735574e-05, "loss": 0.0079, "step": 11138 }, { "epoch": 0.22, "learning_rate": 4.639707832747834e-05, "loss": 0.001, "step": 11140 }, { "epoch": 0.22, "learning_rate": 4.639643136722112e-05, "loss": 0.0012, "step": 11142 }, { "epoch": 0.22, "learning_rate": 4.639578440696388e-05, "loss": 0.0089, "step": 11144 }, { "epoch": 0.22, "learning_rate": 4.639513744670665e-05, "loss": 0.0019, "step": 11146 }, { "epoch": 0.22, "learning_rate": 4.639449048644942e-05, "loss": 0.0208, "step": 11148 }, { "epoch": 0.22, "learning_rate": 4.639384352619219e-05, "loss": 0.0081, "step": 11150 }, { "epoch": 0.22, "learning_rate": 4.639319656593496e-05, "loss": 0.0129, "step": 11152 }, { "epoch": 0.22, "learning_rate": 4.639254960567773e-05, "loss": 0.0013, "step": 11154 }, { "epoch": 0.22, "learning_rate": 4.6391902645420496e-05, "loss": 0.0019, "step": 11156 }, { "epoch": 0.22, "learning_rate": 4.6391255685163265e-05, "loss": 0.0175, "step": 11158 }, { "epoch": 0.22, "learning_rate": 4.6390608724906034e-05, "loss": 0.0024, "step": 11160 }, { "epoch": 0.22, "learning_rate": 4.6389961764648796e-05, "loss": 0.057, "step": 11162 }, { "epoch": 0.22, "learning_rate": 4.638931480439157e-05, "loss": 0.0011, "step": 11164 }, { "epoch": 0.22, "learning_rate": 4.6388667844134334e-05, "loss": 0.0023, "step": 11166 }, { "epoch": 0.22, "learning_rate": 4.6388020883877103e-05, "loss": 0.0012, "step": 11168 }, { "epoch": 0.22, "learning_rate": 4.638737392361987e-05, "loss": 0.0052, "step": 11170 }, { "epoch": 0.22, "learning_rate": 4.638672696336264e-05, "loss": 0.0196, "step": 11172 }, { "epoch": 0.22, "learning_rate": 4.638608000310541e-05, "loss": 0.0041, "step": 11174 }, { "epoch": 0.22, "learning_rate": 4.638543304284818e-05, "loss": 0.0045, "step": 11176 }, { "epoch": 0.22, "learning_rate": 4.638478608259095e-05, "loss": 0.0045, "step": 11178 }, { "epoch": 0.22, "learning_rate": 4.638413912233372e-05, "loss": 0.0105, "step": 11180 }, { "epoch": 0.22, "learning_rate": 4.638349216207649e-05, "loss": 0.0188, "step": 11182 }, { "epoch": 0.22, "learning_rate": 4.638284520181925e-05, "loss": 0.0025, "step": 11184 }, { "epoch": 0.22, "learning_rate": 4.6382198241562025e-05, "loss": 0.0122, "step": 11186 }, { "epoch": 0.22, "learning_rate": 4.6381551281304794e-05, "loss": 0.0029, "step": 11188 }, { "epoch": 0.22, "learning_rate": 4.638090432104756e-05, "loss": 0.007, "step": 11190 }, { "epoch": 0.22, "learning_rate": 4.638025736079033e-05, "loss": 0.0051, "step": 11192 }, { "epoch": 0.22, "learning_rate": 4.6379610400533095e-05, "loss": 0.019, "step": 11194 }, { "epoch": 0.22, "learning_rate": 4.637896344027587e-05, "loss": 0.0217, "step": 11196 }, { "epoch": 0.22, "learning_rate": 4.637831648001863e-05, "loss": 0.0039, "step": 11198 }, { "epoch": 0.22, "learning_rate": 4.63776695197614e-05, "loss": 0.0271, "step": 11200 }, { "epoch": 0.22, "learning_rate": 4.637702255950417e-05, "loss": 0.0018, "step": 11202 }, { "epoch": 0.22, "learning_rate": 4.637637559924694e-05, "loss": 0.0144, "step": 11204 }, { "epoch": 0.22, "learning_rate": 4.637572863898971e-05, "loss": 0.0016, "step": 11206 }, { "epoch": 0.22, "learning_rate": 4.637508167873248e-05, "loss": 0.0298, "step": 11208 }, { "epoch": 0.22, "learning_rate": 4.637443471847525e-05, "loss": 0.0135, "step": 11210 }, { "epoch": 0.22, "learning_rate": 4.637378775821801e-05, "loss": 0.0214, "step": 11212 }, { "epoch": 0.22, "learning_rate": 4.6373140797960786e-05, "loss": 0.0016, "step": 11214 }, { "epoch": 0.22, "learning_rate": 4.637249383770355e-05, "loss": 0.0043, "step": 11216 }, { "epoch": 0.22, "learning_rate": 4.6371846877446324e-05, "loss": 0.0007, "step": 11218 }, { "epoch": 0.22, "learning_rate": 4.637119991718909e-05, "loss": 0.0146, "step": 11220 }, { "epoch": 0.22, "learning_rate": 4.6370552956931855e-05, "loss": 0.0081, "step": 11222 }, { "epoch": 0.22, "learning_rate": 4.636990599667463e-05, "loss": 0.0171, "step": 11224 }, { "epoch": 0.22, "learning_rate": 4.636925903641739e-05, "loss": 0.0017, "step": 11226 }, { "epoch": 0.22, "learning_rate": 4.636861207616016e-05, "loss": 0.0218, "step": 11228 }, { "epoch": 0.22, "learning_rate": 4.636796511590293e-05, "loss": 0.0209, "step": 11230 }, { "epoch": 0.22, "learning_rate": 4.63673181556457e-05, "loss": 0.0009, "step": 11232 }, { "epoch": 0.22, "learning_rate": 4.636667119538847e-05, "loss": 0.012, "step": 11234 }, { "epoch": 0.22, "learning_rate": 4.636602423513124e-05, "loss": 0.0045, "step": 11236 }, { "epoch": 0.22, "learning_rate": 4.636537727487401e-05, "loss": 0.0065, "step": 11238 }, { "epoch": 0.22, "learning_rate": 4.636473031461678e-05, "loss": 0.0316, "step": 11240 }, { "epoch": 0.22, "learning_rate": 4.6364083354359546e-05, "loss": 0.0084, "step": 11242 }, { "epoch": 0.22, "learning_rate": 4.636343639410231e-05, "loss": 0.031, "step": 11244 }, { "epoch": 0.22, "learning_rate": 4.6362789433845084e-05, "loss": 0.0018, "step": 11246 }, { "epoch": 0.22, "learning_rate": 4.6362142473587847e-05, "loss": 0.0064, "step": 11248 }, { "epoch": 0.22, "learning_rate": 4.6361495513330616e-05, "loss": 0.0065, "step": 11250 }, { "epoch": 0.22, "learning_rate": 4.636084855307339e-05, "loss": 0.0019, "step": 11252 }, { "epoch": 0.22, "learning_rate": 4.6360201592816154e-05, "loss": 0.002, "step": 11254 }, { "epoch": 0.22, "learning_rate": 4.635955463255892e-05, "loss": 0.0255, "step": 11256 }, { "epoch": 0.22, "learning_rate": 4.635890767230169e-05, "loss": 0.0069, "step": 11258 }, { "epoch": 0.22, "learning_rate": 4.635826071204446e-05, "loss": 0.0303, "step": 11260 }, { "epoch": 0.22, "learning_rate": 4.635761375178723e-05, "loss": 0.0243, "step": 11262 }, { "epoch": 0.22, "learning_rate": 4.635696679153e-05, "loss": 0.0102, "step": 11264 }, { "epoch": 0.22, "learning_rate": 4.635631983127277e-05, "loss": 0.004, "step": 11266 }, { "epoch": 0.22, "learning_rate": 4.635567287101554e-05, "loss": 0.0319, "step": 11268 }, { "epoch": 0.22, "learning_rate": 4.6355025910758306e-05, "loss": 0.0119, "step": 11270 }, { "epoch": 0.22, "learning_rate": 4.635437895050107e-05, "loss": 0.0081, "step": 11272 }, { "epoch": 0.22, "learning_rate": 4.6353731990243845e-05, "loss": 0.0011, "step": 11274 }, { "epoch": 0.22, "learning_rate": 4.635308502998661e-05, "loss": 0.0046, "step": 11276 }, { "epoch": 0.22, "learning_rate": 4.635243806972938e-05, "loss": 0.0027, "step": 11278 }, { "epoch": 0.22, "learning_rate": 4.6351791109472145e-05, "loss": 0.0015, "step": 11280 }, { "epoch": 0.22, "learning_rate": 4.6351144149214914e-05, "loss": 0.0048, "step": 11282 }, { "epoch": 0.22, "learning_rate": 4.635049718895769e-05, "loss": 0.0104, "step": 11284 }, { "epoch": 0.22, "learning_rate": 4.634985022870045e-05, "loss": 0.0248, "step": 11286 }, { "epoch": 0.22, "learning_rate": 4.634920326844322e-05, "loss": 0.0027, "step": 11288 }, { "epoch": 0.22, "learning_rate": 4.634855630818599e-05, "loss": 0.0013, "step": 11290 }, { "epoch": 0.22, "learning_rate": 4.634790934792876e-05, "loss": 0.0086, "step": 11292 }, { "epoch": 0.22, "learning_rate": 4.634726238767152e-05, "loss": 0.2262, "step": 11294 }, { "epoch": 0.22, "learning_rate": 4.63466154274143e-05, "loss": 0.0052, "step": 11296 }, { "epoch": 0.22, "learning_rate": 4.634596846715707e-05, "loss": 0.0096, "step": 11298 }, { "epoch": 0.22, "learning_rate": 4.6345321506899836e-05, "loss": 0.0147, "step": 11300 }, { "epoch": 0.22, "learning_rate": 4.6344674546642605e-05, "loss": 0.0054, "step": 11302 }, { "epoch": 0.22, "learning_rate": 4.634402758638537e-05, "loss": 0.003, "step": 11304 }, { "epoch": 0.22, "learning_rate": 4.634338062612814e-05, "loss": 0.0111, "step": 11306 }, { "epoch": 0.22, "learning_rate": 4.6342733665870906e-05, "loss": 0.0109, "step": 11308 }, { "epoch": 0.22, "learning_rate": 4.6342086705613675e-05, "loss": 0.0116, "step": 11310 }, { "epoch": 0.22, "learning_rate": 4.6341439745356444e-05, "loss": 0.0182, "step": 11312 }, { "epoch": 0.22, "learning_rate": 4.634079278509921e-05, "loss": 0.012, "step": 11314 }, { "epoch": 0.22, "learning_rate": 4.634014582484198e-05, "loss": 0.0128, "step": 11316 }, { "epoch": 0.22, "learning_rate": 4.633949886458475e-05, "loss": 0.0356, "step": 11318 }, { "epoch": 0.22, "learning_rate": 4.633885190432752e-05, "loss": 0.0233, "step": 11320 }, { "epoch": 0.22, "learning_rate": 4.633820494407029e-05, "loss": 0.0023, "step": 11322 }, { "epoch": 0.22, "learning_rate": 4.633755798381306e-05, "loss": 0.0009, "step": 11324 }, { "epoch": 0.22, "learning_rate": 4.633691102355582e-05, "loss": 0.02, "step": 11326 }, { "epoch": 0.22, "learning_rate": 4.6336264063298596e-05, "loss": 0.0139, "step": 11328 }, { "epoch": 0.22, "learning_rate": 4.6335617103041365e-05, "loss": 0.0011, "step": 11330 }, { "epoch": 0.22, "learning_rate": 4.633497014278413e-05, "loss": 0.0006, "step": 11332 }, { "epoch": 0.22, "learning_rate": 4.6334323182526904e-05, "loss": 0.008, "step": 11334 }, { "epoch": 0.22, "learning_rate": 4.6333676222269666e-05, "loss": 0.0026, "step": 11336 }, { "epoch": 0.22, "learning_rate": 4.633302926201244e-05, "loss": 0.0103, "step": 11338 }, { "epoch": 0.22, "learning_rate": 4.6332382301755204e-05, "loss": 0.0336, "step": 11340 }, { "epoch": 0.22, "learning_rate": 4.633173534149797e-05, "loss": 0.0239, "step": 11342 }, { "epoch": 0.22, "learning_rate": 4.633108838124074e-05, "loss": 0.0129, "step": 11344 }, { "epoch": 0.22, "learning_rate": 4.633044142098351e-05, "loss": 0.001, "step": 11346 }, { "epoch": 0.22, "learning_rate": 4.632979446072628e-05, "loss": 0.0023, "step": 11348 }, { "epoch": 0.22, "learning_rate": 4.632914750046905e-05, "loss": 0.0072, "step": 11350 }, { "epoch": 0.22, "learning_rate": 4.632850054021182e-05, "loss": 0.0087, "step": 11352 }, { "epoch": 0.22, "learning_rate": 4.632785357995458e-05, "loss": 0.0031, "step": 11354 }, { "epoch": 0.22, "learning_rate": 4.632720661969736e-05, "loss": 0.0078, "step": 11356 }, { "epoch": 0.22, "learning_rate": 4.632655965944012e-05, "loss": 0.0032, "step": 11358 }, { "epoch": 0.22, "learning_rate": 4.6325912699182895e-05, "loss": 0.0043, "step": 11360 }, { "epoch": 0.22, "learning_rate": 4.632526573892566e-05, "loss": 0.0184, "step": 11362 }, { "epoch": 0.22, "learning_rate": 4.6324618778668426e-05, "loss": 0.0048, "step": 11364 }, { "epoch": 0.22, "learning_rate": 4.63239718184112e-05, "loss": 0.0011, "step": 11366 }, { "epoch": 0.22, "learning_rate": 4.6323324858153964e-05, "loss": 0.0066, "step": 11368 }, { "epoch": 0.22, "learning_rate": 4.6322677897896734e-05, "loss": 0.0053, "step": 11370 }, { "epoch": 0.22, "learning_rate": 4.63220309376395e-05, "loss": 0.0011, "step": 11372 }, { "epoch": 0.22, "learning_rate": 4.632138397738227e-05, "loss": 0.0206, "step": 11374 }, { "epoch": 0.22, "learning_rate": 4.632073701712504e-05, "loss": 0.0017, "step": 11376 }, { "epoch": 0.22, "learning_rate": 4.632009005686781e-05, "loss": 0.0094, "step": 11378 }, { "epoch": 0.22, "learning_rate": 4.631944309661058e-05, "loss": 0.001, "step": 11380 }, { "epoch": 0.22, "learning_rate": 4.631879613635335e-05, "loss": 0.0094, "step": 11382 }, { "epoch": 0.22, "learning_rate": 4.631814917609612e-05, "loss": 0.0016, "step": 11384 }, { "epoch": 0.22, "learning_rate": 4.631750221583888e-05, "loss": 0.0234, "step": 11386 }, { "epoch": 0.22, "learning_rate": 4.6316855255581655e-05, "loss": 0.0018, "step": 11388 }, { "epoch": 0.22, "learning_rate": 4.631620829532442e-05, "loss": 0.0051, "step": 11390 }, { "epoch": 0.22, "learning_rate": 4.631556133506719e-05, "loss": 0.0029, "step": 11392 }, { "epoch": 0.22, "learning_rate": 4.6314914374809956e-05, "loss": 0.0186, "step": 11394 }, { "epoch": 0.22, "learning_rate": 4.6314267414552725e-05, "loss": 0.0009, "step": 11396 }, { "epoch": 0.22, "learning_rate": 4.6313620454295494e-05, "loss": 0.0025, "step": 11398 }, { "epoch": 0.22, "learning_rate": 4.631297349403826e-05, "loss": 0.0045, "step": 11400 }, { "epoch": 0.22, "learning_rate": 4.631232653378103e-05, "loss": 0.0144, "step": 11402 }, { "epoch": 0.22, "learning_rate": 4.63116795735238e-05, "loss": 0.0011, "step": 11404 }, { "epoch": 0.22, "learning_rate": 4.631103261326657e-05, "loss": 0.018, "step": 11406 }, { "epoch": 0.22, "learning_rate": 4.631038565300934e-05, "loss": 0.0065, "step": 11408 }, { "epoch": 0.22, "learning_rate": 4.630973869275211e-05, "loss": 0.0026, "step": 11410 }, { "epoch": 0.22, "learning_rate": 4.630909173249488e-05, "loss": 0.0038, "step": 11412 }, { "epoch": 0.22, "learning_rate": 4.630844477223764e-05, "loss": 0.0147, "step": 11414 }, { "epoch": 0.22, "learning_rate": 4.6307797811980416e-05, "loss": 0.001, "step": 11416 }, { "epoch": 0.22, "learning_rate": 4.630715085172318e-05, "loss": 0.0008, "step": 11418 }, { "epoch": 0.22, "learning_rate": 4.6306503891465954e-05, "loss": 0.0016, "step": 11420 }, { "epoch": 0.22, "learning_rate": 4.6305856931208716e-05, "loss": 0.0067, "step": 11422 }, { "epoch": 0.22, "learning_rate": 4.6305209970951485e-05, "loss": 0.0076, "step": 11424 }, { "epoch": 0.22, "learning_rate": 4.6304563010694254e-05, "loss": 0.0029, "step": 11426 }, { "epoch": 0.22, "learning_rate": 4.6303916050437023e-05, "loss": 0.0257, "step": 11428 }, { "epoch": 0.22, "learning_rate": 4.630326909017979e-05, "loss": 0.0113, "step": 11430 }, { "epoch": 0.22, "learning_rate": 4.630262212992256e-05, "loss": 0.0028, "step": 11432 }, { "epoch": 0.22, "learning_rate": 4.630197516966533e-05, "loss": 0.0046, "step": 11434 }, { "epoch": 0.22, "learning_rate": 4.630132820940809e-05, "loss": 0.0021, "step": 11436 }, { "epoch": 0.22, "learning_rate": 4.630068124915087e-05, "loss": 0.0016, "step": 11438 }, { "epoch": 0.22, "learning_rate": 4.630003428889363e-05, "loss": 0.0035, "step": 11440 }, { "epoch": 0.22, "learning_rate": 4.629938732863641e-05, "loss": 0.0017, "step": 11442 }, { "epoch": 0.22, "learning_rate": 4.6298740368379176e-05, "loss": 0.0254, "step": 11444 }, { "epoch": 0.22, "learning_rate": 4.629809340812194e-05, "loss": 0.004, "step": 11446 }, { "epoch": 0.22, "learning_rate": 4.6297446447864714e-05, "loss": 0.0022, "step": 11448 }, { "epoch": 0.22, "learning_rate": 4.6296799487607477e-05, "loss": 0.0059, "step": 11450 }, { "epoch": 0.22, "learning_rate": 4.6296152527350246e-05, "loss": 0.0197, "step": 11452 }, { "epoch": 0.22, "learning_rate": 4.6295505567093015e-05, "loss": 0.0115, "step": 11454 }, { "epoch": 0.22, "learning_rate": 4.6294858606835784e-05, "loss": 0.0011, "step": 11456 }, { "epoch": 0.22, "learning_rate": 4.629421164657855e-05, "loss": 0.0376, "step": 11458 }, { "epoch": 0.22, "learning_rate": 4.629356468632132e-05, "loss": 0.0049, "step": 11460 }, { "epoch": 0.22, "learning_rate": 4.629291772606409e-05, "loss": 0.019, "step": 11462 }, { "epoch": 0.22, "learning_rate": 4.629227076580686e-05, "loss": 0.0087, "step": 11464 }, { "epoch": 0.22, "learning_rate": 4.629162380554963e-05, "loss": 0.0121, "step": 11466 }, { "epoch": 0.22, "learning_rate": 4.629097684529239e-05, "loss": 0.0042, "step": 11468 }, { "epoch": 0.22, "learning_rate": 4.629032988503517e-05, "loss": 0.0143, "step": 11470 }, { "epoch": 0.22, "learning_rate": 4.628968292477793e-05, "loss": 0.0154, "step": 11472 }, { "epoch": 0.22, "learning_rate": 4.62890359645207e-05, "loss": 0.0023, "step": 11474 }, { "epoch": 0.22, "learning_rate": 4.6288389004263475e-05, "loss": 0.01, "step": 11476 }, { "epoch": 0.22, "learning_rate": 4.628774204400624e-05, "loss": 0.0013, "step": 11478 }, { "epoch": 0.22, "learning_rate": 4.628709508374901e-05, "loss": 0.0096, "step": 11480 }, { "epoch": 0.22, "learning_rate": 4.6286448123491775e-05, "loss": 0.0088, "step": 11482 }, { "epoch": 0.22, "learning_rate": 4.6285801163234544e-05, "loss": 0.0028, "step": 11484 }, { "epoch": 0.22, "learning_rate": 4.628515420297731e-05, "loss": 0.0175, "step": 11486 }, { "epoch": 0.22, "learning_rate": 4.628450724272008e-05, "loss": 0.003, "step": 11488 }, { "epoch": 0.22, "learning_rate": 4.628386028246285e-05, "loss": 0.0012, "step": 11490 }, { "epoch": 0.22, "learning_rate": 4.628321332220562e-05, "loss": 0.0213, "step": 11492 }, { "epoch": 0.22, "learning_rate": 4.628256636194839e-05, "loss": 0.0104, "step": 11494 }, { "epoch": 0.22, "learning_rate": 4.628191940169115e-05, "loss": 0.0093, "step": 11496 }, { "epoch": 0.22, "learning_rate": 4.628127244143393e-05, "loss": 0.014, "step": 11498 }, { "epoch": 0.22, "learning_rate": 4.628062548117669e-05, "loss": 0.0066, "step": 11500 }, { "epoch": 0.22, "learning_rate": 4.6279978520919466e-05, "loss": 0.0108, "step": 11502 }, { "epoch": 0.22, "learning_rate": 4.627933156066223e-05, "loss": 0.0109, "step": 11504 }, { "epoch": 0.22, "learning_rate": 4.6278684600405e-05, "loss": 0.0007, "step": 11506 }, { "epoch": 0.22, "learning_rate": 4.627803764014777e-05, "loss": 0.0154, "step": 11508 }, { "epoch": 0.22, "learning_rate": 4.6277390679890536e-05, "loss": 0.0141, "step": 11510 }, { "epoch": 0.22, "learning_rate": 4.6276743719633305e-05, "loss": 0.0069, "step": 11512 }, { "epoch": 0.22, "learning_rate": 4.6276096759376074e-05, "loss": 0.0032, "step": 11514 }, { "epoch": 0.22, "learning_rate": 4.627544979911884e-05, "loss": 0.0102, "step": 11516 }, { "epoch": 0.22, "learning_rate": 4.6274802838861605e-05, "loss": 0.0005, "step": 11518 }, { "epoch": 0.22, "learning_rate": 4.627415587860438e-05, "loss": 0.0138, "step": 11520 }, { "epoch": 0.22, "learning_rate": 4.627350891834715e-05, "loss": 0.0017, "step": 11522 }, { "epoch": 0.22, "learning_rate": 4.627286195808992e-05, "loss": 0.0027, "step": 11524 }, { "epoch": 0.22, "learning_rate": 4.627221499783269e-05, "loss": 0.0046, "step": 11526 }, { "epoch": 0.22, "learning_rate": 4.627156803757545e-05, "loss": 0.0016, "step": 11528 }, { "epoch": 0.22, "learning_rate": 4.6270921077318226e-05, "loss": 0.0006, "step": 11530 }, { "epoch": 0.22, "learning_rate": 4.627027411706099e-05, "loss": 0.0096, "step": 11532 }, { "epoch": 0.22, "learning_rate": 4.626962715680376e-05, "loss": 0.0159, "step": 11534 }, { "epoch": 0.22, "learning_rate": 4.626898019654653e-05, "loss": 0.0074, "step": 11536 }, { "epoch": 0.22, "learning_rate": 4.6268333236289296e-05, "loss": 0.0134, "step": 11538 }, { "epoch": 0.22, "learning_rate": 4.6267686276032065e-05, "loss": 0.0084, "step": 11540 }, { "epoch": 0.22, "learning_rate": 4.6267039315774834e-05, "loss": 0.0225, "step": 11542 }, { "epoch": 0.22, "learning_rate": 4.62663923555176e-05, "loss": 0.0417, "step": 11544 }, { "epoch": 0.22, "learning_rate": 4.626574539526037e-05, "loss": 0.0285, "step": 11546 }, { "epoch": 0.22, "learning_rate": 4.626509843500314e-05, "loss": 0.0193, "step": 11548 }, { "epoch": 0.22, "learning_rate": 4.6264451474745904e-05, "loss": 0.0017, "step": 11550 }, { "epoch": 0.22, "learning_rate": 4.626380451448868e-05, "loss": 0.0056, "step": 11552 }, { "epoch": 0.22, "learning_rate": 4.626315755423145e-05, "loss": 0.0131, "step": 11554 }, { "epoch": 0.22, "learning_rate": 4.626251059397421e-05, "loss": 0.0063, "step": 11556 }, { "epoch": 0.22, "learning_rate": 4.626186363371699e-05, "loss": 0.0026, "step": 11558 }, { "epoch": 0.22, "learning_rate": 4.626121667345975e-05, "loss": 0.0036, "step": 11560 }, { "epoch": 0.22, "learning_rate": 4.6260569713202525e-05, "loss": 0.002, "step": 11562 }, { "epoch": 0.22, "learning_rate": 4.625992275294529e-05, "loss": 0.0009, "step": 11564 }, { "epoch": 0.22, "learning_rate": 4.6259275792688056e-05, "loss": 0.0009, "step": 11566 }, { "epoch": 0.22, "learning_rate": 4.6258628832430825e-05, "loss": 0.0177, "step": 11568 }, { "epoch": 0.22, "learning_rate": 4.6257981872173595e-05, "loss": 0.0039, "step": 11570 }, { "epoch": 0.22, "learning_rate": 4.6257334911916364e-05, "loss": 0.0272, "step": 11572 }, { "epoch": 0.22, "learning_rate": 4.625668795165913e-05, "loss": 0.0172, "step": 11574 }, { "epoch": 0.22, "learning_rate": 4.62560409914019e-05, "loss": 0.005, "step": 11576 }, { "epoch": 0.22, "learning_rate": 4.6255394031144664e-05, "loss": 0.0132, "step": 11578 }, { "epoch": 0.22, "learning_rate": 4.625474707088744e-05, "loss": 0.0009, "step": 11580 }, { "epoch": 0.22, "learning_rate": 4.62541001106302e-05, "loss": 0.002, "step": 11582 }, { "epoch": 0.22, "learning_rate": 4.625345315037298e-05, "loss": 0.0014, "step": 11584 }, { "epoch": 0.22, "learning_rate": 4.625280619011575e-05, "loss": 0.0038, "step": 11586 }, { "epoch": 0.22, "learning_rate": 4.625215922985851e-05, "loss": 0.0011, "step": 11588 }, { "epoch": 0.22, "learning_rate": 4.6251512269601285e-05, "loss": 0.0008, "step": 11590 }, { "epoch": 0.22, "learning_rate": 4.625086530934405e-05, "loss": 0.0005, "step": 11592 }, { "epoch": 0.23, "learning_rate": 4.625021834908682e-05, "loss": 0.0138, "step": 11594 }, { "epoch": 0.23, "learning_rate": 4.6249571388829586e-05, "loss": 0.0106, "step": 11596 }, { "epoch": 0.23, "learning_rate": 4.6248924428572355e-05, "loss": 0.0053, "step": 11598 }, { "epoch": 0.23, "learning_rate": 4.6248277468315124e-05, "loss": 0.0136, "step": 11600 }, { "epoch": 0.23, "learning_rate": 4.624763050805789e-05, "loss": 0.012, "step": 11602 }, { "epoch": 0.23, "learning_rate": 4.624698354780066e-05, "loss": 0.0074, "step": 11604 }, { "epoch": 0.23, "learning_rate": 4.624633658754343e-05, "loss": 0.0145, "step": 11606 }, { "epoch": 0.23, "learning_rate": 4.62456896272862e-05, "loss": 0.0013, "step": 11608 }, { "epoch": 0.23, "learning_rate": 4.624504266702896e-05, "loss": 0.0132, "step": 11610 }, { "epoch": 0.23, "learning_rate": 4.624439570677174e-05, "loss": 0.0102, "step": 11612 }, { "epoch": 0.23, "learning_rate": 4.62437487465145e-05, "loss": 0.0127, "step": 11614 }, { "epoch": 0.23, "learning_rate": 4.624310178625727e-05, "loss": 0.002, "step": 11616 }, { "epoch": 0.23, "learning_rate": 4.624245482600004e-05, "loss": 0.0016, "step": 11618 }, { "epoch": 0.23, "learning_rate": 4.624180786574281e-05, "loss": 0.0076, "step": 11620 }, { "epoch": 0.23, "learning_rate": 4.6241160905485584e-05, "loss": 0.0009, "step": 11622 }, { "epoch": 0.23, "learning_rate": 4.6240513945228346e-05, "loss": 0.0025, "step": 11624 }, { "epoch": 0.23, "learning_rate": 4.6239866984971115e-05, "loss": 0.0242, "step": 11626 }, { "epoch": 0.23, "learning_rate": 4.6239220024713884e-05, "loss": 0.0015, "step": 11628 }, { "epoch": 0.23, "learning_rate": 4.6238573064456654e-05, "loss": 0.0042, "step": 11630 }, { "epoch": 0.23, "learning_rate": 4.623792610419942e-05, "loss": 0.0098, "step": 11632 }, { "epoch": 0.23, "learning_rate": 4.623727914394219e-05, "loss": 0.0013, "step": 11634 }, { "epoch": 0.23, "learning_rate": 4.623663218368496e-05, "loss": 0.0066, "step": 11636 }, { "epoch": 0.23, "learning_rate": 4.623598522342772e-05, "loss": 0.0015, "step": 11638 }, { "epoch": 0.23, "learning_rate": 4.62353382631705e-05, "loss": 0.0008, "step": 11640 }, { "epoch": 0.23, "learning_rate": 4.623469130291326e-05, "loss": 0.0035, "step": 11642 }, { "epoch": 0.23, "learning_rate": 4.623404434265604e-05, "loss": 0.001, "step": 11644 }, { "epoch": 0.23, "learning_rate": 4.62333973823988e-05, "loss": 0.0108, "step": 11646 }, { "epoch": 0.23, "learning_rate": 4.623275042214157e-05, "loss": 0.0044, "step": 11648 }, { "epoch": 0.23, "learning_rate": 4.623210346188434e-05, "loss": 0.0033, "step": 11650 }, { "epoch": 0.23, "learning_rate": 4.623145650162711e-05, "loss": 0.0177, "step": 11652 }, { "epoch": 0.23, "learning_rate": 4.6230809541369876e-05, "loss": 0.0122, "step": 11654 }, { "epoch": 0.23, "learning_rate": 4.6230162581112645e-05, "loss": 0.0148, "step": 11656 }, { "epoch": 0.23, "learning_rate": 4.6229515620855414e-05, "loss": 0.0049, "step": 11658 }, { "epoch": 0.23, "learning_rate": 4.6228868660598176e-05, "loss": 0.0017, "step": 11660 }, { "epoch": 0.23, "learning_rate": 4.622822170034095e-05, "loss": 0.0177, "step": 11662 }, { "epoch": 0.23, "learning_rate": 4.6227574740083714e-05, "loss": 0.0064, "step": 11664 }, { "epoch": 0.23, "learning_rate": 4.622692777982649e-05, "loss": 0.009, "step": 11666 }, { "epoch": 0.23, "learning_rate": 4.622628081956926e-05, "loss": 0.0059, "step": 11668 }, { "epoch": 0.23, "learning_rate": 4.622563385931202e-05, "loss": 0.0053, "step": 11670 }, { "epoch": 0.23, "learning_rate": 4.62249868990548e-05, "loss": 0.014, "step": 11672 }, { "epoch": 0.23, "learning_rate": 4.622433993879756e-05, "loss": 0.0025, "step": 11674 }, { "epoch": 0.23, "learning_rate": 4.622369297854033e-05, "loss": 0.0055, "step": 11676 }, { "epoch": 0.23, "learning_rate": 4.62230460182831e-05, "loss": 0.0043, "step": 11678 }, { "epoch": 0.23, "learning_rate": 4.622239905802587e-05, "loss": 0.0048, "step": 11680 }, { "epoch": 0.23, "learning_rate": 4.6221752097768636e-05, "loss": 0.0024, "step": 11682 }, { "epoch": 0.23, "learning_rate": 4.6221105137511405e-05, "loss": 0.0052, "step": 11684 }, { "epoch": 0.23, "learning_rate": 4.6220458177254174e-05, "loss": 0.0027, "step": 11686 }, { "epoch": 0.23, "learning_rate": 4.6219811216996943e-05, "loss": 0.0024, "step": 11688 }, { "epoch": 0.23, "learning_rate": 4.621916425673971e-05, "loss": 0.0112, "step": 11690 }, { "epoch": 0.23, "learning_rate": 4.6218517296482475e-05, "loss": 0.0046, "step": 11692 }, { "epoch": 0.23, "learning_rate": 4.621787033622525e-05, "loss": 0.0279, "step": 11694 }, { "epoch": 0.23, "learning_rate": 4.621722337596801e-05, "loss": 0.0219, "step": 11696 }, { "epoch": 0.23, "learning_rate": 4.621657641571078e-05, "loss": 0.008, "step": 11698 }, { "epoch": 0.23, "learning_rate": 4.621592945545356e-05, "loss": 0.01, "step": 11700 }, { "epoch": 0.23, "learning_rate": 4.621528249519632e-05, "loss": 0.0042, "step": 11702 }, { "epoch": 0.23, "learning_rate": 4.6214635534939096e-05, "loss": 0.0034, "step": 11704 }, { "epoch": 0.23, "learning_rate": 4.621398857468186e-05, "loss": 0.0249, "step": 11706 }, { "epoch": 0.23, "learning_rate": 4.621334161442463e-05, "loss": 0.0031, "step": 11708 }, { "epoch": 0.23, "learning_rate": 4.6212694654167397e-05, "loss": 0.0008, "step": 11710 }, { "epoch": 0.23, "learning_rate": 4.6212047693910166e-05, "loss": 0.001, "step": 11712 }, { "epoch": 0.23, "learning_rate": 4.6211400733652935e-05, "loss": 0.0182, "step": 11714 }, { "epoch": 0.23, "learning_rate": 4.6210753773395704e-05, "loss": 0.017, "step": 11716 }, { "epoch": 0.23, "learning_rate": 4.621010681313847e-05, "loss": 0.0041, "step": 11718 }, { "epoch": 0.23, "learning_rate": 4.6209459852881235e-05, "loss": 0.0052, "step": 11720 }, { "epoch": 0.23, "learning_rate": 4.620881289262401e-05, "loss": 0.0032, "step": 11722 }, { "epoch": 0.23, "learning_rate": 4.6208165932366773e-05, "loss": 0.0112, "step": 11724 }, { "epoch": 0.23, "learning_rate": 4.620751897210955e-05, "loss": 0.1246, "step": 11726 }, { "epoch": 0.23, "learning_rate": 4.620687201185231e-05, "loss": 0.0111, "step": 11728 }, { "epoch": 0.23, "learning_rate": 4.620622505159508e-05, "loss": 0.0135, "step": 11730 }, { "epoch": 0.23, "learning_rate": 4.6205578091337857e-05, "loss": 0.0072, "step": 11732 }, { "epoch": 0.23, "learning_rate": 4.620493113108062e-05, "loss": 0.0071, "step": 11734 }, { "epoch": 0.23, "learning_rate": 4.620428417082339e-05, "loss": 0.0163, "step": 11736 }, { "epoch": 0.23, "learning_rate": 4.620363721056616e-05, "loss": 0.0092, "step": 11738 }, { "epoch": 0.23, "learning_rate": 4.6202990250308926e-05, "loss": 0.0161, "step": 11740 }, { "epoch": 0.23, "learning_rate": 4.620234329005169e-05, "loss": 0.0013, "step": 11742 }, { "epoch": 0.23, "learning_rate": 4.6201696329794464e-05, "loss": 0.0445, "step": 11744 }, { "epoch": 0.23, "learning_rate": 4.620104936953723e-05, "loss": 0.0092, "step": 11746 }, { "epoch": 0.23, "learning_rate": 4.620040240928e-05, "loss": 0.0176, "step": 11748 }, { "epoch": 0.23, "learning_rate": 4.619975544902277e-05, "loss": 0.0023, "step": 11750 }, { "epoch": 0.23, "learning_rate": 4.6199108488765534e-05, "loss": 0.0185, "step": 11752 }, { "epoch": 0.23, "learning_rate": 4.619846152850831e-05, "loss": 0.0016, "step": 11754 }, { "epoch": 0.23, "learning_rate": 4.619781456825107e-05, "loss": 0.0231, "step": 11756 }, { "epoch": 0.23, "learning_rate": 4.619716760799384e-05, "loss": 0.0023, "step": 11758 }, { "epoch": 0.23, "learning_rate": 4.619652064773661e-05, "loss": 0.016, "step": 11760 }, { "epoch": 0.23, "learning_rate": 4.619587368747938e-05, "loss": 0.001, "step": 11762 }, { "epoch": 0.23, "learning_rate": 4.6195226727222155e-05, "loss": 0.0247, "step": 11764 }, { "epoch": 0.23, "learning_rate": 4.619457976696492e-05, "loss": 0.0093, "step": 11766 }, { "epoch": 0.23, "learning_rate": 4.6193932806707686e-05, "loss": 0.0004, "step": 11768 }, { "epoch": 0.23, "learning_rate": 4.6193285846450456e-05, "loss": 0.0047, "step": 11770 }, { "epoch": 0.23, "learning_rate": 4.6192638886193225e-05, "loss": 0.0024, "step": 11772 }, { "epoch": 0.23, "learning_rate": 4.619199192593599e-05, "loss": 0.0053, "step": 11774 }, { "epoch": 0.23, "learning_rate": 4.619134496567876e-05, "loss": 0.0052, "step": 11776 }, { "epoch": 0.23, "learning_rate": 4.619069800542153e-05, "loss": 0.0005, "step": 11778 }, { "epoch": 0.23, "learning_rate": 4.6190051045164294e-05, "loss": 0.0077, "step": 11780 }, { "epoch": 0.23, "learning_rate": 4.618940408490707e-05, "loss": 0.0117, "step": 11782 }, { "epoch": 0.23, "learning_rate": 4.618875712464983e-05, "loss": 0.0036, "step": 11784 }, { "epoch": 0.23, "learning_rate": 4.618811016439261e-05, "loss": 0.0145, "step": 11786 }, { "epoch": 0.23, "learning_rate": 4.618746320413537e-05, "loss": 0.0094, "step": 11788 }, { "epoch": 0.23, "learning_rate": 4.618681624387814e-05, "loss": 0.0039, "step": 11790 }, { "epoch": 0.23, "learning_rate": 4.618616928362091e-05, "loss": 0.0126, "step": 11792 }, { "epoch": 0.23, "learning_rate": 4.618552232336368e-05, "loss": 0.0007, "step": 11794 }, { "epoch": 0.23, "learning_rate": 4.618487536310645e-05, "loss": 0.001, "step": 11796 }, { "epoch": 0.23, "learning_rate": 4.6184228402849216e-05, "loss": 0.0121, "step": 11798 }, { "epoch": 0.23, "learning_rate": 4.6183581442591985e-05, "loss": 0.0049, "step": 11800 }, { "epoch": 0.23, "learning_rate": 4.618293448233475e-05, "loss": 0.009, "step": 11802 }, { "epoch": 0.23, "learning_rate": 4.618228752207752e-05, "loss": 0.0004, "step": 11804 }, { "epoch": 0.23, "learning_rate": 4.6181640561820286e-05, "loss": 0.0054, "step": 11806 }, { "epoch": 0.23, "learning_rate": 4.618099360156306e-05, "loss": 0.0268, "step": 11808 }, { "epoch": 0.23, "learning_rate": 4.618034664130583e-05, "loss": 0.0103, "step": 11810 }, { "epoch": 0.23, "learning_rate": 4.617969968104859e-05, "loss": 0.0071, "step": 11812 }, { "epoch": 0.23, "learning_rate": 4.617905272079137e-05, "loss": 0.0007, "step": 11814 }, { "epoch": 0.23, "learning_rate": 4.617840576053413e-05, "loss": 0.0009, "step": 11816 }, { "epoch": 0.23, "learning_rate": 4.61777588002769e-05, "loss": 0.0023, "step": 11818 }, { "epoch": 0.23, "learning_rate": 4.617711184001967e-05, "loss": 0.0101, "step": 11820 }, { "epoch": 0.23, "learning_rate": 4.617646487976244e-05, "loss": 0.0038, "step": 11822 }, { "epoch": 0.23, "learning_rate": 4.617581791950521e-05, "loss": 0.002, "step": 11824 }, { "epoch": 0.23, "learning_rate": 4.6175170959247976e-05, "loss": 0.0093, "step": 11826 }, { "epoch": 0.23, "learning_rate": 4.6174523998990745e-05, "loss": 0.0174, "step": 11828 }, { "epoch": 0.23, "learning_rate": 4.6173877038733515e-05, "loss": 0.0034, "step": 11830 }, { "epoch": 0.23, "learning_rate": 4.6173230078476284e-05, "loss": 0.007, "step": 11832 }, { "epoch": 0.23, "learning_rate": 4.6172583118219046e-05, "loss": 0.0331, "step": 11834 }, { "epoch": 0.23, "learning_rate": 4.617193615796182e-05, "loss": 0.0096, "step": 11836 }, { "epoch": 0.23, "learning_rate": 4.6171289197704584e-05, "loss": 0.0028, "step": 11838 }, { "epoch": 0.23, "learning_rate": 4.617064223744735e-05, "loss": 0.0024, "step": 11840 }, { "epoch": 0.23, "learning_rate": 4.616999527719012e-05, "loss": 0.0013, "step": 11842 }, { "epoch": 0.23, "learning_rate": 4.616934831693289e-05, "loss": 0.0239, "step": 11844 }, { "epoch": 0.23, "learning_rate": 4.616870135667567e-05, "loss": 0.014, "step": 11846 }, { "epoch": 0.23, "learning_rate": 4.616805439641843e-05, "loss": 0.0103, "step": 11848 }, { "epoch": 0.23, "learning_rate": 4.61674074361612e-05, "loss": 0.0017, "step": 11850 }, { "epoch": 0.23, "learning_rate": 4.616676047590397e-05, "loss": 0.006, "step": 11852 }, { "epoch": 0.23, "learning_rate": 4.616611351564674e-05, "loss": 0.0311, "step": 11854 }, { "epoch": 0.23, "learning_rate": 4.6165466555389506e-05, "loss": 0.0112, "step": 11856 }, { "epoch": 0.23, "learning_rate": 4.6164819595132275e-05, "loss": 0.0109, "step": 11858 }, { "epoch": 0.23, "learning_rate": 4.6164172634875044e-05, "loss": 0.0147, "step": 11860 }, { "epoch": 0.23, "learning_rate": 4.6163525674617806e-05, "loss": 0.0008, "step": 11862 }, { "epoch": 0.23, "learning_rate": 4.616287871436058e-05, "loss": 0.002, "step": 11864 }, { "epoch": 0.23, "learning_rate": 4.6162231754103345e-05, "loss": 0.0047, "step": 11866 }, { "epoch": 0.23, "learning_rate": 4.616158479384612e-05, "loss": 0.0132, "step": 11868 }, { "epoch": 0.23, "learning_rate": 4.616093783358888e-05, "loss": 0.0085, "step": 11870 }, { "epoch": 0.23, "learning_rate": 4.616029087333165e-05, "loss": 0.0112, "step": 11872 }, { "epoch": 0.23, "learning_rate": 4.615964391307442e-05, "loss": 0.0016, "step": 11874 }, { "epoch": 0.23, "learning_rate": 4.615899695281719e-05, "loss": 0.0141, "step": 11876 }, { "epoch": 0.23, "learning_rate": 4.615834999255996e-05, "loss": 0.0062, "step": 11878 }, { "epoch": 0.23, "learning_rate": 4.615770303230273e-05, "loss": 0.0018, "step": 11880 }, { "epoch": 0.23, "learning_rate": 4.61570560720455e-05, "loss": 0.0176, "step": 11882 }, { "epoch": 0.23, "learning_rate": 4.615640911178826e-05, "loss": 0.009, "step": 11884 }, { "epoch": 0.23, "learning_rate": 4.6155762151531035e-05, "loss": 0.0058, "step": 11886 }, { "epoch": 0.23, "learning_rate": 4.6155115191273804e-05, "loss": 0.0188, "step": 11888 }, { "epoch": 0.23, "learning_rate": 4.6154468231016574e-05, "loss": 0.0038, "step": 11890 }, { "epoch": 0.23, "learning_rate": 4.615382127075934e-05, "loss": 0.0024, "step": 11892 }, { "epoch": 0.23, "learning_rate": 4.6153174310502105e-05, "loss": 0.0018, "step": 11894 }, { "epoch": 0.23, "learning_rate": 4.615252735024488e-05, "loss": 0.0012, "step": 11896 }, { "epoch": 0.23, "learning_rate": 4.615188038998764e-05, "loss": 0.0076, "step": 11898 }, { "epoch": 0.23, "learning_rate": 4.615123342973041e-05, "loss": 0.0087, "step": 11900 }, { "epoch": 0.23, "learning_rate": 4.615058646947318e-05, "loss": 0.0154, "step": 11902 }, { "epoch": 0.23, "learning_rate": 4.614993950921595e-05, "loss": 0.0094, "step": 11904 }, { "epoch": 0.23, "learning_rate": 4.614929254895872e-05, "loss": 0.002, "step": 11906 }, { "epoch": 0.23, "learning_rate": 4.614864558870149e-05, "loss": 0.0124, "step": 11908 }, { "epoch": 0.23, "learning_rate": 4.614799862844426e-05, "loss": 0.002, "step": 11910 }, { "epoch": 0.23, "learning_rate": 4.614735166818703e-05, "loss": 0.0011, "step": 11912 }, { "epoch": 0.23, "learning_rate": 4.6146704707929796e-05, "loss": 0.0209, "step": 11914 }, { "epoch": 0.23, "learning_rate": 4.614605774767256e-05, "loss": 0.0161, "step": 11916 }, { "epoch": 0.23, "learning_rate": 4.6145410787415334e-05, "loss": 0.0004, "step": 11918 }, { "epoch": 0.23, "learning_rate": 4.6144763827158096e-05, "loss": 0.0033, "step": 11920 }, { "epoch": 0.23, "learning_rate": 4.6144116866900865e-05, "loss": 0.0067, "step": 11922 }, { "epoch": 0.23, "learning_rate": 4.614346990664364e-05, "loss": 0.0077, "step": 11924 }, { "epoch": 0.23, "learning_rate": 4.6142822946386403e-05, "loss": 0.0037, "step": 11926 }, { "epoch": 0.23, "learning_rate": 4.614217598612918e-05, "loss": 0.0015, "step": 11928 }, { "epoch": 0.23, "learning_rate": 4.614152902587194e-05, "loss": 0.0049, "step": 11930 }, { "epoch": 0.23, "learning_rate": 4.614088206561471e-05, "loss": 0.0023, "step": 11932 }, { "epoch": 0.23, "learning_rate": 4.614023510535748e-05, "loss": 0.0009, "step": 11934 }, { "epoch": 0.23, "learning_rate": 4.613958814510025e-05, "loss": 0.022, "step": 11936 }, { "epoch": 0.23, "learning_rate": 4.613894118484302e-05, "loss": 0.0028, "step": 11938 }, { "epoch": 0.23, "learning_rate": 4.613829422458579e-05, "loss": 0.0019, "step": 11940 }, { "epoch": 0.23, "learning_rate": 4.6137647264328556e-05, "loss": 0.0039, "step": 11942 }, { "epoch": 0.23, "learning_rate": 4.613700030407132e-05, "loss": 0.0023, "step": 11944 }, { "epoch": 0.23, "learning_rate": 4.6136353343814094e-05, "loss": 0.0134, "step": 11946 }, { "epoch": 0.23, "learning_rate": 4.613570638355686e-05, "loss": 0.0016, "step": 11948 }, { "epoch": 0.23, "learning_rate": 4.613505942329963e-05, "loss": 0.0022, "step": 11950 }, { "epoch": 0.23, "learning_rate": 4.6134412463042395e-05, "loss": 0.0104, "step": 11952 }, { "epoch": 0.23, "learning_rate": 4.6133765502785164e-05, "loss": 0.0215, "step": 11954 }, { "epoch": 0.23, "learning_rate": 4.613311854252794e-05, "loss": 0.0089, "step": 11956 }, { "epoch": 0.23, "learning_rate": 4.61324715822707e-05, "loss": 0.0069, "step": 11958 }, { "epoch": 0.23, "learning_rate": 4.613182462201347e-05, "loss": 0.0132, "step": 11960 }, { "epoch": 0.23, "learning_rate": 4.613117766175624e-05, "loss": 0.0062, "step": 11962 }, { "epoch": 0.23, "learning_rate": 4.613053070149901e-05, "loss": 0.0004, "step": 11964 }, { "epoch": 0.23, "learning_rate": 4.612988374124178e-05, "loss": 0.0097, "step": 11966 }, { "epoch": 0.23, "learning_rate": 4.612923678098455e-05, "loss": 0.0032, "step": 11968 }, { "epoch": 0.23, "learning_rate": 4.6128589820727317e-05, "loss": 0.0152, "step": 11970 }, { "epoch": 0.23, "learning_rate": 4.6127942860470086e-05, "loss": 0.0108, "step": 11972 }, { "epoch": 0.23, "learning_rate": 4.6127295900212855e-05, "loss": 0.0112, "step": 11974 }, { "epoch": 0.23, "learning_rate": 4.612664893995562e-05, "loss": 0.0155, "step": 11976 }, { "epoch": 0.23, "learning_rate": 4.612600197969839e-05, "loss": 0.0043, "step": 11978 }, { "epoch": 0.23, "learning_rate": 4.6125355019441155e-05, "loss": 0.0023, "step": 11980 }, { "epoch": 0.23, "learning_rate": 4.6124708059183924e-05, "loss": 0.0134, "step": 11982 }, { "epoch": 0.23, "learning_rate": 4.612406109892669e-05, "loss": 0.0005, "step": 11984 }, { "epoch": 0.23, "learning_rate": 4.612341413866946e-05, "loss": 0.0005, "step": 11986 }, { "epoch": 0.23, "learning_rate": 4.612276717841224e-05, "loss": 0.0029, "step": 11988 }, { "epoch": 0.23, "learning_rate": 4.6122120218155e-05, "loss": 0.0015, "step": 11990 }, { "epoch": 0.23, "learning_rate": 4.612147325789777e-05, "loss": 0.014, "step": 11992 }, { "epoch": 0.23, "learning_rate": 4.612082629764054e-05, "loss": 0.0039, "step": 11994 }, { "epoch": 0.23, "learning_rate": 4.612017933738331e-05, "loss": 0.001, "step": 11996 }, { "epoch": 0.23, "learning_rate": 4.611953237712607e-05, "loss": 0.0009, "step": 11998 }, { "epoch": 0.23, "learning_rate": 4.6118885416868846e-05, "loss": 0.0012, "step": 12000 }, { "epoch": 0.23, "learning_rate": 4.6118238456611615e-05, "loss": 0.0012, "step": 12002 }, { "epoch": 0.23, "learning_rate": 4.611759149635438e-05, "loss": 0.0049, "step": 12004 }, { "epoch": 0.23, "learning_rate": 4.611694453609715e-05, "loss": 0.0258, "step": 12006 }, { "epoch": 0.23, "learning_rate": 4.6116297575839916e-05, "loss": 0.0012, "step": 12008 }, { "epoch": 0.23, "learning_rate": 4.611565061558269e-05, "loss": 0.0073, "step": 12010 }, { "epoch": 0.23, "learning_rate": 4.6115003655325454e-05, "loss": 0.0101, "step": 12012 }, { "epoch": 0.23, "learning_rate": 4.611435669506822e-05, "loss": 0.0012, "step": 12014 }, { "epoch": 0.23, "learning_rate": 4.611370973481099e-05, "loss": 0.0099, "step": 12016 }, { "epoch": 0.23, "learning_rate": 4.611306277455376e-05, "loss": 0.0274, "step": 12018 }, { "epoch": 0.23, "learning_rate": 4.611241581429653e-05, "loss": 0.004, "step": 12020 }, { "epoch": 0.23, "learning_rate": 4.61117688540393e-05, "loss": 0.0018, "step": 12022 }, { "epoch": 0.23, "learning_rate": 4.611112189378207e-05, "loss": 0.0042, "step": 12024 }, { "epoch": 0.23, "learning_rate": 4.611047493352483e-05, "loss": 0.0016, "step": 12026 }, { "epoch": 0.23, "learning_rate": 4.6109827973267606e-05, "loss": 0.016, "step": 12028 }, { "epoch": 0.23, "learning_rate": 4.610918101301037e-05, "loss": 0.0515, "step": 12030 }, { "epoch": 0.23, "learning_rate": 4.6108534052753145e-05, "loss": 0.0037, "step": 12032 }, { "epoch": 0.23, "learning_rate": 4.6107887092495914e-05, "loss": 0.0016, "step": 12034 }, { "epoch": 0.23, "learning_rate": 4.6107240132238676e-05, "loss": 0.0041, "step": 12036 }, { "epoch": 0.23, "learning_rate": 4.610659317198145e-05, "loss": 0.0012, "step": 12038 }, { "epoch": 0.23, "learning_rate": 4.6105946211724214e-05, "loss": 0.0076, "step": 12040 }, { "epoch": 0.23, "learning_rate": 4.610529925146698e-05, "loss": 0.0133, "step": 12042 }, { "epoch": 0.23, "learning_rate": 4.610465229120975e-05, "loss": 0.0064, "step": 12044 }, { "epoch": 0.23, "learning_rate": 4.610400533095252e-05, "loss": 0.0107, "step": 12046 }, { "epoch": 0.23, "learning_rate": 4.610335837069529e-05, "loss": 0.0047, "step": 12048 }, { "epoch": 0.23, "learning_rate": 4.610271141043806e-05, "loss": 0.0012, "step": 12050 }, { "epoch": 0.23, "learning_rate": 4.610206445018083e-05, "loss": 0.0534, "step": 12052 }, { "epoch": 0.23, "learning_rate": 4.61014174899236e-05, "loss": 0.0136, "step": 12054 }, { "epoch": 0.23, "learning_rate": 4.610077052966637e-05, "loss": 0.077, "step": 12056 }, { "epoch": 0.23, "learning_rate": 4.610012356940913e-05, "loss": 0.0048, "step": 12058 }, { "epoch": 0.23, "learning_rate": 4.6099476609151905e-05, "loss": 0.0051, "step": 12060 }, { "epoch": 0.23, "learning_rate": 4.609882964889467e-05, "loss": 0.0093, "step": 12062 }, { "epoch": 0.23, "learning_rate": 4.6098182688637436e-05, "loss": 0.0048, "step": 12064 }, { "epoch": 0.23, "learning_rate": 4.609753572838021e-05, "loss": 0.0328, "step": 12066 }, { "epoch": 0.23, "learning_rate": 4.6096888768122975e-05, "loss": 0.0028, "step": 12068 }, { "epoch": 0.23, "learning_rate": 4.609624180786575e-05, "loss": 0.0647, "step": 12070 }, { "epoch": 0.23, "learning_rate": 4.609559484760851e-05, "loss": 0.0276, "step": 12072 }, { "epoch": 0.23, "learning_rate": 4.609494788735128e-05, "loss": 0.0116, "step": 12074 }, { "epoch": 0.23, "learning_rate": 4.609430092709405e-05, "loss": 0.0022, "step": 12076 }, { "epoch": 0.23, "learning_rate": 4.609365396683682e-05, "loss": 0.0029, "step": 12078 }, { "epoch": 0.23, "learning_rate": 4.609300700657959e-05, "loss": 0.0153, "step": 12080 }, { "epoch": 0.23, "learning_rate": 4.609236004632236e-05, "loss": 0.1149, "step": 12082 }, { "epoch": 0.23, "learning_rate": 4.609203656619374e-05, "loss": 0.1521, "step": 12084 }, { "epoch": 0.23, "learning_rate": 4.6091389605936505e-05, "loss": 0.0256, "step": 12086 }, { "epoch": 0.23, "learning_rate": 4.609074264567928e-05, "loss": 0.0363, "step": 12088 }, { "epoch": 0.23, "learning_rate": 4.609009568542205e-05, "loss": 0.006, "step": 12090 }, { "epoch": 0.23, "learning_rate": 4.608944872516481e-05, "loss": 0.0016, "step": 12092 }, { "epoch": 0.23, "learning_rate": 4.608880176490759e-05, "loss": 0.0059, "step": 12094 }, { "epoch": 0.23, "learning_rate": 4.608815480465035e-05, "loss": 0.0157, "step": 12096 }, { "epoch": 0.23, "learning_rate": 4.608750784439312e-05, "loss": 0.0048, "step": 12098 }, { "epoch": 0.23, "learning_rate": 4.608686088413589e-05, "loss": 0.01, "step": 12100 }, { "epoch": 0.23, "learning_rate": 4.608621392387866e-05, "loss": 0.0035, "step": 12102 }, { "epoch": 0.23, "learning_rate": 4.608556696362143e-05, "loss": 0.0085, "step": 12104 }, { "epoch": 0.23, "learning_rate": 4.6084920003364196e-05, "loss": 0.0249, "step": 12106 }, { "epoch": 0.24, "learning_rate": 4.6084273043106965e-05, "loss": 0.0446, "step": 12108 }, { "epoch": 0.24, "learning_rate": 4.6083626082849734e-05, "loss": 0.0092, "step": 12110 }, { "epoch": 0.24, "learning_rate": 4.60829791225925e-05, "loss": 0.0079, "step": 12112 }, { "epoch": 0.24, "learning_rate": 4.6082332162335265e-05, "loss": 0.0019, "step": 12114 }, { "epoch": 0.24, "learning_rate": 4.608168520207804e-05, "loss": 0.002, "step": 12116 }, { "epoch": 0.24, "learning_rate": 4.6081038241820804e-05, "loss": 0.0018, "step": 12118 }, { "epoch": 0.24, "learning_rate": 4.608039128156357e-05, "loss": 0.0128, "step": 12120 }, { "epoch": 0.24, "learning_rate": 4.607974432130634e-05, "loss": 0.0016, "step": 12122 }, { "epoch": 0.24, "learning_rate": 4.607909736104911e-05, "loss": 0.008, "step": 12124 }, { "epoch": 0.24, "learning_rate": 4.607845040079189e-05, "loss": 0.0222, "step": 12126 }, { "epoch": 0.24, "learning_rate": 4.607780344053465e-05, "loss": 0.0008, "step": 12128 }, { "epoch": 0.24, "learning_rate": 4.607715648027742e-05, "loss": 0.0162, "step": 12130 }, { "epoch": 0.24, "learning_rate": 4.607650952002019e-05, "loss": 0.0108, "step": 12132 }, { "epoch": 0.24, "learning_rate": 4.6075862559762956e-05, "loss": 0.0048, "step": 12134 }, { "epoch": 0.24, "learning_rate": 4.6075215599505725e-05, "loss": 0.0082, "step": 12136 }, { "epoch": 0.24, "learning_rate": 4.6074568639248494e-05, "loss": 0.0012, "step": 12138 }, { "epoch": 0.24, "learning_rate": 4.6073921678991264e-05, "loss": 0.0063, "step": 12140 }, { "epoch": 0.24, "learning_rate": 4.607327471873403e-05, "loss": 0.0158, "step": 12142 }, { "epoch": 0.24, "learning_rate": 4.60726277584768e-05, "loss": 0.0065, "step": 12144 }, { "epoch": 0.24, "learning_rate": 4.6071980798219564e-05, "loss": 0.0021, "step": 12146 }, { "epoch": 0.24, "learning_rate": 4.607133383796234e-05, "loss": 0.0023, "step": 12148 }, { "epoch": 0.24, "learning_rate": 4.60706868777051e-05, "loss": 0.004, "step": 12150 }, { "epoch": 0.24, "learning_rate": 4.607003991744787e-05, "loss": 0.0368, "step": 12152 }, { "epoch": 0.24, "learning_rate": 4.606939295719064e-05, "loss": 0.0146, "step": 12154 }, { "epoch": 0.24, "learning_rate": 4.606874599693341e-05, "loss": 0.0298, "step": 12156 }, { "epoch": 0.24, "learning_rate": 4.606809903667618e-05, "loss": 0.0095, "step": 12158 }, { "epoch": 0.24, "learning_rate": 4.606745207641895e-05, "loss": 0.0164, "step": 12160 }, { "epoch": 0.24, "learning_rate": 4.606680511616172e-05, "loss": 0.011, "step": 12162 }, { "epoch": 0.24, "learning_rate": 4.6066158155904486e-05, "loss": 0.0018, "step": 12164 }, { "epoch": 0.24, "learning_rate": 4.6065511195647255e-05, "loss": 0.0044, "step": 12166 }, { "epoch": 0.24, "learning_rate": 4.6064864235390024e-05, "loss": 0.0088, "step": 12168 }, { "epoch": 0.24, "learning_rate": 4.606421727513279e-05, "loss": 0.0052, "step": 12170 }, { "epoch": 0.24, "learning_rate": 4.606357031487556e-05, "loss": 0.0025, "step": 12172 }, { "epoch": 0.24, "learning_rate": 4.6062923354618324e-05, "loss": 0.0037, "step": 12174 }, { "epoch": 0.24, "learning_rate": 4.60622763943611e-05, "loss": 0.0694, "step": 12176 }, { "epoch": 0.24, "learning_rate": 4.606162943410386e-05, "loss": 0.0022, "step": 12178 }, { "epoch": 0.24, "learning_rate": 4.606098247384663e-05, "loss": 0.002, "step": 12180 }, { "epoch": 0.24, "learning_rate": 4.60603355135894e-05, "loss": 0.0111, "step": 12182 }, { "epoch": 0.24, "learning_rate": 4.605968855333217e-05, "loss": 0.001, "step": 12184 }, { "epoch": 0.24, "learning_rate": 4.605904159307494e-05, "loss": 0.0023, "step": 12186 }, { "epoch": 0.24, "learning_rate": 4.605839463281771e-05, "loss": 0.0006, "step": 12188 }, { "epoch": 0.24, "learning_rate": 4.605774767256048e-05, "loss": 0.0005, "step": 12190 }, { "epoch": 0.24, "learning_rate": 4.6057100712303246e-05, "loss": 0.004, "step": 12192 }, { "epoch": 0.24, "learning_rate": 4.6056453752046015e-05, "loss": 0.0017, "step": 12194 }, { "epoch": 0.24, "learning_rate": 4.605580679178878e-05, "loss": 0.0632, "step": 12196 }, { "epoch": 0.24, "learning_rate": 4.6055159831531553e-05, "loss": 0.0007, "step": 12198 }, { "epoch": 0.24, "learning_rate": 4.6054512871274316e-05, "loss": 0.0176, "step": 12200 }, { "epoch": 0.24, "learning_rate": 4.605386591101709e-05, "loss": 0.0157, "step": 12202 }, { "epoch": 0.24, "learning_rate": 4.605321895075986e-05, "loss": 0.0116, "step": 12204 }, { "epoch": 0.24, "learning_rate": 4.605257199050262e-05, "loss": 0.0033, "step": 12206 }, { "epoch": 0.24, "learning_rate": 4.60519250302454e-05, "loss": 0.008, "step": 12208 }, { "epoch": 0.24, "learning_rate": 4.605127806998816e-05, "loss": 0.0031, "step": 12210 }, { "epoch": 0.24, "learning_rate": 4.605063110973093e-05, "loss": 0.0009, "step": 12212 }, { "epoch": 0.24, "learning_rate": 4.60499841494737e-05, "loss": 0.0391, "step": 12214 }, { "epoch": 0.24, "learning_rate": 4.604933718921647e-05, "loss": 0.0037, "step": 12216 }, { "epoch": 0.24, "learning_rate": 4.604869022895924e-05, "loss": 0.0172, "step": 12218 }, { "epoch": 0.24, "learning_rate": 4.6048043268702007e-05, "loss": 0.0072, "step": 12220 }, { "epoch": 0.24, "learning_rate": 4.6047396308444776e-05, "loss": 0.0047, "step": 12222 }, { "epoch": 0.24, "learning_rate": 4.6046749348187545e-05, "loss": 0.0048, "step": 12224 }, { "epoch": 0.24, "learning_rate": 4.6046102387930314e-05, "loss": 0.0304, "step": 12226 }, { "epoch": 0.24, "learning_rate": 4.6045455427673076e-05, "loss": 0.0078, "step": 12228 }, { "epoch": 0.24, "learning_rate": 4.604480846741585e-05, "loss": 0.0074, "step": 12230 }, { "epoch": 0.24, "learning_rate": 4.6044161507158614e-05, "loss": 0.0193, "step": 12232 }, { "epoch": 0.24, "learning_rate": 4.6043514546901383e-05, "loss": 0.0078, "step": 12234 }, { "epoch": 0.24, "learning_rate": 4.604286758664416e-05, "loss": 0.0073, "step": 12236 }, { "epoch": 0.24, "learning_rate": 4.604222062638692e-05, "loss": 0.0036, "step": 12238 }, { "epoch": 0.24, "learning_rate": 4.604157366612969e-05, "loss": 0.0023, "step": 12240 }, { "epoch": 0.24, "learning_rate": 4.604092670587246e-05, "loss": 0.0005, "step": 12242 }, { "epoch": 0.24, "learning_rate": 4.604027974561523e-05, "loss": 0.0044, "step": 12244 }, { "epoch": 0.24, "learning_rate": 4.6039632785358e-05, "loss": 0.0091, "step": 12246 }, { "epoch": 0.24, "learning_rate": 4.603898582510077e-05, "loss": 0.0072, "step": 12248 }, { "epoch": 0.24, "learning_rate": 4.6038338864843536e-05, "loss": 0.0011, "step": 12250 }, { "epoch": 0.24, "learning_rate": 4.6037691904586305e-05, "loss": 0.0018, "step": 12252 }, { "epoch": 0.24, "learning_rate": 4.6037044944329074e-05, "loss": 0.0122, "step": 12254 }, { "epoch": 0.24, "learning_rate": 4.6036397984071837e-05, "loss": 0.0085, "step": 12256 }, { "epoch": 0.24, "learning_rate": 4.603575102381461e-05, "loss": 0.0072, "step": 12258 }, { "epoch": 0.24, "learning_rate": 4.6035104063557375e-05, "loss": 0.0128, "step": 12260 }, { "epoch": 0.24, "learning_rate": 4.6034457103300144e-05, "loss": 0.0022, "step": 12262 }, { "epoch": 0.24, "learning_rate": 4.603381014304291e-05, "loss": 0.0132, "step": 12264 }, { "epoch": 0.24, "learning_rate": 4.603316318278568e-05, "loss": 0.0055, "step": 12266 }, { "epoch": 0.24, "learning_rate": 4.603251622252846e-05, "loss": 0.0066, "step": 12268 }, { "epoch": 0.24, "learning_rate": 4.603186926227122e-05, "loss": 0.0236, "step": 12270 }, { "epoch": 0.24, "learning_rate": 4.603122230201399e-05, "loss": 0.0168, "step": 12272 }, { "epoch": 0.24, "learning_rate": 4.603057534175676e-05, "loss": 0.0008, "step": 12274 }, { "epoch": 0.24, "learning_rate": 4.602992838149953e-05, "loss": 0.0026, "step": 12276 }, { "epoch": 0.24, "learning_rate": 4.602928142124229e-05, "loss": 0.004, "step": 12278 }, { "epoch": 0.24, "learning_rate": 4.6028634460985066e-05, "loss": 0.0034, "step": 12280 }, { "epoch": 0.24, "learning_rate": 4.6027987500727835e-05, "loss": 0.0007, "step": 12282 }, { "epoch": 0.24, "learning_rate": 4.6027340540470604e-05, "loss": 0.0103, "step": 12284 }, { "epoch": 0.24, "learning_rate": 4.602669358021337e-05, "loss": 0.0048, "step": 12286 }, { "epoch": 0.24, "learning_rate": 4.6026046619956135e-05, "loss": 0.0095, "step": 12288 }, { "epoch": 0.24, "learning_rate": 4.602539965969891e-05, "loss": 0.0047, "step": 12290 }, { "epoch": 0.24, "learning_rate": 4.602475269944167e-05, "loss": 0.0052, "step": 12292 }, { "epoch": 0.24, "learning_rate": 4.602410573918444e-05, "loss": 0.0022, "step": 12294 }, { "epoch": 0.24, "learning_rate": 4.602345877892721e-05, "loss": 0.0024, "step": 12296 }, { "epoch": 0.24, "learning_rate": 4.602281181866998e-05, "loss": 0.0061, "step": 12298 }, { "epoch": 0.24, "learning_rate": 4.602216485841275e-05, "loss": 0.0086, "step": 12300 }, { "epoch": 0.24, "learning_rate": 4.602151789815552e-05, "loss": 0.0442, "step": 12302 }, { "epoch": 0.24, "learning_rate": 4.602087093789829e-05, "loss": 0.0061, "step": 12304 }, { "epoch": 0.24, "learning_rate": 4.602022397764106e-05, "loss": 0.0099, "step": 12306 }, { "epoch": 0.24, "learning_rate": 4.6019577017383826e-05, "loss": 0.0039, "step": 12308 }, { "epoch": 0.24, "learning_rate": 4.601893005712659e-05, "loss": 0.0009, "step": 12310 }, { "epoch": 0.24, "learning_rate": 4.6018283096869364e-05, "loss": 0.0048, "step": 12312 }, { "epoch": 0.24, "learning_rate": 4.601763613661213e-05, "loss": 0.0061, "step": 12314 }, { "epoch": 0.24, "learning_rate": 4.6016989176354896e-05, "loss": 0.0137, "step": 12316 }, { "epoch": 0.24, "learning_rate": 4.601634221609767e-05, "loss": 0.0036, "step": 12318 }, { "epoch": 0.24, "learning_rate": 4.6015695255840434e-05, "loss": 0.0009, "step": 12320 }, { "epoch": 0.24, "learning_rate": 4.60150482955832e-05, "loss": 0.0085, "step": 12322 }, { "epoch": 0.24, "learning_rate": 4.601440133532597e-05, "loss": 0.0027, "step": 12324 }, { "epoch": 0.24, "learning_rate": 4.601375437506874e-05, "loss": 0.0008, "step": 12326 }, { "epoch": 0.24, "learning_rate": 4.601310741481151e-05, "loss": 0.0014, "step": 12328 }, { "epoch": 0.24, "learning_rate": 4.601246045455428e-05, "loss": 0.0166, "step": 12330 }, { "epoch": 0.24, "learning_rate": 4.601181349429705e-05, "loss": 0.002, "step": 12332 }, { "epoch": 0.24, "learning_rate": 4.601116653403982e-05, "loss": 0.0017, "step": 12334 }, { "epoch": 0.24, "learning_rate": 4.6010519573782586e-05, "loss": 0.0094, "step": 12336 }, { "epoch": 0.24, "learning_rate": 4.600987261352535e-05, "loss": 0.001, "step": 12338 }, { "epoch": 0.24, "learning_rate": 4.6009225653268125e-05, "loss": 0.0156, "step": 12340 }, { "epoch": 0.24, "learning_rate": 4.600857869301089e-05, "loss": 0.0055, "step": 12342 }, { "epoch": 0.24, "learning_rate": 4.600793173275366e-05, "loss": 0.0055, "step": 12344 }, { "epoch": 0.24, "learning_rate": 4.600728477249643e-05, "loss": 0.0158, "step": 12346 }, { "epoch": 0.24, "learning_rate": 4.6006637812239194e-05, "loss": 0.0008, "step": 12348 }, { "epoch": 0.24, "learning_rate": 4.600599085198197e-05, "loss": 0.0007, "step": 12350 }, { "epoch": 0.24, "learning_rate": 4.600534389172473e-05, "loss": 0.0187, "step": 12352 }, { "epoch": 0.24, "learning_rate": 4.60046969314675e-05, "loss": 0.012, "step": 12354 }, { "epoch": 0.24, "learning_rate": 4.600404997121027e-05, "loss": 0.0009, "step": 12356 }, { "epoch": 0.24, "learning_rate": 4.600340301095304e-05, "loss": 0.0283, "step": 12358 }, { "epoch": 0.24, "learning_rate": 4.600275605069581e-05, "loss": 0.0134, "step": 12360 }, { "epoch": 0.24, "learning_rate": 4.600210909043858e-05, "loss": 0.0016, "step": 12362 }, { "epoch": 0.24, "learning_rate": 4.600146213018135e-05, "loss": 0.0072, "step": 12364 }, { "epoch": 0.24, "learning_rate": 4.6000815169924116e-05, "loss": 0.0085, "step": 12366 }, { "epoch": 0.24, "learning_rate": 4.6000168209666885e-05, "loss": 0.0115, "step": 12368 }, { "epoch": 0.24, "learning_rate": 4.599952124940965e-05, "loss": 0.0028, "step": 12370 }, { "epoch": 0.24, "learning_rate": 4.599887428915242e-05, "loss": 0.0005, "step": 12372 }, { "epoch": 0.24, "learning_rate": 4.5998227328895185e-05, "loss": 0.0018, "step": 12374 }, { "epoch": 0.24, "learning_rate": 4.5997580368637955e-05, "loss": 0.0035, "step": 12376 }, { "epoch": 0.24, "learning_rate": 4.5996933408380724e-05, "loss": 0.0018, "step": 12378 }, { "epoch": 0.24, "learning_rate": 4.599628644812349e-05, "loss": 0.0028, "step": 12380 }, { "epoch": 0.24, "learning_rate": 4.599563948786626e-05, "loss": 0.0007, "step": 12382 }, { "epoch": 0.24, "learning_rate": 4.599499252760903e-05, "loss": 0.0051, "step": 12384 }, { "epoch": 0.24, "learning_rate": 4.59943455673518e-05, "loss": 0.0023, "step": 12386 }, { "epoch": 0.24, "learning_rate": 4.599369860709457e-05, "loss": 0.0029, "step": 12388 }, { "epoch": 0.24, "learning_rate": 4.599305164683734e-05, "loss": 0.0047, "step": 12390 }, { "epoch": 0.24, "learning_rate": 4.599240468658011e-05, "loss": 0.0041, "step": 12392 }, { "epoch": 0.24, "learning_rate": 4.5991757726322876e-05, "loss": 0.0299, "step": 12394 }, { "epoch": 0.24, "learning_rate": 4.5991110766065645e-05, "loss": 0.0013, "step": 12396 }, { "epoch": 0.24, "learning_rate": 4.599046380580841e-05, "loss": 0.0018, "step": 12398 }, { "epoch": 0.24, "learning_rate": 4.5989816845551184e-05, "loss": 0.0038, "step": 12400 }, { "epoch": 0.24, "learning_rate": 4.5989169885293946e-05, "loss": 0.0009, "step": 12402 }, { "epoch": 0.24, "learning_rate": 4.5988522925036715e-05, "loss": 0.013, "step": 12404 }, { "epoch": 0.24, "learning_rate": 4.5987875964779484e-05, "loss": 0.0024, "step": 12406 }, { "epoch": 0.24, "learning_rate": 4.598722900452225e-05, "loss": 0.0061, "step": 12408 }, { "epoch": 0.24, "learning_rate": 4.598658204426502e-05, "loss": 0.0055, "step": 12410 }, { "epoch": 0.24, "learning_rate": 4.598593508400779e-05, "loss": 0.0014, "step": 12412 }, { "epoch": 0.24, "learning_rate": 4.598528812375056e-05, "loss": 0.0076, "step": 12414 }, { "epoch": 0.24, "learning_rate": 4.598464116349333e-05, "loss": 0.0008, "step": 12416 }, { "epoch": 0.24, "learning_rate": 4.59839942032361e-05, "loss": 0.0215, "step": 12418 }, { "epoch": 0.24, "learning_rate": 4.598334724297886e-05, "loss": 0.0066, "step": 12420 }, { "epoch": 0.24, "learning_rate": 4.598270028272164e-05, "loss": 0.005, "step": 12422 }, { "epoch": 0.24, "learning_rate": 4.598237680259302e-05, "loss": 0.2539, "step": 12424 }, { "epoch": 0.24, "learning_rate": 4.5981729842335784e-05, "loss": 0.0099, "step": 12426 }, { "epoch": 0.24, "learning_rate": 4.598108288207856e-05, "loss": 0.0096, "step": 12428 }, { "epoch": 0.24, "learning_rate": 4.598043592182132e-05, "loss": 0.0725, "step": 12430 }, { "epoch": 0.24, "learning_rate": 4.597978896156409e-05, "loss": 0.0045, "step": 12432 }, { "epoch": 0.24, "learning_rate": 4.597914200130686e-05, "loss": 0.0123, "step": 12434 }, { "epoch": 0.24, "learning_rate": 4.597849504104963e-05, "loss": 0.0072, "step": 12436 }, { "epoch": 0.24, "learning_rate": 4.5977848080792405e-05, "loss": 0.0004, "step": 12438 }, { "epoch": 0.24, "learning_rate": 4.597720112053517e-05, "loss": 0.0121, "step": 12440 }, { "epoch": 0.24, "learning_rate": 4.5976554160277936e-05, "loss": 0.0153, "step": 12442 }, { "epoch": 0.24, "learning_rate": 4.5975907200020705e-05, "loss": 0.0028, "step": 12444 }, { "epoch": 0.24, "learning_rate": 4.5975260239763474e-05, "loss": 0.0008, "step": 12446 }, { "epoch": 0.24, "learning_rate": 4.5974613279506243e-05, "loss": 0.0205, "step": 12448 }, { "epoch": 0.24, "learning_rate": 4.597396631924901e-05, "loss": 0.0011, "step": 12450 }, { "epoch": 0.24, "learning_rate": 4.597331935899178e-05, "loss": 0.0141, "step": 12452 }, { "epoch": 0.24, "learning_rate": 4.5972672398734544e-05, "loss": 0.0006, "step": 12454 }, { "epoch": 0.24, "learning_rate": 4.597202543847732e-05, "loss": 0.0024, "step": 12456 }, { "epoch": 0.24, "learning_rate": 4.597137847822008e-05, "loss": 0.0155, "step": 12458 }, { "epoch": 0.24, "learning_rate": 4.597073151796286e-05, "loss": 0.0047, "step": 12460 }, { "epoch": 0.24, "learning_rate": 4.597008455770562e-05, "loss": 0.0071, "step": 12462 }, { "epoch": 0.24, "learning_rate": 4.596943759744839e-05, "loss": 0.0028, "step": 12464 }, { "epoch": 0.24, "learning_rate": 4.596879063719116e-05, "loss": 0.0101, "step": 12466 }, { "epoch": 0.24, "learning_rate": 4.596814367693393e-05, "loss": 0.0101, "step": 12468 }, { "epoch": 0.24, "learning_rate": 4.5967496716676697e-05, "loss": 0.0082, "step": 12470 }, { "epoch": 0.24, "learning_rate": 4.5966849756419466e-05, "loss": 0.0012, "step": 12472 }, { "epoch": 0.24, "learning_rate": 4.5966202796162235e-05, "loss": 0.0085, "step": 12474 }, { "epoch": 0.24, "learning_rate": 4.5965555835905e-05, "loss": 0.0008, "step": 12476 }, { "epoch": 0.24, "learning_rate": 4.596490887564777e-05, "loss": 0.0122, "step": 12478 }, { "epoch": 0.24, "learning_rate": 4.5964261915390535e-05, "loss": 0.0015, "step": 12480 }, { "epoch": 0.24, "learning_rate": 4.596361495513331e-05, "loss": 0.0498, "step": 12482 }, { "epoch": 0.24, "learning_rate": 4.596296799487608e-05, "loss": 0.0134, "step": 12484 }, { "epoch": 0.24, "learning_rate": 4.596232103461884e-05, "loss": 0.0094, "step": 12486 }, { "epoch": 0.24, "learning_rate": 4.596167407436162e-05, "loss": 0.0041, "step": 12488 }, { "epoch": 0.24, "learning_rate": 4.596102711410438e-05, "loss": 0.0118, "step": 12490 }, { "epoch": 0.24, "learning_rate": 4.596038015384715e-05, "loss": 0.0027, "step": 12492 }, { "epoch": 0.24, "learning_rate": 4.595973319358992e-05, "loss": 0.0126, "step": 12494 }, { "epoch": 0.24, "learning_rate": 4.595908623333269e-05, "loss": 0.0068, "step": 12496 }, { "epoch": 0.24, "learning_rate": 4.595843927307546e-05, "loss": 0.0007, "step": 12498 }, { "epoch": 0.24, "learning_rate": 4.5957792312818226e-05, "loss": 0.0039, "step": 12500 }, { "epoch": 0.24, "learning_rate": 4.5957145352560995e-05, "loss": 0.0027, "step": 12502 }, { "epoch": 0.24, "learning_rate": 4.5956498392303764e-05, "loss": 0.021, "step": 12504 }, { "epoch": 0.24, "learning_rate": 4.595585143204653e-05, "loss": 0.0022, "step": 12506 }, { "epoch": 0.24, "learning_rate": 4.5955204471789296e-05, "loss": 0.0511, "step": 12508 }, { "epoch": 0.24, "learning_rate": 4.595455751153207e-05, "loss": 0.0009, "step": 12510 }, { "epoch": 0.24, "learning_rate": 4.5953910551274834e-05, "loss": 0.0027, "step": 12512 }, { "epoch": 0.24, "learning_rate": 4.59532635910176e-05, "loss": 0.0006, "step": 12514 }, { "epoch": 0.24, "learning_rate": 4.595261663076038e-05, "loss": 0.0006, "step": 12516 }, { "epoch": 0.24, "learning_rate": 4.595196967050314e-05, "loss": 0.0031, "step": 12518 }, { "epoch": 0.24, "learning_rate": 4.595132271024592e-05, "loss": 0.0081, "step": 12520 }, { "epoch": 0.24, "learning_rate": 4.595067574998868e-05, "loss": 0.0018, "step": 12522 }, { "epoch": 0.24, "learning_rate": 4.595002878973145e-05, "loss": 0.0051, "step": 12524 }, { "epoch": 0.24, "learning_rate": 4.594938182947422e-05, "loss": 0.001, "step": 12526 }, { "epoch": 0.24, "learning_rate": 4.5948734869216987e-05, "loss": 0.0087, "step": 12528 }, { "epoch": 0.24, "learning_rate": 4.5948087908959756e-05, "loss": 0.0208, "step": 12530 }, { "epoch": 0.24, "learning_rate": 4.5947440948702525e-05, "loss": 0.0075, "step": 12532 }, { "epoch": 0.24, "learning_rate": 4.5946793988445294e-05, "loss": 0.0031, "step": 12534 }, { "epoch": 0.24, "learning_rate": 4.5946147028188056e-05, "loss": 0.0065, "step": 12536 }, { "epoch": 0.24, "learning_rate": 4.594550006793083e-05, "loss": 0.0035, "step": 12538 }, { "epoch": 0.24, "learning_rate": 4.5944853107673594e-05, "loss": 0.009, "step": 12540 }, { "epoch": 0.24, "learning_rate": 4.594420614741637e-05, "loss": 0.0352, "step": 12542 }, { "epoch": 0.24, "learning_rate": 4.594355918715913e-05, "loss": 0.037, "step": 12544 }, { "epoch": 0.24, "learning_rate": 4.59429122269019e-05, "loss": 0.0026, "step": 12546 }, { "epoch": 0.24, "learning_rate": 4.594226526664468e-05, "loss": 0.0075, "step": 12548 }, { "epoch": 0.24, "learning_rate": 4.594161830638744e-05, "loss": 0.0119, "step": 12550 }, { "epoch": 0.24, "learning_rate": 4.594097134613021e-05, "loss": 0.0063, "step": 12552 }, { "epoch": 0.24, "learning_rate": 4.594032438587298e-05, "loss": 0.0012, "step": 12554 }, { "epoch": 0.24, "learning_rate": 4.593967742561575e-05, "loss": 0.0018, "step": 12556 }, { "epoch": 0.24, "learning_rate": 4.5939030465358516e-05, "loss": 0.0061, "step": 12558 }, { "epoch": 0.24, "learning_rate": 4.5938383505101285e-05, "loss": 0.0033, "step": 12560 }, { "epoch": 0.24, "learning_rate": 4.5937736544844054e-05, "loss": 0.0006, "step": 12562 }, { "epoch": 0.24, "learning_rate": 4.593708958458682e-05, "loss": 0.0105, "step": 12564 }, { "epoch": 0.24, "learning_rate": 4.593644262432959e-05, "loss": 0.0005, "step": 12566 }, { "epoch": 0.24, "learning_rate": 4.5935795664072355e-05, "loss": 0.0011, "step": 12568 }, { "epoch": 0.24, "learning_rate": 4.593514870381513e-05, "loss": 0.0017, "step": 12570 }, { "epoch": 0.24, "learning_rate": 4.593450174355789e-05, "loss": 0.0326, "step": 12572 }, { "epoch": 0.24, "learning_rate": 4.593385478330066e-05, "loss": 0.0028, "step": 12574 }, { "epoch": 0.24, "learning_rate": 4.593320782304343e-05, "loss": 0.0113, "step": 12576 }, { "epoch": 0.24, "learning_rate": 4.59325608627862e-05, "loss": 0.0102, "step": 12578 }, { "epoch": 0.24, "learning_rate": 4.593191390252897e-05, "loss": 0.0008, "step": 12580 }, { "epoch": 0.24, "learning_rate": 4.593126694227174e-05, "loss": 0.0452, "step": 12582 }, { "epoch": 0.24, "learning_rate": 4.593061998201451e-05, "loss": 0.0064, "step": 12584 }, { "epoch": 0.24, "learning_rate": 4.5929973021757276e-05, "loss": 0.0039, "step": 12586 }, { "epoch": 0.24, "learning_rate": 4.5929326061500045e-05, "loss": 0.0114, "step": 12588 }, { "epoch": 0.24, "learning_rate": 4.592867910124281e-05, "loss": 0.0108, "step": 12590 }, { "epoch": 0.24, "learning_rate": 4.5928032140985584e-05, "loss": 0.0033, "step": 12592 }, { "epoch": 0.24, "learning_rate": 4.592738518072835e-05, "loss": 0.02, "step": 12594 }, { "epoch": 0.24, "learning_rate": 4.5926738220471115e-05, "loss": 0.0029, "step": 12596 }, { "epoch": 0.24, "learning_rate": 4.592609126021389e-05, "loss": 0.0038, "step": 12598 }, { "epoch": 0.24, "learning_rate": 4.592544429995665e-05, "loss": 0.0057, "step": 12600 }, { "epoch": 0.24, "learning_rate": 4.592479733969943e-05, "loss": 0.0058, "step": 12602 }, { "epoch": 0.24, "learning_rate": 4.592415037944219e-05, "loss": 0.0035, "step": 12604 }, { "epoch": 0.24, "learning_rate": 4.592350341918496e-05, "loss": 0.0084, "step": 12606 }, { "epoch": 0.24, "learning_rate": 4.592285645892773e-05, "loss": 0.0006, "step": 12608 }, { "epoch": 0.24, "learning_rate": 4.59222094986705e-05, "loss": 0.0091, "step": 12610 }, { "epoch": 0.24, "learning_rate": 4.592156253841327e-05, "loss": 0.0285, "step": 12612 }, { "epoch": 0.24, "learning_rate": 4.592091557815604e-05, "loss": 0.0084, "step": 12614 }, { "epoch": 0.24, "learning_rate": 4.5920268617898806e-05, "loss": 0.0021, "step": 12616 }, { "epoch": 0.24, "learning_rate": 4.591962165764157e-05, "loss": 0.0043, "step": 12618 }, { "epoch": 0.24, "learning_rate": 4.5918974697384344e-05, "loss": 0.0194, "step": 12620 }, { "epoch": 0.24, "learning_rate": 4.5918327737127106e-05, "loss": 0.0131, "step": 12622 }, { "epoch": 0.25, "learning_rate": 4.591768077686988e-05, "loss": 0.0021, "step": 12624 }, { "epoch": 0.25, "learning_rate": 4.5917033816612645e-05, "loss": 0.0106, "step": 12626 }, { "epoch": 0.25, "learning_rate": 4.5916386856355414e-05, "loss": 0.0169, "step": 12628 }, { "epoch": 0.25, "learning_rate": 4.591573989609819e-05, "loss": 0.0068, "step": 12630 }, { "epoch": 0.25, "learning_rate": 4.591509293584095e-05, "loss": 0.009, "step": 12632 }, { "epoch": 0.25, "learning_rate": 4.591444597558372e-05, "loss": 0.0098, "step": 12634 }, { "epoch": 0.25, "learning_rate": 4.591379901532649e-05, "loss": 0.0019, "step": 12636 }, { "epoch": 0.25, "learning_rate": 4.591315205506926e-05, "loss": 0.0212, "step": 12638 }, { "epoch": 0.25, "learning_rate": 4.591250509481203e-05, "loss": 0.0393, "step": 12640 }, { "epoch": 0.25, "learning_rate": 4.59118581345548e-05, "loss": 0.0342, "step": 12642 }, { "epoch": 0.25, "learning_rate": 4.5911211174297566e-05, "loss": 0.0131, "step": 12644 }, { "epoch": 0.25, "learning_rate": 4.5910564214040335e-05, "loss": 0.0033, "step": 12646 }, { "epoch": 0.25, "learning_rate": 4.5909917253783104e-05, "loss": 0.0192, "step": 12648 }, { "epoch": 0.25, "learning_rate": 4.590927029352587e-05, "loss": 0.0257, "step": 12650 }, { "epoch": 0.25, "learning_rate": 4.590862333326864e-05, "loss": 0.0033, "step": 12652 }, { "epoch": 0.25, "learning_rate": 4.5907976373011405e-05, "loss": 0.0093, "step": 12654 }, { "epoch": 0.25, "learning_rate": 4.5907329412754174e-05, "loss": 0.0046, "step": 12656 }, { "epoch": 0.25, "learning_rate": 4.590668245249694e-05, "loss": 0.0109, "step": 12658 }, { "epoch": 0.25, "learning_rate": 4.590603549223971e-05, "loss": 0.001, "step": 12660 }, { "epoch": 0.25, "learning_rate": 4.590538853198249e-05, "loss": 0.0036, "step": 12662 }, { "epoch": 0.25, "learning_rate": 4.590474157172525e-05, "loss": 0.0357, "step": 12664 }, { "epoch": 0.25, "learning_rate": 4.590409461146802e-05, "loss": 0.041, "step": 12666 }, { "epoch": 0.25, "learning_rate": 4.590344765121079e-05, "loss": 0.0419, "step": 12668 }, { "epoch": 0.25, "learning_rate": 4.590280069095356e-05, "loss": 0.0149, "step": 12670 }, { "epoch": 0.25, "learning_rate": 4.590215373069633e-05, "loss": 0.0081, "step": 12672 }, { "epoch": 0.25, "learning_rate": 4.5901506770439096e-05, "loss": 0.0241, "step": 12674 }, { "epoch": 0.25, "learning_rate": 4.5900859810181865e-05, "loss": 0.0014, "step": 12676 }, { "epoch": 0.25, "learning_rate": 4.590021284992463e-05, "loss": 0.0373, "step": 12678 }, { "epoch": 0.25, "learning_rate": 4.58995658896674e-05, "loss": 0.0104, "step": 12680 }, { "epoch": 0.25, "learning_rate": 4.5898918929410165e-05, "loss": 0.0008, "step": 12682 }, { "epoch": 0.25, "learning_rate": 4.589827196915294e-05, "loss": 0.0051, "step": 12684 }, { "epoch": 0.25, "learning_rate": 4.5897625008895704e-05, "loss": 0.003, "step": 12686 }, { "epoch": 0.25, "learning_rate": 4.589697804863847e-05, "loss": 0.0376, "step": 12688 }, { "epoch": 0.25, "learning_rate": 4.589633108838124e-05, "loss": 0.0255, "step": 12690 }, { "epoch": 0.25, "learning_rate": 4.589568412812401e-05, "loss": 0.0087, "step": 12692 }, { "epoch": 0.25, "learning_rate": 4.589503716786678e-05, "loss": 0.0194, "step": 12694 }, { "epoch": 0.25, "learning_rate": 4.589439020760955e-05, "loss": 0.0118, "step": 12696 }, { "epoch": 0.25, "learning_rate": 4.589374324735232e-05, "loss": 0.0061, "step": 12698 }, { "epoch": 0.25, "learning_rate": 4.589309628709509e-05, "loss": 0.0038, "step": 12700 }, { "epoch": 0.25, "learning_rate": 4.5892449326837856e-05, "loss": 0.0032, "step": 12702 }, { "epoch": 0.25, "learning_rate": 4.589180236658062e-05, "loss": 0.011, "step": 12704 }, { "epoch": 0.25, "learning_rate": 4.5891155406323394e-05, "loss": 0.0065, "step": 12706 }, { "epoch": 0.25, "learning_rate": 4.5890508446066163e-05, "loss": 0.0024, "step": 12708 }, { "epoch": 0.25, "learning_rate": 4.5889861485808926e-05, "loss": 0.0092, "step": 12710 }, { "epoch": 0.25, "learning_rate": 4.58892145255517e-05, "loss": 0.005, "step": 12712 }, { "epoch": 0.25, "learning_rate": 4.5888567565294464e-05, "loss": 0.003, "step": 12714 }, { "epoch": 0.25, "learning_rate": 4.588792060503723e-05, "loss": 0.0025, "step": 12716 }, { "epoch": 0.25, "learning_rate": 4.588727364478e-05, "loss": 0.01, "step": 12718 }, { "epoch": 0.25, "learning_rate": 4.588662668452277e-05, "loss": 0.036, "step": 12720 }, { "epoch": 0.25, "learning_rate": 4.588597972426554e-05, "loss": 0.0003, "step": 12722 }, { "epoch": 0.25, "learning_rate": 4.588533276400831e-05, "loss": 0.0035, "step": 12724 }, { "epoch": 0.25, "learning_rate": 4.588468580375108e-05, "loss": 0.002, "step": 12726 }, { "epoch": 0.25, "learning_rate": 4.588403884349385e-05, "loss": 0.0103, "step": 12728 }, { "epoch": 0.25, "learning_rate": 4.5883391883236617e-05, "loss": 0.0042, "step": 12730 }, { "epoch": 0.25, "learning_rate": 4.588274492297938e-05, "loss": 0.0018, "step": 12732 }, { "epoch": 0.25, "learning_rate": 4.5882097962722155e-05, "loss": 0.0012, "step": 12734 }, { "epoch": 0.25, "learning_rate": 4.588145100246492e-05, "loss": 0.0007, "step": 12736 }, { "epoch": 0.25, "learning_rate": 4.5880804042207686e-05, "loss": 0.0821, "step": 12738 }, { "epoch": 0.25, "learning_rate": 4.588015708195046e-05, "loss": 0.0004, "step": 12740 }, { "epoch": 0.25, "learning_rate": 4.5879510121693224e-05, "loss": 0.0117, "step": 12742 }, { "epoch": 0.25, "learning_rate": 4.5878863161436e-05, "loss": 0.0024, "step": 12744 }, { "epoch": 0.25, "learning_rate": 4.587821620117876e-05, "loss": 0.0124, "step": 12746 }, { "epoch": 0.25, "learning_rate": 4.587756924092153e-05, "loss": 0.024, "step": 12748 }, { "epoch": 0.25, "learning_rate": 4.58769222806643e-05, "loss": 0.0093, "step": 12750 }, { "epoch": 0.25, "learning_rate": 4.587627532040707e-05, "loss": 0.0181, "step": 12752 }, { "epoch": 0.25, "learning_rate": 4.587562836014984e-05, "loss": 0.0108, "step": 12754 }, { "epoch": 0.25, "learning_rate": 4.587498139989261e-05, "loss": 0.0003, "step": 12756 }, { "epoch": 0.25, "learning_rate": 4.587433443963538e-05, "loss": 0.0123, "step": 12758 }, { "epoch": 0.25, "learning_rate": 4.587368747937814e-05, "loss": 0.0007, "step": 12760 }, { "epoch": 0.25, "learning_rate": 4.5873040519120915e-05, "loss": 0.0016, "step": 12762 }, { "epoch": 0.25, "learning_rate": 4.587239355886368e-05, "loss": 0.0012, "step": 12764 }, { "epoch": 0.25, "learning_rate": 4.587174659860645e-05, "loss": 0.0034, "step": 12766 }, { "epoch": 0.25, "learning_rate": 4.5871099638349216e-05, "loss": 0.0113, "step": 12768 }, { "epoch": 0.25, "learning_rate": 4.5870452678091985e-05, "loss": 0.0322, "step": 12770 }, { "epoch": 0.25, "learning_rate": 4.586980571783476e-05, "loss": 0.0034, "step": 12772 }, { "epoch": 0.25, "learning_rate": 4.586915875757752e-05, "loss": 0.0013, "step": 12774 }, { "epoch": 0.25, "learning_rate": 4.586851179732029e-05, "loss": 0.0029, "step": 12776 }, { "epoch": 0.25, "learning_rate": 4.586786483706306e-05, "loss": 0.0009, "step": 12778 }, { "epoch": 0.25, "learning_rate": 4.586721787680583e-05, "loss": 0.009, "step": 12780 }, { "epoch": 0.25, "learning_rate": 4.58665709165486e-05, "loss": 0.0018, "step": 12782 }, { "epoch": 0.25, "learning_rate": 4.586592395629137e-05, "loss": 0.0118, "step": 12784 }, { "epoch": 0.25, "learning_rate": 4.586527699603414e-05, "loss": 0.0201, "step": 12786 }, { "epoch": 0.25, "learning_rate": 4.5864630035776906e-05, "loss": 0.0128, "step": 12788 }, { "epoch": 0.25, "learning_rate": 4.5863983075519676e-05, "loss": 0.0151, "step": 12790 }, { "epoch": 0.25, "learning_rate": 4.586333611526244e-05, "loss": 0.0016, "step": 12792 }, { "epoch": 0.25, "learning_rate": 4.5862689155005214e-05, "loss": 0.0071, "step": 12794 }, { "epoch": 0.25, "learning_rate": 4.5862042194747976e-05, "loss": 0.0169, "step": 12796 }, { "epoch": 0.25, "learning_rate": 4.5861395234490745e-05, "loss": 0.0075, "step": 12798 }, { "epoch": 0.25, "learning_rate": 4.5860748274233514e-05, "loss": 0.0122, "step": 12800 }, { "epoch": 0.25, "learning_rate": 4.586010131397628e-05, "loss": 0.0443, "step": 12802 }, { "epoch": 0.25, "learning_rate": 4.585945435371905e-05, "loss": 0.0068, "step": 12804 }, { "epoch": 0.25, "learning_rate": 4.585880739346182e-05, "loss": 0.0082, "step": 12806 }, { "epoch": 0.25, "learning_rate": 4.585816043320459e-05, "loss": 0.0011, "step": 12808 }, { "epoch": 0.25, "learning_rate": 4.585751347294736e-05, "loss": 0.0061, "step": 12810 }, { "epoch": 0.25, "learning_rate": 4.585686651269013e-05, "loss": 0.0032, "step": 12812 }, { "epoch": 0.25, "learning_rate": 4.585621955243289e-05, "loss": 0.0002, "step": 12814 }, { "epoch": 0.25, "learning_rate": 4.585557259217567e-05, "loss": 0.0041, "step": 12816 }, { "epoch": 0.25, "learning_rate": 4.5854925631918436e-05, "loss": 0.0008, "step": 12818 }, { "epoch": 0.25, "learning_rate": 4.58542786716612e-05, "loss": 0.0412, "step": 12820 }, { "epoch": 0.25, "learning_rate": 4.5853631711403974e-05, "loss": 0.0152, "step": 12822 }, { "epoch": 0.25, "learning_rate": 4.5852984751146736e-05, "loss": 0.0017, "step": 12824 }, { "epoch": 0.25, "learning_rate": 4.585233779088951e-05, "loss": 0.0025, "step": 12826 }, { "epoch": 0.25, "learning_rate": 4.5851690830632275e-05, "loss": 0.0093, "step": 12828 }, { "epoch": 0.25, "learning_rate": 4.5851043870375044e-05, "loss": 0.0391, "step": 12830 }, { "epoch": 0.25, "learning_rate": 4.585039691011781e-05, "loss": 0.0166, "step": 12832 }, { "epoch": 0.25, "learning_rate": 4.584974994986058e-05, "loss": 0.0149, "step": 12834 }, { "epoch": 0.25, "learning_rate": 4.584910298960335e-05, "loss": 0.001, "step": 12836 }, { "epoch": 0.25, "learning_rate": 4.584845602934612e-05, "loss": 0.0069, "step": 12838 }, { "epoch": 0.25, "learning_rate": 4.584780906908889e-05, "loss": 0.006, "step": 12840 }, { "epoch": 0.25, "learning_rate": 4.584716210883166e-05, "loss": 0.016, "step": 12842 }, { "epoch": 0.25, "learning_rate": 4.584651514857443e-05, "loss": 0.0037, "step": 12844 }, { "epoch": 0.25, "learning_rate": 4.584586818831719e-05, "loss": 0.0111, "step": 12846 }, { "epoch": 0.25, "learning_rate": 4.5845221228059965e-05, "loss": 0.0082, "step": 12848 }, { "epoch": 0.25, "learning_rate": 4.5844574267802735e-05, "loss": 0.0068, "step": 12850 }, { "epoch": 0.25, "learning_rate": 4.58439273075455e-05, "loss": 0.0031, "step": 12852 }, { "epoch": 0.25, "learning_rate": 4.584328034728827e-05, "loss": 0.002, "step": 12854 }, { "epoch": 0.25, "learning_rate": 4.5842633387031035e-05, "loss": 0.0046, "step": 12856 }, { "epoch": 0.25, "learning_rate": 4.5841986426773804e-05, "loss": 0.0016, "step": 12858 }, { "epoch": 0.25, "learning_rate": 4.584133946651657e-05, "loss": 0.0108, "step": 12860 }, { "epoch": 0.25, "learning_rate": 4.584069250625934e-05, "loss": 0.0304, "step": 12862 }, { "epoch": 0.25, "learning_rate": 4.584004554600211e-05, "loss": 0.0017, "step": 12864 }, { "epoch": 0.25, "learning_rate": 4.583939858574488e-05, "loss": 0.0058, "step": 12866 }, { "epoch": 0.25, "learning_rate": 4.583875162548765e-05, "loss": 0.0127, "step": 12868 }, { "epoch": 0.25, "learning_rate": 4.583810466523042e-05, "loss": 0.0009, "step": 12870 }, { "epoch": 0.25, "learning_rate": 4.583745770497319e-05, "loss": 0.0003, "step": 12872 }, { "epoch": 0.25, "learning_rate": 4.583681074471595e-05, "loss": 0.0296, "step": 12874 }, { "epoch": 0.25, "learning_rate": 4.5836163784458726e-05, "loss": 0.003, "step": 12876 }, { "epoch": 0.25, "learning_rate": 4.583551682420149e-05, "loss": 0.0008, "step": 12878 }, { "epoch": 0.25, "learning_rate": 4.583486986394426e-05, "loss": 0.0025, "step": 12880 }, { "epoch": 0.25, "learning_rate": 4.5834222903687026e-05, "loss": 0.0008, "step": 12882 }, { "epoch": 0.25, "learning_rate": 4.5833575943429795e-05, "loss": 0.0128, "step": 12884 }, { "epoch": 0.25, "learning_rate": 4.583292898317257e-05, "loss": 0.0008, "step": 12886 }, { "epoch": 0.25, "learning_rate": 4.5832282022915334e-05, "loss": 0.0057, "step": 12888 }, { "epoch": 0.25, "learning_rate": 4.58316350626581e-05, "loss": 0.0056, "step": 12890 }, { "epoch": 0.25, "learning_rate": 4.583098810240087e-05, "loss": 0.012, "step": 12892 }, { "epoch": 0.25, "learning_rate": 4.583034114214364e-05, "loss": 0.0006, "step": 12894 }, { "epoch": 0.25, "learning_rate": 4.582969418188641e-05, "loss": 0.0355, "step": 12896 }, { "epoch": 0.25, "learning_rate": 4.582904722162918e-05, "loss": 0.0154, "step": 12898 }, { "epoch": 0.25, "learning_rate": 4.582840026137195e-05, "loss": 0.0083, "step": 12900 }, { "epoch": 0.25, "learning_rate": 4.582775330111471e-05, "loss": 0.0007, "step": 12902 }, { "epoch": 0.25, "learning_rate": 4.5827106340857486e-05, "loss": 0.0219, "step": 12904 }, { "epoch": 0.25, "learning_rate": 4.582645938060025e-05, "loss": 0.001, "step": 12906 }, { "epoch": 0.25, "learning_rate": 4.5825812420343024e-05, "loss": 0.0091, "step": 12908 }, { "epoch": 0.25, "learning_rate": 4.582516546008579e-05, "loss": 0.0081, "step": 12910 }, { "epoch": 0.25, "learning_rate": 4.5824518499828556e-05, "loss": 0.0053, "step": 12912 }, { "epoch": 0.25, "learning_rate": 4.5823871539571325e-05, "loss": 0.0038, "step": 12914 }, { "epoch": 0.25, "learning_rate": 4.5823224579314094e-05, "loss": 0.0142, "step": 12916 }, { "epoch": 0.25, "learning_rate": 4.582257761905686e-05, "loss": 0.0043, "step": 12918 }, { "epoch": 0.25, "learning_rate": 4.582193065879963e-05, "loss": 0.0012, "step": 12920 }, { "epoch": 0.25, "learning_rate": 4.58212836985424e-05, "loss": 0.0032, "step": 12922 }, { "epoch": 0.25, "learning_rate": 4.582063673828517e-05, "loss": 0.0004, "step": 12924 }, { "epoch": 0.25, "learning_rate": 4.581998977802794e-05, "loss": 0.0182, "step": 12926 }, { "epoch": 0.25, "learning_rate": 4.581934281777071e-05, "loss": 0.0027, "step": 12928 }, { "epoch": 0.25, "learning_rate": 4.581869585751348e-05, "loss": 0.0014, "step": 12930 }, { "epoch": 0.25, "learning_rate": 4.581804889725625e-05, "loss": 0.015, "step": 12932 }, { "epoch": 0.25, "learning_rate": 4.581740193699901e-05, "loss": 0.0246, "step": 12934 }, { "epoch": 0.25, "learning_rate": 4.5816754976741785e-05, "loss": 0.0023, "step": 12936 }, { "epoch": 0.25, "learning_rate": 4.581610801648455e-05, "loss": 0.0014, "step": 12938 }, { "epoch": 0.25, "learning_rate": 4.5815461056227316e-05, "loss": 0.064, "step": 12940 }, { "epoch": 0.25, "learning_rate": 4.5814814095970085e-05, "loss": 0.0012, "step": 12942 }, { "epoch": 0.25, "learning_rate": 4.5814167135712854e-05, "loss": 0.0025, "step": 12944 }, { "epoch": 0.25, "learning_rate": 4.5813520175455623e-05, "loss": 0.0084, "step": 12946 }, { "epoch": 0.25, "learning_rate": 4.581287321519839e-05, "loss": 0.0034, "step": 12948 }, { "epoch": 0.25, "learning_rate": 4.581222625494116e-05, "loss": 0.0016, "step": 12950 }, { "epoch": 0.25, "learning_rate": 4.581157929468393e-05, "loss": 0.0012, "step": 12952 }, { "epoch": 0.25, "learning_rate": 4.58109323344267e-05, "loss": 0.0065, "step": 12954 }, { "epoch": 0.25, "learning_rate": 4.581028537416946e-05, "loss": 0.0008, "step": 12956 }, { "epoch": 0.25, "learning_rate": 4.580963841391224e-05, "loss": 0.0044, "step": 12958 }, { "epoch": 0.25, "learning_rate": 4.5808991453655e-05, "loss": 0.0132, "step": 12960 }, { "epoch": 0.25, "learning_rate": 4.580834449339777e-05, "loss": 0.002, "step": 12962 }, { "epoch": 0.25, "learning_rate": 4.5807697533140545e-05, "loss": 0.0081, "step": 12964 }, { "epoch": 0.25, "learning_rate": 4.580705057288331e-05, "loss": 0.0053, "step": 12966 }, { "epoch": 0.25, "learning_rate": 4.5806403612626083e-05, "loss": 0.0014, "step": 12968 }, { "epoch": 0.25, "learning_rate": 4.5805756652368846e-05, "loss": 0.0025, "step": 12970 }, { "epoch": 0.25, "learning_rate": 4.5805109692111615e-05, "loss": 0.0036, "step": 12972 }, { "epoch": 0.25, "learning_rate": 4.5804462731854384e-05, "loss": 0.0033, "step": 12974 }, { "epoch": 0.25, "learning_rate": 4.580381577159715e-05, "loss": 0.0068, "step": 12976 }, { "epoch": 0.25, "learning_rate": 4.580316881133992e-05, "loss": 0.0036, "step": 12978 }, { "epoch": 0.25, "learning_rate": 4.580252185108269e-05, "loss": 0.0014, "step": 12980 }, { "epoch": 0.25, "learning_rate": 4.580187489082546e-05, "loss": 0.0087, "step": 12982 }, { "epoch": 0.25, "learning_rate": 4.580122793056823e-05, "loss": 0.0184, "step": 12984 }, { "epoch": 0.25, "learning_rate": 4.5800580970311e-05, "loss": 0.0022, "step": 12986 }, { "epoch": 0.25, "learning_rate": 4.579993401005376e-05, "loss": 0.0146, "step": 12988 }, { "epoch": 0.25, "learning_rate": 4.5799287049796537e-05, "loss": 0.0049, "step": 12990 }, { "epoch": 0.25, "learning_rate": 4.57986400895393e-05, "loss": 0.003, "step": 12992 }, { "epoch": 0.25, "learning_rate": 4.579799312928207e-05, "loss": 0.0062, "step": 12994 }, { "epoch": 0.25, "learning_rate": 4.5797346169024844e-05, "loss": 0.0035, "step": 12996 }, { "epoch": 0.25, "learning_rate": 4.5796699208767606e-05, "loss": 0.0204, "step": 12998 }, { "epoch": 0.25, "learning_rate": 4.5796052248510375e-05, "loss": 0.0081, "step": 13000 }, { "epoch": 0.25, "learning_rate": 4.5795405288253144e-05, "loss": 0.0075, "step": 13002 }, { "epoch": 0.25, "learning_rate": 4.579475832799591e-05, "loss": 0.002, "step": 13004 }, { "epoch": 0.25, "learning_rate": 4.579411136773868e-05, "loss": 0.0159, "step": 13006 }, { "epoch": 0.25, "learning_rate": 4.579346440748145e-05, "loss": 0.001, "step": 13008 }, { "epoch": 0.25, "learning_rate": 4.579281744722422e-05, "loss": 0.0185, "step": 13010 }, { "epoch": 0.25, "learning_rate": 4.579217048696699e-05, "loss": 0.008, "step": 13012 }, { "epoch": 0.25, "learning_rate": 4.579152352670976e-05, "loss": 0.011, "step": 13014 }, { "epoch": 0.25, "learning_rate": 4.579087656645252e-05, "loss": 0.0213, "step": 13016 }, { "epoch": 0.25, "learning_rate": 4.57902296061953e-05, "loss": 0.0146, "step": 13018 }, { "epoch": 0.25, "learning_rate": 4.578958264593806e-05, "loss": 0.0102, "step": 13020 }, { "epoch": 0.25, "learning_rate": 4.578893568568083e-05, "loss": 0.0026, "step": 13022 }, { "epoch": 0.25, "learning_rate": 4.57882887254236e-05, "loss": 0.0205, "step": 13024 }, { "epoch": 0.25, "learning_rate": 4.5787641765166367e-05, "loss": 0.0111, "step": 13026 }, { "epoch": 0.25, "learning_rate": 4.578699480490914e-05, "loss": 0.0005, "step": 13028 }, { "epoch": 0.25, "learning_rate": 4.5786347844651905e-05, "loss": 0.0007, "step": 13030 }, { "epoch": 0.25, "learning_rate": 4.5785700884394674e-05, "loss": 0.0111, "step": 13032 }, { "epoch": 0.25, "learning_rate": 4.578505392413744e-05, "loss": 0.0009, "step": 13034 }, { "epoch": 0.25, "learning_rate": 4.578440696388021e-05, "loss": 0.0008, "step": 13036 }, { "epoch": 0.25, "learning_rate": 4.5783760003622974e-05, "loss": 0.0366, "step": 13038 }, { "epoch": 0.25, "learning_rate": 4.578311304336575e-05, "loss": 0.0027, "step": 13040 }, { "epoch": 0.25, "learning_rate": 4.578246608310852e-05, "loss": 0.0009, "step": 13042 }, { "epoch": 0.25, "learning_rate": 4.578181912285128e-05, "loss": 0.023, "step": 13044 }, { "epoch": 0.25, "learning_rate": 4.578117216259406e-05, "loss": 0.0075, "step": 13046 }, { "epoch": 0.25, "learning_rate": 4.578052520233682e-05, "loss": 0.0013, "step": 13048 }, { "epoch": 0.25, "learning_rate": 4.5779878242079596e-05, "loss": 0.0076, "step": 13050 }, { "epoch": 0.25, "learning_rate": 4.577923128182236e-05, "loss": 0.0062, "step": 13052 }, { "epoch": 0.25, "learning_rate": 4.577858432156513e-05, "loss": 0.0006, "step": 13054 }, { "epoch": 0.25, "learning_rate": 4.5777937361307896e-05, "loss": 0.0615, "step": 13056 }, { "epoch": 0.25, "learning_rate": 4.5777290401050665e-05, "loss": 0.0055, "step": 13058 }, { "epoch": 0.25, "learning_rate": 4.5776643440793434e-05, "loss": 0.0078, "step": 13060 }, { "epoch": 0.25, "learning_rate": 4.57759964805362e-05, "loss": 0.0012, "step": 13062 }, { "epoch": 0.25, "learning_rate": 4.577534952027897e-05, "loss": 0.0004, "step": 13064 }, { "epoch": 0.25, "learning_rate": 4.577470256002174e-05, "loss": 0.0065, "step": 13066 }, { "epoch": 0.25, "learning_rate": 4.577405559976451e-05, "loss": 0.039, "step": 13068 }, { "epoch": 0.25, "learning_rate": 4.577340863950727e-05, "loss": 0.0115, "step": 13070 }, { "epoch": 0.25, "learning_rate": 4.577276167925005e-05, "loss": 0.0031, "step": 13072 }, { "epoch": 0.25, "learning_rate": 4.577211471899282e-05, "loss": 0.0134, "step": 13074 }, { "epoch": 0.25, "learning_rate": 4.577146775873558e-05, "loss": 0.0218, "step": 13076 }, { "epoch": 0.25, "learning_rate": 4.5770820798478356e-05, "loss": 0.0029, "step": 13078 }, { "epoch": 0.25, "learning_rate": 4.577017383822112e-05, "loss": 0.0018, "step": 13080 }, { "epoch": 0.25, "learning_rate": 4.576952687796389e-05, "loss": 0.0092, "step": 13082 }, { "epoch": 0.25, "learning_rate": 4.5768879917706656e-05, "loss": 0.0043, "step": 13084 }, { "epoch": 0.25, "learning_rate": 4.5768232957449426e-05, "loss": 0.0096, "step": 13086 }, { "epoch": 0.25, "learning_rate": 4.5767585997192195e-05, "loss": 0.0073, "step": 13088 }, { "epoch": 0.25, "learning_rate": 4.5766939036934964e-05, "loss": 0.0019, "step": 13090 }, { "epoch": 0.25, "learning_rate": 4.576629207667773e-05, "loss": 0.0012, "step": 13092 }, { "epoch": 0.25, "learning_rate": 4.57656451164205e-05, "loss": 0.0047, "step": 13094 }, { "epoch": 0.25, "learning_rate": 4.576499815616327e-05, "loss": 0.0097, "step": 13096 }, { "epoch": 0.25, "learning_rate": 4.576435119590603e-05, "loss": 0.0016, "step": 13098 }, { "epoch": 0.25, "learning_rate": 4.576370423564881e-05, "loss": 0.0039, "step": 13100 }, { "epoch": 0.25, "learning_rate": 4.576305727539157e-05, "loss": 0.0016, "step": 13102 }, { "epoch": 0.25, "learning_rate": 4.576241031513434e-05, "loss": 0.0011, "step": 13104 }, { "epoch": 0.25, "learning_rate": 4.576176335487711e-05, "loss": 0.0074, "step": 13106 }, { "epoch": 0.25, "learning_rate": 4.576111639461988e-05, "loss": 0.0016, "step": 13108 }, { "epoch": 0.25, "learning_rate": 4.5760469434362655e-05, "loss": 0.0057, "step": 13110 }, { "epoch": 0.25, "learning_rate": 4.575982247410542e-05, "loss": 0.0033, "step": 13112 }, { "epoch": 0.25, "learning_rate": 4.5759175513848186e-05, "loss": 0.0035, "step": 13114 }, { "epoch": 0.25, "learning_rate": 4.5758528553590955e-05, "loss": 0.0154, "step": 13116 }, { "epoch": 0.25, "learning_rate": 4.5757881593333724e-05, "loss": 0.0238, "step": 13118 }, { "epoch": 0.25, "learning_rate": 4.575723463307649e-05, "loss": 0.0015, "step": 13120 }, { "epoch": 0.25, "learning_rate": 4.575658767281926e-05, "loss": 0.0106, "step": 13122 }, { "epoch": 0.25, "learning_rate": 4.575594071256203e-05, "loss": 0.0097, "step": 13124 }, { "epoch": 0.25, "learning_rate": 4.57552937523048e-05, "loss": 0.0041, "step": 13126 }, { "epoch": 0.25, "learning_rate": 4.575464679204757e-05, "loss": 0.0045, "step": 13128 }, { "epoch": 0.25, "learning_rate": 4.575399983179033e-05, "loss": 0.0014, "step": 13130 }, { "epoch": 0.25, "learning_rate": 4.575335287153311e-05, "loss": 0.0017, "step": 13132 }, { "epoch": 0.25, "learning_rate": 4.575270591127587e-05, "loss": 0.0242, "step": 13134 }, { "epoch": 0.25, "learning_rate": 4.575205895101864e-05, "loss": 0.0145, "step": 13136 }, { "epoch": 0.25, "learning_rate": 4.575141199076141e-05, "loss": 0.0023, "step": 13138 }, { "epoch": 0.26, "learning_rate": 4.575076503050418e-05, "loss": 0.0018, "step": 13140 }, { "epoch": 0.26, "learning_rate": 4.5750118070246946e-05, "loss": 0.0023, "step": 13142 }, { "epoch": 0.26, "learning_rate": 4.5749471109989715e-05, "loss": 0.0003, "step": 13144 }, { "epoch": 0.26, "learning_rate": 4.5748824149732484e-05, "loss": 0.0008, "step": 13146 }, { "epoch": 0.26, "learning_rate": 4.5748177189475254e-05, "loss": 0.0046, "step": 13148 }, { "epoch": 0.26, "learning_rate": 4.574753022921802e-05, "loss": 0.0011, "step": 13150 }, { "epoch": 0.26, "learning_rate": 4.574688326896079e-05, "loss": 0.0004, "step": 13152 }, { "epoch": 0.26, "learning_rate": 4.574623630870356e-05, "loss": 0.0043, "step": 13154 }, { "epoch": 0.26, "learning_rate": 4.574558934844633e-05, "loss": 0.0002, "step": 13156 }, { "epoch": 0.26, "learning_rate": 4.574494238818909e-05, "loss": 0.002, "step": 13158 }, { "epoch": 0.26, "learning_rate": 4.574429542793187e-05, "loss": 0.0022, "step": 13160 }, { "epoch": 0.26, "learning_rate": 4.574364846767463e-05, "loss": 0.0067, "step": 13162 }, { "epoch": 0.26, "learning_rate": 4.57430015074174e-05, "loss": 0.014, "step": 13164 }, { "epoch": 0.26, "learning_rate": 4.574235454716017e-05, "loss": 0.0004, "step": 13166 }, { "epoch": 0.26, "learning_rate": 4.574170758690294e-05, "loss": 0.002, "step": 13168 }, { "epoch": 0.26, "learning_rate": 4.574106062664571e-05, "loss": 0.0008, "step": 13170 }, { "epoch": 0.26, "learning_rate": 4.5740413666388476e-05, "loss": 0.0004, "step": 13172 }, { "epoch": 0.26, "learning_rate": 4.5739766706131245e-05, "loss": 0.0027, "step": 13174 }, { "epoch": 0.26, "learning_rate": 4.5739119745874014e-05, "loss": 0.0052, "step": 13176 }, { "epoch": 0.26, "learning_rate": 4.573847278561678e-05, "loss": 0.0009, "step": 13178 }, { "epoch": 0.26, "learning_rate": 4.5737825825359545e-05, "loss": 0.0801, "step": 13180 }, { "epoch": 0.26, "learning_rate": 4.573717886510232e-05, "loss": 0.0026, "step": 13182 }, { "epoch": 0.26, "learning_rate": 4.5736531904845084e-05, "loss": 0.0005, "step": 13184 }, { "epoch": 0.26, "learning_rate": 4.573588494458785e-05, "loss": 0.0009, "step": 13186 }, { "epoch": 0.26, "learning_rate": 4.573523798433063e-05, "loss": 0.0242, "step": 13188 }, { "epoch": 0.26, "learning_rate": 4.573459102407339e-05, "loss": 0.0148, "step": 13190 }, { "epoch": 0.26, "learning_rate": 4.573394406381617e-05, "loss": 0.001, "step": 13192 }, { "epoch": 0.26, "learning_rate": 4.573329710355893e-05, "loss": 0.0166, "step": 13194 }, { "epoch": 0.26, "learning_rate": 4.57326501433017e-05, "loss": 0.0011, "step": 13196 }, { "epoch": 0.26, "learning_rate": 4.573200318304447e-05, "loss": 0.0019, "step": 13198 }, { "epoch": 0.26, "learning_rate": 4.5731356222787236e-05, "loss": 0.0113, "step": 13200 }, { "epoch": 0.26, "learning_rate": 4.5730709262530005e-05, "loss": 0.0061, "step": 13202 }, { "epoch": 0.26, "learning_rate": 4.5730062302272774e-05, "loss": 0.0048, "step": 13204 }, { "epoch": 0.26, "learning_rate": 4.5729415342015543e-05, "loss": 0.0103, "step": 13206 }, { "epoch": 0.26, "learning_rate": 4.572876838175831e-05, "loss": 0.0196, "step": 13208 }, { "epoch": 0.26, "learning_rate": 4.572812142150108e-05, "loss": 0.0099, "step": 13210 }, { "epoch": 0.26, "learning_rate": 4.5727474461243844e-05, "loss": 0.0092, "step": 13212 }, { "epoch": 0.26, "learning_rate": 4.572682750098662e-05, "loss": 0.0112, "step": 13214 }, { "epoch": 0.26, "learning_rate": 4.572618054072938e-05, "loss": 0.0014, "step": 13216 }, { "epoch": 0.26, "learning_rate": 4.572553358047215e-05, "loss": 0.0003, "step": 13218 }, { "epoch": 0.26, "learning_rate": 4.572488662021493e-05, "loss": 0.0011, "step": 13220 }, { "epoch": 0.26, "learning_rate": 4.572423965995769e-05, "loss": 0.004, "step": 13222 }, { "epoch": 0.26, "learning_rate": 4.572359269970046e-05, "loss": 0.0081, "step": 13224 }, { "epoch": 0.26, "learning_rate": 4.572294573944323e-05, "loss": 0.0063, "step": 13226 }, { "epoch": 0.26, "learning_rate": 4.5722298779186e-05, "loss": 0.0244, "step": 13228 }, { "epoch": 0.26, "learning_rate": 4.5721651818928766e-05, "loss": 0.0017, "step": 13230 }, { "epoch": 0.26, "learning_rate": 4.5721004858671535e-05, "loss": 0.0008, "step": 13232 }, { "epoch": 0.26, "learning_rate": 4.5720357898414304e-05, "loss": 0.0179, "step": 13234 }, { "epoch": 0.26, "learning_rate": 4.571971093815707e-05, "loss": 0.0023, "step": 13236 }, { "epoch": 0.26, "learning_rate": 4.571906397789984e-05, "loss": 0.0078, "step": 13238 }, { "epoch": 0.26, "learning_rate": 4.5718417017642604e-05, "loss": 0.0125, "step": 13240 }, { "epoch": 0.26, "learning_rate": 4.571777005738538e-05, "loss": 0.0422, "step": 13242 }, { "epoch": 0.26, "learning_rate": 4.571712309712814e-05, "loss": 0.0043, "step": 13244 }, { "epoch": 0.26, "learning_rate": 4.571647613687091e-05, "loss": 0.0012, "step": 13246 }, { "epoch": 0.26, "learning_rate": 4.571582917661368e-05, "loss": 0.0118, "step": 13248 }, { "epoch": 0.26, "learning_rate": 4.571518221635645e-05, "loss": 0.004, "step": 13250 }, { "epoch": 0.26, "learning_rate": 4.5714535256099226e-05, "loss": 0.0003, "step": 13252 }, { "epoch": 0.26, "learning_rate": 4.571388829584199e-05, "loss": 0.0004, "step": 13254 }, { "epoch": 0.26, "learning_rate": 4.571324133558476e-05, "loss": 0.0021, "step": 13256 }, { "epoch": 0.26, "learning_rate": 4.5712594375327526e-05, "loss": 0.024, "step": 13258 }, { "epoch": 0.26, "learning_rate": 4.5711947415070295e-05, "loss": 0.0034, "step": 13260 }, { "epoch": 0.26, "learning_rate": 4.571130045481306e-05, "loss": 0.0036, "step": 13262 }, { "epoch": 0.26, "learning_rate": 4.571065349455583e-05, "loss": 0.0015, "step": 13264 }, { "epoch": 0.26, "learning_rate": 4.57100065342986e-05, "loss": 0.0006, "step": 13266 }, { "epoch": 0.26, "learning_rate": 4.570935957404137e-05, "loss": 0.0133, "step": 13268 }, { "epoch": 0.26, "learning_rate": 4.570871261378414e-05, "loss": 0.0045, "step": 13270 }, { "epoch": 0.26, "learning_rate": 4.57080656535269e-05, "loss": 0.0006, "step": 13272 }, { "epoch": 0.26, "learning_rate": 4.570741869326968e-05, "loss": 0.0017, "step": 13274 }, { "epoch": 0.26, "learning_rate": 4.570677173301244e-05, "loss": 0.0005, "step": 13276 }, { "epoch": 0.26, "learning_rate": 4.570612477275521e-05, "loss": 0.01, "step": 13278 }, { "epoch": 0.26, "learning_rate": 4.570547781249798e-05, "loss": 0.0025, "step": 13280 }, { "epoch": 0.26, "learning_rate": 4.570483085224075e-05, "loss": 0.0004, "step": 13282 }, { "epoch": 0.26, "learning_rate": 4.570418389198352e-05, "loss": 0.0098, "step": 13284 }, { "epoch": 0.26, "learning_rate": 4.5703536931726287e-05, "loss": 0.0011, "step": 13286 }, { "epoch": 0.26, "learning_rate": 4.5702889971469056e-05, "loss": 0.0012, "step": 13288 }, { "epoch": 0.26, "learning_rate": 4.5702243011211825e-05, "loss": 0.0039, "step": 13290 }, { "epoch": 0.26, "learning_rate": 4.5701596050954594e-05, "loss": 0.0017, "step": 13292 }, { "epoch": 0.26, "learning_rate": 4.5700949090697356e-05, "loss": 0.0035, "step": 13294 }, { "epoch": 0.26, "learning_rate": 4.570030213044013e-05, "loss": 0.0004, "step": 13296 }, { "epoch": 0.26, "learning_rate": 4.56996551701829e-05, "loss": 0.0014, "step": 13298 }, { "epoch": 0.26, "learning_rate": 4.569900820992566e-05, "loss": 0.0007, "step": 13300 }, { "epoch": 0.26, "learning_rate": 4.569836124966844e-05, "loss": 0.0025, "step": 13302 }, { "epoch": 0.26, "learning_rate": 4.56977142894112e-05, "loss": 0.0102, "step": 13304 }, { "epoch": 0.26, "learning_rate": 4.569706732915397e-05, "loss": 0.003, "step": 13306 }, { "epoch": 0.26, "learning_rate": 4.569642036889674e-05, "loss": 0.0051, "step": 13308 }, { "epoch": 0.26, "learning_rate": 4.569577340863951e-05, "loss": 0.0027, "step": 13310 }, { "epoch": 0.26, "learning_rate": 4.569512644838228e-05, "loss": 0.0089, "step": 13312 }, { "epoch": 0.26, "learning_rate": 4.569447948812505e-05, "loss": 0.0065, "step": 13314 }, { "epoch": 0.26, "learning_rate": 4.5693832527867816e-05, "loss": 0.0006, "step": 13316 }, { "epoch": 0.26, "learning_rate": 4.5693185567610585e-05, "loss": 0.0007, "step": 13318 }, { "epoch": 0.26, "learning_rate": 4.5692538607353354e-05, "loss": 0.0039, "step": 13320 }, { "epoch": 0.26, "learning_rate": 4.5691891647096116e-05, "loss": 0.0006, "step": 13322 }, { "epoch": 0.26, "learning_rate": 4.569124468683889e-05, "loss": 0.0297, "step": 13324 }, { "epoch": 0.26, "learning_rate": 4.5690597726581655e-05, "loss": 0.0137, "step": 13326 }, { "epoch": 0.26, "learning_rate": 4.5689950766324424e-05, "loss": 0.0018, "step": 13328 }, { "epoch": 0.26, "learning_rate": 4.56893038060672e-05, "loss": 0.0038, "step": 13330 }, { "epoch": 0.26, "learning_rate": 4.568865684580996e-05, "loss": 0.0003, "step": 13332 }, { "epoch": 0.26, "learning_rate": 4.568800988555274e-05, "loss": 0.0021, "step": 13334 }, { "epoch": 0.26, "learning_rate": 4.56873629252955e-05, "loss": 0.0071, "step": 13336 }, { "epoch": 0.26, "learning_rate": 4.568671596503827e-05, "loss": 0.0076, "step": 13338 }, { "epoch": 0.26, "learning_rate": 4.568606900478104e-05, "loss": 0.0056, "step": 13340 }, { "epoch": 0.26, "learning_rate": 4.568542204452381e-05, "loss": 0.0063, "step": 13342 }, { "epoch": 0.26, "learning_rate": 4.5684775084266576e-05, "loss": 0.0075, "step": 13344 }, { "epoch": 0.26, "learning_rate": 4.5684128124009345e-05, "loss": 0.0006, "step": 13346 }, { "epoch": 0.26, "learning_rate": 4.5683481163752115e-05, "loss": 0.0114, "step": 13348 }, { "epoch": 0.26, "learning_rate": 4.5682834203494884e-05, "loss": 0.0028, "step": 13350 }, { "epoch": 0.26, "learning_rate": 4.568218724323765e-05, "loss": 0.0031, "step": 13352 }, { "epoch": 0.26, "learning_rate": 4.5681540282980415e-05, "loss": 0.0071, "step": 13354 }, { "epoch": 0.26, "learning_rate": 4.568089332272319e-05, "loss": 0.0032, "step": 13356 }, { "epoch": 0.26, "learning_rate": 4.568024636246595e-05, "loss": 0.0023, "step": 13358 }, { "epoch": 0.26, "learning_rate": 4.567959940220872e-05, "loss": 0.0063, "step": 13360 }, { "epoch": 0.26, "learning_rate": 4.567895244195149e-05, "loss": 0.001, "step": 13362 }, { "epoch": 0.26, "learning_rate": 4.567830548169426e-05, "loss": 0.0092, "step": 13364 }, { "epoch": 0.26, "learning_rate": 4.567765852143703e-05, "loss": 0.004, "step": 13366 }, { "epoch": 0.26, "learning_rate": 4.56770115611798e-05, "loss": 0.0014, "step": 13368 }, { "epoch": 0.26, "learning_rate": 4.567636460092257e-05, "loss": 0.0054, "step": 13370 }, { "epoch": 0.26, "learning_rate": 4.567571764066534e-05, "loss": 0.0004, "step": 13372 }, { "epoch": 0.26, "learning_rate": 4.5675070680408106e-05, "loss": 0.0089, "step": 13374 }, { "epoch": 0.26, "learning_rate": 4.5674423720150875e-05, "loss": 0.0005, "step": 13376 }, { "epoch": 0.26, "learning_rate": 4.5673776759893644e-05, "loss": 0.0008, "step": 13378 }, { "epoch": 0.26, "learning_rate": 4.567312979963641e-05, "loss": 0.0008, "step": 13380 }, { "epoch": 0.26, "learning_rate": 4.5672482839379175e-05, "loss": 0.0021, "step": 13382 }, { "epoch": 0.26, "learning_rate": 4.567183587912195e-05, "loss": 0.0227, "step": 13384 }, { "epoch": 0.26, "learning_rate": 4.5671188918864714e-05, "loss": 0.0012, "step": 13386 }, { "epoch": 0.26, "learning_rate": 4.567054195860748e-05, "loss": 0.0008, "step": 13388 }, { "epoch": 0.26, "learning_rate": 4.566989499835025e-05, "loss": 0.0078, "step": 13390 }, { "epoch": 0.26, "learning_rate": 4.566924803809302e-05, "loss": 0.0006, "step": 13392 }, { "epoch": 0.26, "learning_rate": 4.566860107783579e-05, "loss": 0.015, "step": 13394 }, { "epoch": 0.26, "learning_rate": 4.566795411757856e-05, "loss": 0.0006, "step": 13396 }, { "epoch": 0.26, "learning_rate": 4.566730715732133e-05, "loss": 0.0104, "step": 13398 }, { "epoch": 0.26, "learning_rate": 4.56666601970641e-05, "loss": 0.0096, "step": 13400 }, { "epoch": 0.26, "learning_rate": 4.5666013236806866e-05, "loss": 0.0264, "step": 13402 }, { "epoch": 0.26, "learning_rate": 4.566536627654963e-05, "loss": 0.0013, "step": 13404 }, { "epoch": 0.26, "learning_rate": 4.5664719316292404e-05, "loss": 0.0168, "step": 13406 }, { "epoch": 0.26, "learning_rate": 4.5664072356035174e-05, "loss": 0.0059, "step": 13408 }, { "epoch": 0.26, "learning_rate": 4.566342539577794e-05, "loss": 0.0012, "step": 13410 }, { "epoch": 0.26, "learning_rate": 4.566277843552071e-05, "loss": 0.0031, "step": 13412 }, { "epoch": 0.26, "learning_rate": 4.5662131475263474e-05, "loss": 0.0085, "step": 13414 }, { "epoch": 0.26, "learning_rate": 4.566148451500625e-05, "loss": 0.0051, "step": 13416 }, { "epoch": 0.26, "learning_rate": 4.566083755474901e-05, "loss": 0.0004, "step": 13418 }, { "epoch": 0.26, "learning_rate": 4.566019059449178e-05, "loss": 0.0004, "step": 13420 }, { "epoch": 0.26, "learning_rate": 4.565954363423455e-05, "loss": 0.0071, "step": 13422 }, { "epoch": 0.26, "learning_rate": 4.565889667397732e-05, "loss": 0.001, "step": 13424 }, { "epoch": 0.26, "learning_rate": 4.565824971372009e-05, "loss": 0.0052, "step": 13426 }, { "epoch": 0.26, "learning_rate": 4.565760275346286e-05, "loss": 0.0243, "step": 13428 }, { "epoch": 0.26, "learning_rate": 4.565695579320563e-05, "loss": 0.0009, "step": 13430 }, { "epoch": 0.26, "learning_rate": 4.5656308832948396e-05, "loss": 0.0026, "step": 13432 }, { "epoch": 0.26, "learning_rate": 4.5655661872691165e-05, "loss": 0.0007, "step": 13434 }, { "epoch": 0.26, "learning_rate": 4.565501491243393e-05, "loss": 0.0219, "step": 13436 }, { "epoch": 0.26, "learning_rate": 4.56543679521767e-05, "loss": 0.005, "step": 13438 }, { "epoch": 0.26, "learning_rate": 4.5653720991919465e-05, "loss": 0.0004, "step": 13440 }, { "epoch": 0.26, "learning_rate": 4.5653074031662234e-05, "loss": 0.0036, "step": 13442 }, { "epoch": 0.26, "learning_rate": 4.565242707140501e-05, "loss": 0.0063, "step": 13444 }, { "epoch": 0.26, "learning_rate": 4.565178011114777e-05, "loss": 0.0575, "step": 13446 }, { "epoch": 0.26, "learning_rate": 4.565113315089054e-05, "loss": 0.0105, "step": 13448 }, { "epoch": 0.26, "learning_rate": 4.565048619063331e-05, "loss": 0.0039, "step": 13450 }, { "epoch": 0.26, "learning_rate": 4.564983923037608e-05, "loss": 0.0005, "step": 13452 }, { "epoch": 0.26, "learning_rate": 4.564919227011885e-05, "loss": 0.0014, "step": 13454 }, { "epoch": 0.26, "learning_rate": 4.564854530986162e-05, "loss": 0.0038, "step": 13456 }, { "epoch": 0.26, "learning_rate": 4.564789834960439e-05, "loss": 0.0038, "step": 13458 }, { "epoch": 0.26, "learning_rate": 4.5647251389347156e-05, "loss": 0.006, "step": 13460 }, { "epoch": 0.26, "learning_rate": 4.5646604429089925e-05, "loss": 0.0012, "step": 13462 }, { "epoch": 0.26, "learning_rate": 4.564595746883269e-05, "loss": 0.0024, "step": 13464 }, { "epoch": 0.26, "learning_rate": 4.5645310508575463e-05, "loss": 0.0006, "step": 13466 }, { "epoch": 0.26, "learning_rate": 4.5644663548318226e-05, "loss": 0.032, "step": 13468 }, { "epoch": 0.26, "learning_rate": 4.5644016588060995e-05, "loss": 0.0081, "step": 13470 }, { "epoch": 0.26, "learning_rate": 4.5643369627803764e-05, "loss": 0.001, "step": 13472 }, { "epoch": 0.26, "learning_rate": 4.564272266754653e-05, "loss": 0.0072, "step": 13474 }, { "epoch": 0.26, "learning_rate": 4.564207570728931e-05, "loss": 0.0188, "step": 13476 }, { "epoch": 0.26, "learning_rate": 4.564142874703207e-05, "loss": 0.0013, "step": 13478 }, { "epoch": 0.26, "learning_rate": 4.564078178677484e-05, "loss": 0.0067, "step": 13480 }, { "epoch": 0.26, "learning_rate": 4.564013482651761e-05, "loss": 0.0015, "step": 13482 }, { "epoch": 0.26, "learning_rate": 4.563948786626038e-05, "loss": 0.0004, "step": 13484 }, { "epoch": 0.26, "learning_rate": 4.563884090600314e-05, "loss": 0.0029, "step": 13486 }, { "epoch": 0.26, "learning_rate": 4.5638193945745917e-05, "loss": 0.001, "step": 13488 }, { "epoch": 0.26, "learning_rate": 4.5637546985488686e-05, "loss": 0.0034, "step": 13490 }, { "epoch": 0.26, "learning_rate": 4.5636900025231455e-05, "loss": 0.0279, "step": 13492 }, { "epoch": 0.26, "learning_rate": 4.5636253064974224e-05, "loss": 0.0017, "step": 13494 }, { "epoch": 0.26, "learning_rate": 4.5635606104716986e-05, "loss": 0.0017, "step": 13496 }, { "epoch": 0.26, "learning_rate": 4.563495914445976e-05, "loss": 0.0072, "step": 13498 }, { "epoch": 0.26, "learning_rate": 4.5634312184202524e-05, "loss": 0.0222, "step": 13500 }, { "epoch": 0.26, "learning_rate": 4.5633665223945293e-05, "loss": 0.05, "step": 13502 }, { "epoch": 0.26, "learning_rate": 4.563301826368806e-05, "loss": 0.0079, "step": 13504 }, { "epoch": 0.26, "learning_rate": 4.563237130343083e-05, "loss": 0.0005, "step": 13506 }, { "epoch": 0.26, "learning_rate": 4.56317243431736e-05, "loss": 0.0013, "step": 13508 }, { "epoch": 0.26, "learning_rate": 4.563107738291637e-05, "loss": 0.0038, "step": 13510 }, { "epoch": 0.26, "learning_rate": 4.563043042265914e-05, "loss": 0.0081, "step": 13512 }, { "epoch": 0.26, "learning_rate": 4.562978346240191e-05, "loss": 0.0029, "step": 13514 }, { "epoch": 0.26, "learning_rate": 4.562913650214468e-05, "loss": 0.0003, "step": 13516 }, { "epoch": 0.26, "learning_rate": 4.562848954188744e-05, "loss": 0.0079, "step": 13518 }, { "epoch": 0.26, "learning_rate": 4.5627842581630215e-05, "loss": 0.0196, "step": 13520 }, { "epoch": 0.26, "learning_rate": 4.5627195621372984e-05, "loss": 0.0123, "step": 13522 }, { "epoch": 0.26, "learning_rate": 4.5626548661115747e-05, "loss": 0.0208, "step": 13524 }, { "epoch": 0.26, "learning_rate": 4.562590170085852e-05, "loss": 0.0059, "step": 13526 }, { "epoch": 0.26, "learning_rate": 4.5625254740601285e-05, "loss": 0.0011, "step": 13528 }, { "epoch": 0.26, "learning_rate": 4.5624607780344054e-05, "loss": 0.0015, "step": 13530 }, { "epoch": 0.26, "learning_rate": 4.562396082008682e-05, "loss": 0.0057, "step": 13532 }, { "epoch": 0.26, "learning_rate": 4.562331385982959e-05, "loss": 0.0094, "step": 13534 }, { "epoch": 0.26, "learning_rate": 4.562266689957236e-05, "loss": 0.0089, "step": 13536 }, { "epoch": 0.26, "learning_rate": 4.562201993931513e-05, "loss": 0.0014, "step": 13538 }, { "epoch": 0.26, "learning_rate": 4.56213729790579e-05, "loss": 0.0172, "step": 13540 }, { "epoch": 0.26, "learning_rate": 4.562072601880067e-05, "loss": 0.0097, "step": 13542 }, { "epoch": 0.26, "learning_rate": 4.562007905854344e-05, "loss": 0.0117, "step": 13544 }, { "epoch": 0.26, "learning_rate": 4.56194320982862e-05, "loss": 0.0063, "step": 13546 }, { "epoch": 0.26, "learning_rate": 4.5618785138028976e-05, "loss": 0.0023, "step": 13548 }, { "epoch": 0.26, "learning_rate": 4.561813817777174e-05, "loss": 0.0071, "step": 13550 }, { "epoch": 0.26, "learning_rate": 4.561749121751451e-05, "loss": 0.0017, "step": 13552 }, { "epoch": 0.26, "learning_rate": 4.561684425725728e-05, "loss": 0.0104, "step": 13554 }, { "epoch": 0.26, "learning_rate": 4.5616197297000045e-05, "loss": 0.0159, "step": 13556 }, { "epoch": 0.26, "learning_rate": 4.561555033674282e-05, "loss": 0.0021, "step": 13558 }, { "epoch": 0.26, "learning_rate": 4.561490337648558e-05, "loss": 0.0084, "step": 13560 }, { "epoch": 0.26, "learning_rate": 4.561425641622835e-05, "loss": 0.0043, "step": 13562 }, { "epoch": 0.26, "learning_rate": 4.561360945597112e-05, "loss": 0.0071, "step": 13564 }, { "epoch": 0.26, "learning_rate": 4.561296249571389e-05, "loss": 0.0075, "step": 13566 }, { "epoch": 0.26, "learning_rate": 4.561231553545666e-05, "loss": 0.0005, "step": 13568 }, { "epoch": 0.26, "learning_rate": 4.561166857519943e-05, "loss": 0.01, "step": 13570 }, { "epoch": 0.26, "learning_rate": 4.56110216149422e-05, "loss": 0.04, "step": 13572 }, { "epoch": 0.26, "learning_rate": 4.561037465468497e-05, "loss": 0.0313, "step": 13574 }, { "epoch": 0.26, "learning_rate": 4.5609727694427736e-05, "loss": 0.0291, "step": 13576 }, { "epoch": 0.26, "learning_rate": 4.56090807341705e-05, "loss": 0.0033, "step": 13578 }, { "epoch": 0.26, "learning_rate": 4.5608433773913274e-05, "loss": 0.0285, "step": 13580 }, { "epoch": 0.26, "learning_rate": 4.5607786813656036e-05, "loss": 0.0093, "step": 13582 }, { "epoch": 0.26, "learning_rate": 4.5607139853398806e-05, "loss": 0.0358, "step": 13584 }, { "epoch": 0.26, "learning_rate": 4.560649289314158e-05, "loss": 0.0081, "step": 13586 }, { "epoch": 0.26, "learning_rate": 4.5605845932884344e-05, "loss": 0.0076, "step": 13588 }, { "epoch": 0.26, "learning_rate": 4.560519897262711e-05, "loss": 0.0041, "step": 13590 }, { "epoch": 0.26, "learning_rate": 4.560455201236988e-05, "loss": 0.0144, "step": 13592 }, { "epoch": 0.26, "learning_rate": 4.560390505211265e-05, "loss": 0.0066, "step": 13594 }, { "epoch": 0.26, "learning_rate": 4.560325809185542e-05, "loss": 0.0205, "step": 13596 }, { "epoch": 0.26, "learning_rate": 4.560261113159819e-05, "loss": 0.0113, "step": 13598 }, { "epoch": 0.26, "learning_rate": 4.560196417134096e-05, "loss": 0.0207, "step": 13600 }, { "epoch": 0.26, "learning_rate": 4.560131721108373e-05, "loss": 0.0005, "step": 13602 }, { "epoch": 0.26, "learning_rate": 4.5600670250826496e-05, "loss": 0.0058, "step": 13604 }, { "epoch": 0.26, "learning_rate": 4.560002329056926e-05, "loss": 0.0003, "step": 13606 }, { "epoch": 0.26, "learning_rate": 4.5599376330312035e-05, "loss": 0.0062, "step": 13608 }, { "epoch": 0.26, "learning_rate": 4.55987293700548e-05, "loss": 0.0166, "step": 13610 }, { "epoch": 0.26, "learning_rate": 4.5598082409797566e-05, "loss": 0.0095, "step": 13612 }, { "epoch": 0.26, "learning_rate": 4.5597435449540335e-05, "loss": 0.0031, "step": 13614 }, { "epoch": 0.26, "learning_rate": 4.5596788489283104e-05, "loss": 0.0011, "step": 13616 }, { "epoch": 0.26, "learning_rate": 4.559614152902587e-05, "loss": 0.0012, "step": 13618 }, { "epoch": 0.26, "learning_rate": 4.559549456876864e-05, "loss": 0.0006, "step": 13620 }, { "epoch": 0.26, "learning_rate": 4.559484760851141e-05, "loss": 0.0081, "step": 13622 }, { "epoch": 0.26, "learning_rate": 4.559420064825418e-05, "loss": 0.0028, "step": 13624 }, { "epoch": 0.26, "learning_rate": 4.559355368799695e-05, "loss": 0.0079, "step": 13626 }, { "epoch": 0.26, "learning_rate": 4.559290672773971e-05, "loss": 0.0106, "step": 13628 }, { "epoch": 0.26, "learning_rate": 4.559225976748249e-05, "loss": 0.0124, "step": 13630 }, { "epoch": 0.26, "learning_rate": 4.559161280722526e-05, "loss": 0.0113, "step": 13632 }, { "epoch": 0.26, "learning_rate": 4.5590965846968026e-05, "loss": 0.0537, "step": 13634 }, { "epoch": 0.26, "learning_rate": 4.5590318886710795e-05, "loss": 0.0147, "step": 13636 }, { "epoch": 0.26, "learning_rate": 4.558967192645356e-05, "loss": 0.0011, "step": 13638 }, { "epoch": 0.26, "learning_rate": 4.558902496619633e-05, "loss": 0.0022, "step": 13640 }, { "epoch": 0.26, "learning_rate": 4.5588378005939095e-05, "loss": 0.0019, "step": 13642 }, { "epoch": 0.26, "learning_rate": 4.5587731045681865e-05, "loss": 0.0001, "step": 13644 }, { "epoch": 0.26, "learning_rate": 4.5587084085424634e-05, "loss": 0.0203, "step": 13646 }, { "epoch": 0.26, "learning_rate": 4.55864371251674e-05, "loss": 0.0005, "step": 13648 }, { "epoch": 0.26, "learning_rate": 4.558579016491017e-05, "loss": 0.0055, "step": 13650 }, { "epoch": 0.26, "learning_rate": 4.558514320465294e-05, "loss": 0.0084, "step": 13652 }, { "epoch": 0.27, "learning_rate": 4.558449624439571e-05, "loss": 0.0005, "step": 13654 }, { "epoch": 0.27, "learning_rate": 4.558384928413848e-05, "loss": 0.0295, "step": 13656 }, { "epoch": 0.27, "learning_rate": 4.558320232388125e-05, "loss": 0.0067, "step": 13658 }, { "epoch": 0.27, "learning_rate": 4.558255536362401e-05, "loss": 0.0007, "step": 13660 }, { "epoch": 0.27, "learning_rate": 4.5581908403366786e-05, "loss": 0.0006, "step": 13662 }, { "epoch": 0.27, "learning_rate": 4.558126144310955e-05, "loss": 0.0114, "step": 13664 }, { "epoch": 0.27, "learning_rate": 4.558061448285232e-05, "loss": 0.0023, "step": 13666 }, { "epoch": 0.27, "learning_rate": 4.5579967522595094e-05, "loss": 0.0112, "step": 13668 }, { "epoch": 0.27, "learning_rate": 4.5579320562337856e-05, "loss": 0.0243, "step": 13670 }, { "epoch": 0.27, "learning_rate": 4.5578673602080625e-05, "loss": 0.0131, "step": 13672 }, { "epoch": 0.27, "learning_rate": 4.5578026641823394e-05, "loss": 0.008, "step": 13674 }, { "epoch": 0.27, "learning_rate": 4.557737968156616e-05, "loss": 0.0056, "step": 13676 }, { "epoch": 0.27, "learning_rate": 4.557673272130893e-05, "loss": 0.0041, "step": 13678 }, { "epoch": 0.27, "learning_rate": 4.55760857610517e-05, "loss": 0.005, "step": 13680 }, { "epoch": 0.27, "learning_rate": 4.557543880079447e-05, "loss": 0.0021, "step": 13682 }, { "epoch": 0.27, "learning_rate": 4.557479184053724e-05, "loss": 0.0019, "step": 13684 }, { "epoch": 0.27, "learning_rate": 4.557414488028001e-05, "loss": 0.0199, "step": 13686 }, { "epoch": 0.27, "learning_rate": 4.557349792002277e-05, "loss": 0.0006, "step": 13688 }, { "epoch": 0.27, "learning_rate": 4.557285095976555e-05, "loss": 0.0058, "step": 13690 }, { "epoch": 0.27, "learning_rate": 4.557220399950831e-05, "loss": 0.0042, "step": 13692 }, { "epoch": 0.27, "learning_rate": 4.557155703925108e-05, "loss": 0.0267, "step": 13694 }, { "epoch": 0.27, "learning_rate": 4.557091007899385e-05, "loss": 0.009, "step": 13696 }, { "epoch": 0.27, "learning_rate": 4.5570263118736616e-05, "loss": 0.0048, "step": 13698 }, { "epoch": 0.27, "learning_rate": 4.556961615847939e-05, "loss": 0.0024, "step": 13700 }, { "epoch": 0.27, "learning_rate": 4.5568969198222154e-05, "loss": 0.0022, "step": 13702 }, { "epoch": 0.27, "learning_rate": 4.5568322237964924e-05, "loss": 0.0026, "step": 13704 }, { "epoch": 0.27, "learning_rate": 4.556767527770769e-05, "loss": 0.0115, "step": 13706 }, { "epoch": 0.27, "learning_rate": 4.556702831745046e-05, "loss": 0.0076, "step": 13708 }, { "epoch": 0.27, "learning_rate": 4.556638135719323e-05, "loss": 0.0042, "step": 13710 }, { "epoch": 0.27, "learning_rate": 4.5565734396936e-05, "loss": 0.0191, "step": 13712 }, { "epoch": 0.27, "learning_rate": 4.556508743667877e-05, "loss": 0.0015, "step": 13714 }, { "epoch": 0.27, "learning_rate": 4.556444047642154e-05, "loss": 0.0009, "step": 13716 }, { "epoch": 0.27, "learning_rate": 4.556379351616431e-05, "loss": 0.0016, "step": 13718 }, { "epoch": 0.27, "learning_rate": 4.556314655590707e-05, "loss": 0.0002, "step": 13720 }, { "epoch": 0.27, "learning_rate": 4.5562499595649845e-05, "loss": 0.0131, "step": 13722 }, { "epoch": 0.27, "learning_rate": 4.556185263539261e-05, "loss": 0.0097, "step": 13724 }, { "epoch": 0.27, "learning_rate": 4.556120567513538e-05, "loss": 0.0008, "step": 13726 }, { "epoch": 0.27, "learning_rate": 4.5560558714878146e-05, "loss": 0.0169, "step": 13728 }, { "epoch": 0.27, "learning_rate": 4.5559911754620915e-05, "loss": 0.1463, "step": 13730 }, { "epoch": 0.27, "learning_rate": 4.5559264794363684e-05, "loss": 0.0012, "step": 13732 }, { "epoch": 0.27, "learning_rate": 4.555861783410645e-05, "loss": 0.0016, "step": 13734 }, { "epoch": 0.27, "learning_rate": 4.555797087384922e-05, "loss": 0.0003, "step": 13736 }, { "epoch": 0.27, "learning_rate": 4.555732391359199e-05, "loss": 0.0131, "step": 13738 }, { "epoch": 0.27, "learning_rate": 4.555667695333476e-05, "loss": 0.0054, "step": 13740 }, { "epoch": 0.27, "learning_rate": 4.555602999307752e-05, "loss": 0.014, "step": 13742 }, { "epoch": 0.27, "learning_rate": 4.55553830328203e-05, "loss": 0.0003, "step": 13744 }, { "epoch": 0.27, "learning_rate": 4.555473607256307e-05, "loss": 0.0007, "step": 13746 }, { "epoch": 0.27, "learning_rate": 4.555408911230583e-05, "loss": 0.0009, "step": 13748 }, { "epoch": 0.27, "learning_rate": 4.5553442152048606e-05, "loss": 0.0005, "step": 13750 }, { "epoch": 0.27, "learning_rate": 4.555279519179137e-05, "loss": 0.0138, "step": 13752 }, { "epoch": 0.27, "learning_rate": 4.555214823153414e-05, "loss": 0.0005, "step": 13754 }, { "epoch": 0.27, "learning_rate": 4.5551501271276906e-05, "loss": 0.0011, "step": 13756 }, { "epoch": 0.27, "learning_rate": 4.5550854311019675e-05, "loss": 0.0003, "step": 13758 }, { "epoch": 0.27, "learning_rate": 4.5550207350762444e-05, "loss": 0.0199, "step": 13760 }, { "epoch": 0.27, "learning_rate": 4.5549560390505213e-05, "loss": 0.0021, "step": 13762 }, { "epoch": 0.27, "learning_rate": 4.554891343024798e-05, "loss": 0.0005, "step": 13764 }, { "epoch": 0.27, "learning_rate": 4.554826646999075e-05, "loss": 0.0061, "step": 13766 }, { "epoch": 0.27, "learning_rate": 4.554761950973352e-05, "loss": 0.0049, "step": 13768 }, { "epoch": 0.27, "learning_rate": 4.554697254947628e-05, "loss": 0.0224, "step": 13770 }, { "epoch": 0.27, "learning_rate": 4.554632558921906e-05, "loss": 0.0163, "step": 13772 }, { "epoch": 0.27, "learning_rate": 4.554567862896182e-05, "loss": 0.0005, "step": 13774 }, { "epoch": 0.27, "learning_rate": 4.55450316687046e-05, "loss": 0.0008, "step": 13776 }, { "epoch": 0.27, "learning_rate": 4.5544384708447366e-05, "loss": 0.0103, "step": 13778 }, { "epoch": 0.27, "learning_rate": 4.554373774819013e-05, "loss": 0.0049, "step": 13780 }, { "epoch": 0.27, "learning_rate": 4.5543090787932904e-05, "loss": 0.0006, "step": 13782 }, { "epoch": 0.27, "learning_rate": 4.5542443827675667e-05, "loss": 0.0017, "step": 13784 }, { "epoch": 0.27, "learning_rate": 4.5541796867418436e-05, "loss": 0.0051, "step": 13786 }, { "epoch": 0.27, "learning_rate": 4.5541149907161205e-05, "loss": 0.001, "step": 13788 }, { "epoch": 0.27, "learning_rate": 4.5540502946903974e-05, "loss": 0.0005, "step": 13790 }, { "epoch": 0.27, "learning_rate": 4.553985598664674e-05, "loss": 0.0155, "step": 13792 }, { "epoch": 0.27, "learning_rate": 4.553920902638951e-05, "loss": 0.0007, "step": 13794 }, { "epoch": 0.27, "learning_rate": 4.553856206613228e-05, "loss": 0.0067, "step": 13796 }, { "epoch": 0.27, "learning_rate": 4.553791510587505e-05, "loss": 0.0264, "step": 13798 }, { "epoch": 0.27, "learning_rate": 4.553726814561782e-05, "loss": 0.0003, "step": 13800 }, { "epoch": 0.27, "learning_rate": 4.553662118536058e-05, "loss": 0.0129, "step": 13802 }, { "epoch": 0.27, "learning_rate": 4.553597422510336e-05, "loss": 0.0005, "step": 13804 }, { "epoch": 0.27, "learning_rate": 4.553532726484612e-05, "loss": 0.0318, "step": 13806 }, { "epoch": 0.27, "learning_rate": 4.553468030458889e-05, "loss": 0.0087, "step": 13808 }, { "epoch": 0.27, "learning_rate": 4.5534033344331665e-05, "loss": 0.0044, "step": 13810 }, { "epoch": 0.27, "learning_rate": 4.553338638407443e-05, "loss": 0.0005, "step": 13812 }, { "epoch": 0.27, "learning_rate": 4.5532739423817196e-05, "loss": 0.0006, "step": 13814 }, { "epoch": 0.27, "learning_rate": 4.5532092463559965e-05, "loss": 0.0006, "step": 13816 }, { "epoch": 0.27, "learning_rate": 4.5531445503302734e-05, "loss": 0.0014, "step": 13818 }, { "epoch": 0.27, "learning_rate": 4.55307985430455e-05, "loss": 0.0085, "step": 13820 }, { "epoch": 0.27, "learning_rate": 4.553015158278827e-05, "loss": 0.031, "step": 13822 }, { "epoch": 0.27, "learning_rate": 4.552950462253104e-05, "loss": 0.0029, "step": 13824 }, { "epoch": 0.27, "learning_rate": 4.552885766227381e-05, "loss": 0.001, "step": 13826 }, { "epoch": 0.27, "learning_rate": 4.552821070201658e-05, "loss": 0.0036, "step": 13828 }, { "epoch": 0.27, "learning_rate": 4.552756374175934e-05, "loss": 0.0185, "step": 13830 }, { "epoch": 0.27, "learning_rate": 4.552691678150212e-05, "loss": 0.0002, "step": 13832 }, { "epoch": 0.27, "learning_rate": 4.552626982124488e-05, "loss": 0.0059, "step": 13834 }, { "epoch": 0.27, "learning_rate": 4.552562286098765e-05, "loss": 0.0059, "step": 13836 }, { "epoch": 0.27, "learning_rate": 4.552497590073042e-05, "loss": 0.0137, "step": 13838 }, { "epoch": 0.27, "learning_rate": 4.552432894047319e-05, "loss": 0.0074, "step": 13840 }, { "epoch": 0.27, "learning_rate": 4.5523681980215956e-05, "loss": 0.0293, "step": 13842 }, { "epoch": 0.27, "learning_rate": 4.5523035019958726e-05, "loss": 0.002, "step": 13844 }, { "epoch": 0.27, "learning_rate": 4.5522388059701495e-05, "loss": 0.0063, "step": 13846 }, { "epoch": 0.27, "learning_rate": 4.5521741099444264e-05, "loss": 0.002, "step": 13848 }, { "epoch": 0.27, "learning_rate": 4.552109413918703e-05, "loss": 0.0095, "step": 13850 }, { "epoch": 0.27, "learning_rate": 4.5520447178929795e-05, "loss": 0.0002, "step": 13852 }, { "epoch": 0.27, "learning_rate": 4.551980021867257e-05, "loss": 0.0038, "step": 13854 }, { "epoch": 0.27, "learning_rate": 4.551915325841534e-05, "loss": 0.0007, "step": 13856 }, { "epoch": 0.27, "learning_rate": 4.551850629815811e-05, "loss": 0.0225, "step": 13858 }, { "epoch": 0.27, "learning_rate": 4.551785933790088e-05, "loss": 0.0008, "step": 13860 }, { "epoch": 0.27, "learning_rate": 4.551721237764364e-05, "loss": 0.0142, "step": 13862 }, { "epoch": 0.27, "learning_rate": 4.5516565417386416e-05, "loss": 0.0007, "step": 13864 }, { "epoch": 0.27, "learning_rate": 4.551591845712918e-05, "loss": 0.0094, "step": 13866 }, { "epoch": 0.27, "learning_rate": 4.551527149687195e-05, "loss": 0.0052, "step": 13868 }, { "epoch": 0.27, "learning_rate": 4.551462453661472e-05, "loss": 0.0048, "step": 13870 }, { "epoch": 0.27, "learning_rate": 4.5513977576357486e-05, "loss": 0.0004, "step": 13872 }, { "epoch": 0.27, "learning_rate": 4.5513330616100255e-05, "loss": 0.0052, "step": 13874 }, { "epoch": 0.27, "learning_rate": 4.5512683655843024e-05, "loss": 0.0016, "step": 13876 }, { "epoch": 0.27, "learning_rate": 4.551203669558579e-05, "loss": 0.0101, "step": 13878 }, { "epoch": 0.27, "learning_rate": 4.551138973532856e-05, "loss": 0.01, "step": 13880 }, { "epoch": 0.27, "learning_rate": 4.551074277507133e-05, "loss": 0.0246, "step": 13882 }, { "epoch": 0.27, "learning_rate": 4.5510095814814094e-05, "loss": 0.0065, "step": 13884 }, { "epoch": 0.27, "learning_rate": 4.550944885455687e-05, "loss": 0.0105, "step": 13886 }, { "epoch": 0.27, "learning_rate": 4.550880189429964e-05, "loss": 0.0152, "step": 13888 }, { "epoch": 0.27, "learning_rate": 4.55081549340424e-05, "loss": 0.0004, "step": 13890 }, { "epoch": 0.27, "learning_rate": 4.550750797378518e-05, "loss": 0.0106, "step": 13892 }, { "epoch": 0.27, "learning_rate": 4.550686101352794e-05, "loss": 0.0018, "step": 13894 }, { "epoch": 0.27, "learning_rate": 4.550621405327071e-05, "loss": 0.0121, "step": 13896 }, { "epoch": 0.27, "learning_rate": 4.550556709301348e-05, "loss": 0.0016, "step": 13898 }, { "epoch": 0.27, "learning_rate": 4.5504920132756246e-05, "loss": 0.0005, "step": 13900 }, { "epoch": 0.27, "learning_rate": 4.5504273172499015e-05, "loss": 0.009, "step": 13902 }, { "epoch": 0.27, "learning_rate": 4.5503626212241785e-05, "loss": 0.0133, "step": 13904 }, { "epoch": 0.27, "learning_rate": 4.5502979251984554e-05, "loss": 0.0068, "step": 13906 }, { "epoch": 0.27, "learning_rate": 4.550233229172732e-05, "loss": 0.0135, "step": 13908 }, { "epoch": 0.27, "learning_rate": 4.550168533147009e-05, "loss": 0.0104, "step": 13910 }, { "epoch": 0.27, "learning_rate": 4.5501038371212854e-05, "loss": 0.0017, "step": 13912 }, { "epoch": 0.27, "learning_rate": 4.550039141095563e-05, "loss": 0.004, "step": 13914 }, { "epoch": 0.27, "learning_rate": 4.549974445069839e-05, "loss": 0.0037, "step": 13916 }, { "epoch": 0.27, "learning_rate": 4.549909749044117e-05, "loss": 0.0005, "step": 13918 }, { "epoch": 0.27, "learning_rate": 4.549845053018393e-05, "loss": 0.0012, "step": 13920 }, { "epoch": 0.27, "learning_rate": 4.54978035699267e-05, "loss": 0.0075, "step": 13922 }, { "epoch": 0.27, "learning_rate": 4.5497156609669475e-05, "loss": 0.0295, "step": 13924 }, { "epoch": 0.27, "learning_rate": 4.549650964941224e-05, "loss": 0.0008, "step": 13926 }, { "epoch": 0.27, "learning_rate": 4.549586268915501e-05, "loss": 0.0077, "step": 13928 }, { "epoch": 0.27, "learning_rate": 4.5495215728897776e-05, "loss": 0.0016, "step": 13930 }, { "epoch": 0.27, "learning_rate": 4.5494568768640545e-05, "loss": 0.0069, "step": 13932 }, { "epoch": 0.27, "learning_rate": 4.5493921808383314e-05, "loss": 0.0006, "step": 13934 }, { "epoch": 0.27, "learning_rate": 4.549327484812608e-05, "loss": 0.0017, "step": 13936 }, { "epoch": 0.27, "learning_rate": 4.549262788786885e-05, "loss": 0.0098, "step": 13938 }, { "epoch": 0.27, "learning_rate": 4.549198092761162e-05, "loss": 0.004, "step": 13940 }, { "epoch": 0.27, "learning_rate": 4.549133396735439e-05, "loss": 0.0002, "step": 13942 }, { "epoch": 0.27, "learning_rate": 4.549068700709715e-05, "loss": 0.0004, "step": 13944 }, { "epoch": 0.27, "learning_rate": 4.549004004683993e-05, "loss": 0.0055, "step": 13946 }, { "epoch": 0.27, "learning_rate": 4.548939308658269e-05, "loss": 0.0253, "step": 13948 }, { "epoch": 0.27, "learning_rate": 4.548874612632546e-05, "loss": 0.017, "step": 13950 }, { "epoch": 0.27, "learning_rate": 4.548809916606823e-05, "loss": 0.0006, "step": 13952 }, { "epoch": 0.27, "learning_rate": 4.5487452205811e-05, "loss": 0.0032, "step": 13954 }, { "epoch": 0.27, "learning_rate": 4.548680524555377e-05, "loss": 0.0345, "step": 13956 }, { "epoch": 0.27, "learning_rate": 4.5486158285296536e-05, "loss": 0.0029, "step": 13958 }, { "epoch": 0.27, "learning_rate": 4.5485511325039305e-05, "loss": 0.0014, "step": 13960 }, { "epoch": 0.27, "learning_rate": 4.5484864364782074e-05, "loss": 0.0019, "step": 13962 }, { "epoch": 0.27, "learning_rate": 4.5484217404524843e-05, "loss": 0.0033, "step": 13964 }, { "epoch": 0.27, "learning_rate": 4.5483570444267606e-05, "loss": 0.0201, "step": 13966 }, { "epoch": 0.27, "learning_rate": 4.548292348401038e-05, "loss": 0.0039, "step": 13968 }, { "epoch": 0.27, "learning_rate": 4.548227652375315e-05, "loss": 0.0189, "step": 13970 }, { "epoch": 0.27, "learning_rate": 4.548162956349591e-05, "loss": 0.0426, "step": 13972 }, { "epoch": 0.27, "learning_rate": 4.548098260323869e-05, "loss": 0.0017, "step": 13974 }, { "epoch": 0.27, "learning_rate": 4.548033564298145e-05, "loss": 0.0048, "step": 13976 }, { "epoch": 0.27, "learning_rate": 4.547968868272422e-05, "loss": 0.0057, "step": 13978 }, { "epoch": 0.27, "learning_rate": 4.547904172246699e-05, "loss": 0.0027, "step": 13980 }, { "epoch": 0.27, "learning_rate": 4.547839476220976e-05, "loss": 0.0125, "step": 13982 }, { "epoch": 0.27, "learning_rate": 4.547774780195253e-05, "loss": 0.0177, "step": 13984 }, { "epoch": 0.27, "learning_rate": 4.54771008416953e-05, "loss": 0.0015, "step": 13986 }, { "epoch": 0.27, "learning_rate": 4.5476453881438066e-05, "loss": 0.0059, "step": 13988 }, { "epoch": 0.27, "learning_rate": 4.5475806921180835e-05, "loss": 0.0196, "step": 13990 }, { "epoch": 0.27, "learning_rate": 4.5475159960923604e-05, "loss": 0.0103, "step": 13992 }, { "epoch": 0.27, "learning_rate": 4.5474513000666366e-05, "loss": 0.0033, "step": 13994 }, { "epoch": 0.27, "learning_rate": 4.547386604040914e-05, "loss": 0.0019, "step": 13996 }, { "epoch": 0.27, "learning_rate": 4.5473219080151904e-05, "loss": 0.0095, "step": 13998 }, { "epoch": 0.27, "learning_rate": 4.547257211989468e-05, "loss": 0.011, "step": 14000 }, { "epoch": 0.27, "learning_rate": 4.547192515963745e-05, "loss": 0.0018, "step": 14002 }, { "epoch": 0.27, "learning_rate": 4.547127819938021e-05, "loss": 0.0004, "step": 14004 }, { "epoch": 0.27, "learning_rate": 4.547063123912299e-05, "loss": 0.001, "step": 14006 }, { "epoch": 0.27, "learning_rate": 4.546998427886575e-05, "loss": 0.0003, "step": 14008 }, { "epoch": 0.27, "learning_rate": 4.546933731860852e-05, "loss": 0.0011, "step": 14010 }, { "epoch": 0.27, "learning_rate": 4.546869035835129e-05, "loss": 0.0028, "step": 14012 }, { "epoch": 0.27, "learning_rate": 4.546804339809406e-05, "loss": 0.0079, "step": 14014 }, { "epoch": 0.27, "learning_rate": 4.5467396437836826e-05, "loss": 0.0034, "step": 14016 }, { "epoch": 0.27, "learning_rate": 4.5466749477579595e-05, "loss": 0.0228, "step": 14018 }, { "epoch": 0.27, "learning_rate": 4.5466102517322364e-05, "loss": 0.0012, "step": 14020 }, { "epoch": 0.27, "learning_rate": 4.546545555706513e-05, "loss": 0.0262, "step": 14022 }, { "epoch": 0.27, "learning_rate": 4.54648085968079e-05, "loss": 0.0029, "step": 14024 }, { "epoch": 0.27, "learning_rate": 4.5464161636550665e-05, "loss": 0.0097, "step": 14026 }, { "epoch": 0.27, "learning_rate": 4.546351467629344e-05, "loss": 0.0004, "step": 14028 }, { "epoch": 0.27, "learning_rate": 4.54628677160362e-05, "loss": 0.0089, "step": 14030 }, { "epoch": 0.27, "learning_rate": 4.546222075577897e-05, "loss": 0.0004, "step": 14032 }, { "epoch": 0.27, "learning_rate": 4.546157379552175e-05, "loss": 0.0168, "step": 14034 }, { "epoch": 0.27, "learning_rate": 4.546092683526451e-05, "loss": 0.0299, "step": 14036 }, { "epoch": 0.27, "learning_rate": 4.546027987500728e-05, "loss": 0.0142, "step": 14038 }, { "epoch": 0.27, "learning_rate": 4.545963291475005e-05, "loss": 0.001, "step": 14040 }, { "epoch": 0.27, "learning_rate": 4.545898595449282e-05, "loss": 0.0009, "step": 14042 }, { "epoch": 0.27, "learning_rate": 4.5458338994235587e-05, "loss": 0.0072, "step": 14044 }, { "epoch": 0.27, "learning_rate": 4.5457692033978356e-05, "loss": 0.0013, "step": 14046 }, { "epoch": 0.27, "learning_rate": 4.5457045073721125e-05, "loss": 0.0059, "step": 14048 }, { "epoch": 0.27, "learning_rate": 4.5456398113463894e-05, "loss": 0.003, "step": 14050 }, { "epoch": 0.27, "learning_rate": 4.545575115320666e-05, "loss": 0.0086, "step": 14052 }, { "epoch": 0.27, "learning_rate": 4.5455104192949425e-05, "loss": 0.0034, "step": 14054 }, { "epoch": 0.27, "learning_rate": 4.54544572326922e-05, "loss": 0.0108, "step": 14056 }, { "epoch": 0.27, "learning_rate": 4.545381027243496e-05, "loss": 0.0132, "step": 14058 }, { "epoch": 0.27, "learning_rate": 4.545316331217774e-05, "loss": 0.0056, "step": 14060 }, { "epoch": 0.27, "learning_rate": 4.54525163519205e-05, "loss": 0.0092, "step": 14062 }, { "epoch": 0.27, "learning_rate": 4.545186939166327e-05, "loss": 0.0087, "step": 14064 }, { "epoch": 0.27, "learning_rate": 4.5451222431406046e-05, "loss": 0.0067, "step": 14066 }, { "epoch": 0.27, "learning_rate": 4.545057547114881e-05, "loss": 0.0023, "step": 14068 }, { "epoch": 0.27, "learning_rate": 4.544992851089158e-05, "loss": 0.003, "step": 14070 }, { "epoch": 0.27, "learning_rate": 4.544928155063435e-05, "loss": 0.013, "step": 14072 }, { "epoch": 0.27, "learning_rate": 4.5448634590377116e-05, "loss": 0.0002, "step": 14074 }, { "epoch": 0.27, "learning_rate": 4.544798763011988e-05, "loss": 0.01, "step": 14076 }, { "epoch": 0.27, "learning_rate": 4.5447340669862654e-05, "loss": 0.0006, "step": 14078 }, { "epoch": 0.27, "learning_rate": 4.544669370960542e-05, "loss": 0.0084, "step": 14080 }, { "epoch": 0.27, "learning_rate": 4.544604674934819e-05, "loss": 0.0081, "step": 14082 }, { "epoch": 0.27, "learning_rate": 4.544539978909096e-05, "loss": 0.0011, "step": 14084 }, { "epoch": 0.27, "learning_rate": 4.5444752828833724e-05, "loss": 0.0018, "step": 14086 }, { "epoch": 0.27, "learning_rate": 4.54441058685765e-05, "loss": 0.0317, "step": 14088 }, { "epoch": 0.27, "learning_rate": 4.544345890831926e-05, "loss": 0.0082, "step": 14090 }, { "epoch": 0.27, "learning_rate": 4.544281194806203e-05, "loss": 0.0007, "step": 14092 }, { "epoch": 0.27, "learning_rate": 4.54421649878048e-05, "loss": 0.0029, "step": 14094 }, { "epoch": 0.27, "learning_rate": 4.544151802754757e-05, "loss": 0.0022, "step": 14096 }, { "epoch": 0.27, "learning_rate": 4.544087106729034e-05, "loss": 0.0007, "step": 14098 }, { "epoch": 0.27, "learning_rate": 4.544022410703311e-05, "loss": 0.0196, "step": 14100 }, { "epoch": 0.27, "learning_rate": 4.5439577146775876e-05, "loss": 0.0006, "step": 14102 }, { "epoch": 0.27, "learning_rate": 4.5438930186518646e-05, "loss": 0.0142, "step": 14104 }, { "epoch": 0.27, "learning_rate": 4.5438283226261415e-05, "loss": 0.0048, "step": 14106 }, { "epoch": 0.27, "learning_rate": 4.543763626600418e-05, "loss": 0.0026, "step": 14108 }, { "epoch": 0.27, "learning_rate": 4.543698930574695e-05, "loss": 0.0024, "step": 14110 }, { "epoch": 0.27, "learning_rate": 4.543634234548972e-05, "loss": 0.0023, "step": 14112 }, { "epoch": 0.27, "learning_rate": 4.5435695385232484e-05, "loss": 0.0075, "step": 14114 }, { "epoch": 0.27, "learning_rate": 4.543504842497526e-05, "loss": 0.002, "step": 14116 }, { "epoch": 0.27, "learning_rate": 4.543440146471802e-05, "loss": 0.0031, "step": 14118 }, { "epoch": 0.27, "learning_rate": 4.543375450446079e-05, "loss": 0.0156, "step": 14120 }, { "epoch": 0.27, "learning_rate": 4.543310754420356e-05, "loss": 0.0064, "step": 14122 }, { "epoch": 0.27, "learning_rate": 4.543246058394633e-05, "loss": 0.0008, "step": 14124 }, { "epoch": 0.27, "learning_rate": 4.54318136236891e-05, "loss": 0.002, "step": 14126 }, { "epoch": 0.27, "learning_rate": 4.543116666343187e-05, "loss": 0.0079, "step": 14128 }, { "epoch": 0.27, "learning_rate": 4.543051970317464e-05, "loss": 0.0042, "step": 14130 }, { "epoch": 0.27, "learning_rate": 4.5429872742917406e-05, "loss": 0.0178, "step": 14132 }, { "epoch": 0.27, "learning_rate": 4.5429225782660175e-05, "loss": 0.0104, "step": 14134 }, { "epoch": 0.27, "learning_rate": 4.542857882240294e-05, "loss": 0.0011, "step": 14136 }, { "epoch": 0.27, "learning_rate": 4.542793186214571e-05, "loss": 0.0003, "step": 14138 }, { "epoch": 0.27, "learning_rate": 4.5427284901888475e-05, "loss": 0.0042, "step": 14140 }, { "epoch": 0.27, "learning_rate": 4.542663794163125e-05, "loss": 0.0011, "step": 14142 }, { "epoch": 0.27, "learning_rate": 4.5425990981374014e-05, "loss": 0.0147, "step": 14144 }, { "epoch": 0.27, "learning_rate": 4.542534402111678e-05, "loss": 0.0013, "step": 14146 }, { "epoch": 0.27, "learning_rate": 4.542469706085956e-05, "loss": 0.0022, "step": 14148 }, { "epoch": 0.27, "learning_rate": 4.542405010060232e-05, "loss": 0.0082, "step": 14150 }, { "epoch": 0.27, "learning_rate": 4.542340314034509e-05, "loss": 0.0227, "step": 14152 }, { "epoch": 0.27, "learning_rate": 4.542275618008786e-05, "loss": 0.0039, "step": 14154 }, { "epoch": 0.27, "learning_rate": 4.542210921983063e-05, "loss": 0.015, "step": 14156 }, { "epoch": 0.27, "learning_rate": 4.54214622595734e-05, "loss": 0.0082, "step": 14158 }, { "epoch": 0.27, "learning_rate": 4.5420815299316166e-05, "loss": 0.0023, "step": 14160 }, { "epoch": 0.27, "learning_rate": 4.5420168339058935e-05, "loss": 0.0013, "step": 14162 }, { "epoch": 0.27, "learning_rate": 4.5419521378801704e-05, "loss": 0.0021, "step": 14164 }, { "epoch": 0.27, "learning_rate": 4.5418874418544474e-05, "loss": 0.0232, "step": 14166 }, { "epoch": 0.27, "learning_rate": 4.5418227458287236e-05, "loss": 0.0286, "step": 14168 }, { "epoch": 0.28, "learning_rate": 4.541758049803001e-05, "loss": 0.0067, "step": 14170 }, { "epoch": 0.28, "learning_rate": 4.5416933537772774e-05, "loss": 0.0037, "step": 14172 }, { "epoch": 0.28, "learning_rate": 4.541628657751554e-05, "loss": 0.0006, "step": 14174 }, { "epoch": 0.28, "learning_rate": 4.541563961725831e-05, "loss": 0.001, "step": 14176 }, { "epoch": 0.28, "learning_rate": 4.541499265700108e-05, "loss": 0.0123, "step": 14178 }, { "epoch": 0.28, "learning_rate": 4.541434569674385e-05, "loss": 0.0077, "step": 14180 }, { "epoch": 0.28, "learning_rate": 4.541369873648662e-05, "loss": 0.0176, "step": 14182 }, { "epoch": 0.28, "learning_rate": 4.541305177622939e-05, "loss": 0.0015, "step": 14184 }, { "epoch": 0.28, "learning_rate": 4.541240481597216e-05, "loss": 0.001, "step": 14186 }, { "epoch": 0.28, "learning_rate": 4.541175785571493e-05, "loss": 0.0123, "step": 14188 }, { "epoch": 0.28, "learning_rate": 4.5411110895457696e-05, "loss": 0.0009, "step": 14190 }, { "epoch": 0.28, "learning_rate": 4.5410463935200465e-05, "loss": 0.0007, "step": 14192 }, { "epoch": 0.28, "learning_rate": 4.5409816974943234e-05, "loss": 0.0159, "step": 14194 }, { "epoch": 0.28, "learning_rate": 4.5409170014685996e-05, "loss": 0.0003, "step": 14196 }, { "epoch": 0.28, "learning_rate": 4.540852305442877e-05, "loss": 0.0246, "step": 14198 }, { "epoch": 0.28, "learning_rate": 4.5407876094171534e-05, "loss": 0.0009, "step": 14200 }, { "epoch": 0.28, "learning_rate": 4.540722913391431e-05, "loss": 0.0008, "step": 14202 }, { "epoch": 0.28, "learning_rate": 4.540658217365707e-05, "loss": 0.0007, "step": 14204 }, { "epoch": 0.28, "learning_rate": 4.540593521339984e-05, "loss": 0.0081, "step": 14206 }, { "epoch": 0.28, "learning_rate": 4.540528825314261e-05, "loss": 0.0008, "step": 14208 }, { "epoch": 0.28, "learning_rate": 4.540464129288538e-05, "loss": 0.0141, "step": 14210 }, { "epoch": 0.28, "learning_rate": 4.540399433262815e-05, "loss": 0.0144, "step": 14212 }, { "epoch": 0.28, "learning_rate": 4.540334737237092e-05, "loss": 0.0061, "step": 14214 }, { "epoch": 0.28, "learning_rate": 4.540270041211369e-05, "loss": 0.0015, "step": 14216 }, { "epoch": 0.28, "learning_rate": 4.540205345185645e-05, "loss": 0.0004, "step": 14218 }, { "epoch": 0.28, "learning_rate": 4.5401406491599225e-05, "loss": 0.0076, "step": 14220 }, { "epoch": 0.28, "learning_rate": 4.540075953134199e-05, "loss": 0.0083, "step": 14222 }, { "epoch": 0.28, "learning_rate": 4.5400112571084763e-05, "loss": 0.0008, "step": 14224 }, { "epoch": 0.28, "learning_rate": 4.539946561082753e-05, "loss": 0.0019, "step": 14226 }, { "epoch": 0.28, "learning_rate": 4.5398818650570295e-05, "loss": 0.0074, "step": 14228 }, { "epoch": 0.28, "learning_rate": 4.539817169031307e-05, "loss": 0.002, "step": 14230 }, { "epoch": 0.28, "learning_rate": 4.539752473005583e-05, "loss": 0.0013, "step": 14232 }, { "epoch": 0.28, "learning_rate": 4.53968777697986e-05, "loss": 0.0021, "step": 14234 }, { "epoch": 0.28, "learning_rate": 4.539623080954137e-05, "loss": 0.0281, "step": 14236 }, { "epoch": 0.28, "learning_rate": 4.539558384928414e-05, "loss": 0.0008, "step": 14238 }, { "epoch": 0.28, "learning_rate": 4.539493688902691e-05, "loss": 0.0226, "step": 14240 }, { "epoch": 0.28, "learning_rate": 4.539428992876968e-05, "loss": 0.0184, "step": 14242 }, { "epoch": 0.28, "learning_rate": 4.539364296851245e-05, "loss": 0.0026, "step": 14244 }, { "epoch": 0.28, "learning_rate": 4.539299600825522e-05, "loss": 0.0281, "step": 14246 }, { "epoch": 0.28, "learning_rate": 4.5392349047997986e-05, "loss": 0.0021, "step": 14248 }, { "epoch": 0.28, "learning_rate": 4.539170208774075e-05, "loss": 0.0035, "step": 14250 }, { "epoch": 0.28, "learning_rate": 4.5391055127483524e-05, "loss": 0.0067, "step": 14252 }, { "epoch": 0.28, "learning_rate": 4.5390408167226286e-05, "loss": 0.0259, "step": 14254 }, { "epoch": 0.28, "learning_rate": 4.5389761206969055e-05, "loss": 0.0004, "step": 14256 }, { "epoch": 0.28, "learning_rate": 4.538911424671183e-05, "loss": 0.0091, "step": 14258 }, { "epoch": 0.28, "learning_rate": 4.5388467286454593e-05, "loss": 0.0061, "step": 14260 }, { "epoch": 0.28, "learning_rate": 4.538782032619736e-05, "loss": 0.0028, "step": 14262 }, { "epoch": 0.28, "learning_rate": 4.538717336594013e-05, "loss": 0.0012, "step": 14264 }, { "epoch": 0.28, "learning_rate": 4.53865264056829e-05, "loss": 0.0057, "step": 14266 }, { "epoch": 0.28, "learning_rate": 4.538587944542567e-05, "loss": 0.0108, "step": 14268 }, { "epoch": 0.28, "learning_rate": 4.538523248516844e-05, "loss": 0.0007, "step": 14270 }, { "epoch": 0.28, "learning_rate": 4.538458552491121e-05, "loss": 0.0065, "step": 14272 }, { "epoch": 0.28, "learning_rate": 4.538393856465398e-05, "loss": 0.001, "step": 14274 }, { "epoch": 0.28, "learning_rate": 4.5383291604396746e-05, "loss": 0.1004, "step": 14276 }, { "epoch": 0.28, "learning_rate": 4.538264464413951e-05, "loss": 0.0113, "step": 14278 }, { "epoch": 0.28, "learning_rate": 4.5381997683882284e-05, "loss": 0.0125, "step": 14280 }, { "epoch": 0.28, "learning_rate": 4.5381350723625047e-05, "loss": 0.009, "step": 14282 }, { "epoch": 0.28, "learning_rate": 4.538070376336782e-05, "loss": 0.0033, "step": 14284 }, { "epoch": 0.28, "learning_rate": 4.5380056803110585e-05, "loss": 0.0094, "step": 14286 }, { "epoch": 0.28, "learning_rate": 4.5379409842853354e-05, "loss": 0.0067, "step": 14288 }, { "epoch": 0.28, "learning_rate": 4.537876288259613e-05, "loss": 0.0124, "step": 14290 }, { "epoch": 0.28, "learning_rate": 4.537811592233889e-05, "loss": 0.0004, "step": 14292 }, { "epoch": 0.28, "learning_rate": 4.537746896208166e-05, "loss": 0.0073, "step": 14294 }, { "epoch": 0.28, "learning_rate": 4.537682200182443e-05, "loss": 0.0016, "step": 14296 }, { "epoch": 0.28, "learning_rate": 4.53761750415672e-05, "loss": 0.0187, "step": 14298 }, { "epoch": 0.28, "learning_rate": 4.537552808130996e-05, "loss": 0.0053, "step": 14300 }, { "epoch": 0.28, "learning_rate": 4.537488112105274e-05, "loss": 0.0094, "step": 14302 }, { "epoch": 0.28, "learning_rate": 4.5374234160795507e-05, "loss": 0.0109, "step": 14304 }, { "epoch": 0.28, "learning_rate": 4.5373587200538276e-05, "loss": 0.0034, "step": 14306 }, { "epoch": 0.28, "learning_rate": 4.5372940240281045e-05, "loss": 0.0093, "step": 14308 }, { "epoch": 0.28, "learning_rate": 4.537229328002381e-05, "loss": 0.0067, "step": 14310 }, { "epoch": 0.28, "learning_rate": 4.537164631976658e-05, "loss": 0.001, "step": 14312 }, { "epoch": 0.28, "learning_rate": 4.5370999359509345e-05, "loss": 0.0042, "step": 14314 }, { "epoch": 0.28, "learning_rate": 4.5370352399252114e-05, "loss": 0.0261, "step": 14316 }, { "epoch": 0.28, "learning_rate": 4.536970543899488e-05, "loss": 0.0008, "step": 14318 }, { "epoch": 0.28, "learning_rate": 4.536905847873765e-05, "loss": 0.0119, "step": 14320 }, { "epoch": 0.28, "learning_rate": 4.536841151848042e-05, "loss": 0.0015, "step": 14322 }, { "epoch": 0.28, "learning_rate": 4.536776455822319e-05, "loss": 0.0129, "step": 14324 }, { "epoch": 0.28, "learning_rate": 4.536711759796596e-05, "loss": 0.0104, "step": 14326 }, { "epoch": 0.28, "learning_rate": 4.536647063770873e-05, "loss": 0.002, "step": 14328 }, { "epoch": 0.28, "learning_rate": 4.53658236774515e-05, "loss": 0.0058, "step": 14330 }, { "epoch": 0.28, "learning_rate": 4.536517671719426e-05, "loss": 0.0036, "step": 14332 }, { "epoch": 0.28, "learning_rate": 4.5364529756937036e-05, "loss": 0.0008, "step": 14334 }, { "epoch": 0.28, "learning_rate": 4.5363882796679805e-05, "loss": 0.001, "step": 14336 }, { "epoch": 0.28, "learning_rate": 4.536323583642257e-05, "loss": 0.0011, "step": 14338 }, { "epoch": 0.28, "learning_rate": 4.536258887616534e-05, "loss": 0.0009, "step": 14340 }, { "epoch": 0.28, "learning_rate": 4.5361941915908106e-05, "loss": 0.0135, "step": 14342 }, { "epoch": 0.28, "learning_rate": 4.536129495565088e-05, "loss": 0.001, "step": 14344 }, { "epoch": 0.28, "learning_rate": 4.5360647995393644e-05, "loss": 0.0012, "step": 14346 }, { "epoch": 0.28, "learning_rate": 4.536000103513641e-05, "loss": 0.0154, "step": 14348 }, { "epoch": 0.28, "learning_rate": 4.535935407487918e-05, "loss": 0.0009, "step": 14350 }, { "epoch": 0.28, "learning_rate": 4.535870711462195e-05, "loss": 0.0006, "step": 14352 }, { "epoch": 0.28, "learning_rate": 4.535806015436472e-05, "loss": 0.0098, "step": 14354 }, { "epoch": 0.28, "learning_rate": 4.535741319410749e-05, "loss": 0.0014, "step": 14356 }, { "epoch": 0.28, "learning_rate": 4.535676623385026e-05, "loss": 0.0096, "step": 14358 }, { "epoch": 0.28, "learning_rate": 4.535611927359302e-05, "loss": 0.0034, "step": 14360 }, { "epoch": 0.28, "learning_rate": 4.5355472313335796e-05, "loss": 0.0179, "step": 14362 }, { "epoch": 0.28, "learning_rate": 4.535482535307856e-05, "loss": 0.0079, "step": 14364 }, { "epoch": 0.28, "learning_rate": 4.5354178392821335e-05, "loss": 0.0481, "step": 14366 }, { "epoch": 0.28, "learning_rate": 4.5353531432564104e-05, "loss": 0.003, "step": 14368 }, { "epoch": 0.28, "learning_rate": 4.5352884472306866e-05, "loss": 0.016, "step": 14370 }, { "epoch": 0.28, "learning_rate": 4.535223751204964e-05, "loss": 0.008, "step": 14372 }, { "epoch": 0.28, "learning_rate": 4.5351590551792404e-05, "loss": 0.0077, "step": 14374 }, { "epoch": 0.28, "learning_rate": 4.535094359153517e-05, "loss": 0.0031, "step": 14376 }, { "epoch": 0.28, "learning_rate": 4.535029663127794e-05, "loss": 0.0035, "step": 14378 }, { "epoch": 0.28, "learning_rate": 4.534964967102071e-05, "loss": 0.0176, "step": 14380 }, { "epoch": 0.28, "learning_rate": 4.534900271076348e-05, "loss": 0.0003, "step": 14382 }, { "epoch": 0.28, "learning_rate": 4.534835575050625e-05, "loss": 0.0143, "step": 14384 }, { "epoch": 0.28, "learning_rate": 4.534770879024902e-05, "loss": 0.0002, "step": 14386 }, { "epoch": 0.28, "learning_rate": 4.534706182999179e-05, "loss": 0.0088, "step": 14388 }, { "epoch": 0.28, "learning_rate": 4.534641486973456e-05, "loss": 0.0003, "step": 14390 }, { "epoch": 0.28, "learning_rate": 4.534576790947732e-05, "loss": 0.0005, "step": 14392 }, { "epoch": 0.28, "learning_rate": 4.5345120949220095e-05, "loss": 0.0065, "step": 14394 }, { "epoch": 0.28, "learning_rate": 4.534447398896286e-05, "loss": 0.0006, "step": 14396 }, { "epoch": 0.28, "learning_rate": 4.5343827028705626e-05, "loss": 0.0113, "step": 14398 }, { "epoch": 0.28, "learning_rate": 4.5343180068448395e-05, "loss": 0.0058, "step": 14400 }, { "epoch": 0.28, "learning_rate": 4.5342533108191165e-05, "loss": 0.0047, "step": 14402 }, { "epoch": 0.28, "learning_rate": 4.5341886147933934e-05, "loss": 0.0063, "step": 14404 }, { "epoch": 0.28, "learning_rate": 4.53412391876767e-05, "loss": 0.0045, "step": 14406 }, { "epoch": 0.28, "learning_rate": 4.534059222741947e-05, "loss": 0.0006, "step": 14408 }, { "epoch": 0.28, "learning_rate": 4.533994526716224e-05, "loss": 0.0018, "step": 14410 }, { "epoch": 0.28, "learning_rate": 4.533929830690501e-05, "loss": 0.0017, "step": 14412 }, { "epoch": 0.28, "learning_rate": 4.533865134664778e-05, "loss": 0.0022, "step": 14414 }, { "epoch": 0.28, "learning_rate": 4.533800438639055e-05, "loss": 0.0081, "step": 14416 }, { "epoch": 0.28, "learning_rate": 4.533735742613332e-05, "loss": 0.0143, "step": 14418 }, { "epoch": 0.28, "learning_rate": 4.533671046587608e-05, "loss": 0.0011, "step": 14420 }, { "epoch": 0.28, "learning_rate": 4.5336063505618855e-05, "loss": 0.0015, "step": 14422 }, { "epoch": 0.28, "learning_rate": 4.533541654536162e-05, "loss": 0.0267, "step": 14424 }, { "epoch": 0.28, "learning_rate": 4.5334769585104394e-05, "loss": 0.0054, "step": 14426 }, { "epoch": 0.28, "learning_rate": 4.5334122624847156e-05, "loss": 0.0015, "step": 14428 }, { "epoch": 0.28, "learning_rate": 4.5333475664589925e-05, "loss": 0.0014, "step": 14430 }, { "epoch": 0.28, "learning_rate": 4.5332828704332694e-05, "loss": 0.0004, "step": 14432 }, { "epoch": 0.28, "learning_rate": 4.533218174407546e-05, "loss": 0.0019, "step": 14434 }, { "epoch": 0.28, "learning_rate": 4.533153478381823e-05, "loss": 0.0004, "step": 14436 }, { "epoch": 0.28, "learning_rate": 4.5330887823561e-05, "loss": 0.0025, "step": 14438 }, { "epoch": 0.28, "learning_rate": 4.533024086330377e-05, "loss": 0.0064, "step": 14440 }, { "epoch": 0.28, "learning_rate": 4.532959390304653e-05, "loss": 0.0218, "step": 14442 }, { "epoch": 0.28, "learning_rate": 4.532894694278931e-05, "loss": 0.0234, "step": 14444 }, { "epoch": 0.28, "learning_rate": 4.532829998253207e-05, "loss": 0.0004, "step": 14446 }, { "epoch": 0.28, "learning_rate": 4.532765302227485e-05, "loss": 0.0008, "step": 14448 }, { "epoch": 0.28, "learning_rate": 4.5327006062017616e-05, "loss": 0.0041, "step": 14450 }, { "epoch": 0.28, "learning_rate": 4.532635910176038e-05, "loss": 0.0026, "step": 14452 }, { "epoch": 0.28, "learning_rate": 4.5325712141503154e-05, "loss": 0.0049, "step": 14454 }, { "epoch": 0.28, "learning_rate": 4.5325065181245916e-05, "loss": 0.0092, "step": 14456 }, { "epoch": 0.28, "learning_rate": 4.5324418220988685e-05, "loss": 0.0354, "step": 14458 }, { "epoch": 0.28, "learning_rate": 4.5323771260731454e-05, "loss": 0.0169, "step": 14460 }, { "epoch": 0.28, "learning_rate": 4.5323124300474224e-05, "loss": 0.0203, "step": 14462 }, { "epoch": 0.28, "learning_rate": 4.532247734021699e-05, "loss": 0.0008, "step": 14464 }, { "epoch": 0.28, "learning_rate": 4.532183037995976e-05, "loss": 0.0006, "step": 14466 }, { "epoch": 0.28, "learning_rate": 4.532118341970253e-05, "loss": 0.0028, "step": 14468 }, { "epoch": 0.28, "learning_rate": 4.53205364594453e-05, "loss": 0.0012, "step": 14470 }, { "epoch": 0.28, "learning_rate": 4.531988949918807e-05, "loss": 0.0022, "step": 14472 }, { "epoch": 0.28, "learning_rate": 4.531924253893083e-05, "loss": 0.0014, "step": 14474 }, { "epoch": 0.28, "learning_rate": 4.531859557867361e-05, "loss": 0.002, "step": 14476 }, { "epoch": 0.28, "learning_rate": 4.531794861841637e-05, "loss": 0.0186, "step": 14478 }, { "epoch": 0.28, "learning_rate": 4.531730165815914e-05, "loss": 0.0179, "step": 14480 }, { "epoch": 0.28, "learning_rate": 4.5316654697901914e-05, "loss": 0.0002, "step": 14482 }, { "epoch": 0.28, "learning_rate": 4.531600773764468e-05, "loss": 0.0014, "step": 14484 }, { "epoch": 0.28, "learning_rate": 4.531536077738745e-05, "loss": 0.0015, "step": 14486 }, { "epoch": 0.28, "learning_rate": 4.5314713817130215e-05, "loss": 0.0002, "step": 14488 }, { "epoch": 0.28, "learning_rate": 4.5314066856872984e-05, "loss": 0.0167, "step": 14490 }, { "epoch": 0.28, "learning_rate": 4.531341989661575e-05, "loss": 0.0116, "step": 14492 }, { "epoch": 0.28, "learning_rate": 4.531277293635852e-05, "loss": 0.0006, "step": 14494 }, { "epoch": 0.28, "learning_rate": 4.531212597610129e-05, "loss": 0.0072, "step": 14496 }, { "epoch": 0.28, "learning_rate": 4.531147901584406e-05, "loss": 0.0003, "step": 14498 }, { "epoch": 0.28, "learning_rate": 4.531083205558683e-05, "loss": 0.0016, "step": 14500 }, { "epoch": 0.28, "learning_rate": 4.531018509532959e-05, "loss": 0.0099, "step": 14502 }, { "epoch": 0.28, "learning_rate": 4.530953813507237e-05, "loss": 0.0017, "step": 14504 }, { "epoch": 0.28, "learning_rate": 4.530889117481513e-05, "loss": 0.0007, "step": 14506 }, { "epoch": 0.28, "learning_rate": 4.5308244214557906e-05, "loss": 0.0193, "step": 14508 }, { "epoch": 0.28, "learning_rate": 4.530759725430067e-05, "loss": 0.0237, "step": 14510 }, { "epoch": 0.28, "learning_rate": 4.530695029404344e-05, "loss": 0.0071, "step": 14512 }, { "epoch": 0.28, "learning_rate": 4.530630333378621e-05, "loss": 0.0062, "step": 14514 }, { "epoch": 0.28, "learning_rate": 4.5305656373528975e-05, "loss": 0.0084, "step": 14516 }, { "epoch": 0.28, "learning_rate": 4.5305009413271744e-05, "loss": 0.0026, "step": 14518 }, { "epoch": 0.28, "learning_rate": 4.5304362453014513e-05, "loss": 0.0469, "step": 14520 }, { "epoch": 0.28, "learning_rate": 4.530371549275728e-05, "loss": 0.0013, "step": 14522 }, { "epoch": 0.28, "learning_rate": 4.5303068532500045e-05, "loss": 0.0356, "step": 14524 }, { "epoch": 0.28, "learning_rate": 4.530242157224282e-05, "loss": 0.0006, "step": 14526 }, { "epoch": 0.28, "learning_rate": 4.530177461198559e-05, "loss": 0.0097, "step": 14528 }, { "epoch": 0.28, "learning_rate": 4.530112765172836e-05, "loss": 0.0024, "step": 14530 }, { "epoch": 0.28, "learning_rate": 4.530048069147113e-05, "loss": 0.0079, "step": 14532 }, { "epoch": 0.28, "learning_rate": 4.529983373121389e-05, "loss": 0.0005, "step": 14534 }, { "epoch": 0.28, "learning_rate": 4.5299186770956666e-05, "loss": 0.0372, "step": 14536 }, { "epoch": 0.28, "learning_rate": 4.529853981069943e-05, "loss": 0.0061, "step": 14538 }, { "epoch": 0.28, "learning_rate": 4.52978928504422e-05, "loss": 0.0016, "step": 14540 }, { "epoch": 0.28, "learning_rate": 4.5297245890184967e-05, "loss": 0.0003, "step": 14542 }, { "epoch": 0.28, "learning_rate": 4.5296598929927736e-05, "loss": 0.0015, "step": 14544 }, { "epoch": 0.28, "learning_rate": 4.5295951969670505e-05, "loss": 0.0006, "step": 14546 }, { "epoch": 0.28, "learning_rate": 4.5295305009413274e-05, "loss": 0.0086, "step": 14548 }, { "epoch": 0.28, "learning_rate": 4.529465804915604e-05, "loss": 0.0114, "step": 14550 }, { "epoch": 0.28, "learning_rate": 4.529401108889881e-05, "loss": 0.0022, "step": 14552 }, { "epoch": 0.28, "learning_rate": 4.529336412864158e-05, "loss": 0.0055, "step": 14554 }, { "epoch": 0.28, "learning_rate": 4.529271716838434e-05, "loss": 0.0156, "step": 14556 }, { "epoch": 0.28, "learning_rate": 4.529207020812712e-05, "loss": 0.0084, "step": 14558 }, { "epoch": 0.28, "learning_rate": 4.529142324786989e-05, "loss": 0.0145, "step": 14560 }, { "epoch": 0.28, "learning_rate": 4.529077628761265e-05, "loss": 0.0007, "step": 14562 }, { "epoch": 0.28, "learning_rate": 4.5290129327355426e-05, "loss": 0.0051, "step": 14564 }, { "epoch": 0.28, "learning_rate": 4.528948236709819e-05, "loss": 0.0088, "step": 14566 }, { "epoch": 0.28, "learning_rate": 4.5288835406840965e-05, "loss": 0.0218, "step": 14568 }, { "epoch": 0.28, "learning_rate": 4.528818844658373e-05, "loss": 0.0178, "step": 14570 }, { "epoch": 0.28, "learning_rate": 4.5287541486326496e-05, "loss": 0.0422, "step": 14572 }, { "epoch": 0.28, "learning_rate": 4.5286894526069265e-05, "loss": 0.0043, "step": 14574 }, { "epoch": 0.28, "learning_rate": 4.5286247565812034e-05, "loss": 0.0028, "step": 14576 }, { "epoch": 0.28, "learning_rate": 4.52856006055548e-05, "loss": 0.0033, "step": 14578 }, { "epoch": 0.28, "learning_rate": 4.528495364529757e-05, "loss": 0.0392, "step": 14580 }, { "epoch": 0.28, "learning_rate": 4.528430668504034e-05, "loss": 0.0021, "step": 14582 }, { "epoch": 0.28, "learning_rate": 4.5283659724783104e-05, "loss": 0.0124, "step": 14584 }, { "epoch": 0.28, "learning_rate": 4.528301276452588e-05, "loss": 0.0093, "step": 14586 }, { "epoch": 0.28, "learning_rate": 4.528236580426864e-05, "loss": 0.002, "step": 14588 }, { "epoch": 0.28, "learning_rate": 4.528171884401142e-05, "loss": 0.0006, "step": 14590 }, { "epoch": 0.28, "learning_rate": 4.528107188375419e-05, "loss": 0.0072, "step": 14592 }, { "epoch": 0.28, "learning_rate": 4.528042492349695e-05, "loss": 0.0172, "step": 14594 }, { "epoch": 0.28, "learning_rate": 4.5279777963239725e-05, "loss": 0.0002, "step": 14596 }, { "epoch": 0.28, "learning_rate": 4.527913100298249e-05, "loss": 0.0006, "step": 14598 }, { "epoch": 0.28, "learning_rate": 4.5278484042725256e-05, "loss": 0.0076, "step": 14600 }, { "epoch": 0.28, "learning_rate": 4.5277837082468026e-05, "loss": 0.0451, "step": 14602 }, { "epoch": 0.28, "learning_rate": 4.5277190122210795e-05, "loss": 0.011, "step": 14604 }, { "epoch": 0.28, "learning_rate": 4.5276543161953564e-05, "loss": 0.0042, "step": 14606 }, { "epoch": 0.28, "learning_rate": 4.527589620169633e-05, "loss": 0.0063, "step": 14608 }, { "epoch": 0.28, "learning_rate": 4.52752492414391e-05, "loss": 0.0231, "step": 14610 }, { "epoch": 0.28, "learning_rate": 4.527460228118187e-05, "loss": 0.0007, "step": 14612 }, { "epoch": 0.28, "learning_rate": 4.527395532092464e-05, "loss": 0.0198, "step": 14614 }, { "epoch": 0.28, "learning_rate": 4.52733083606674e-05, "loss": 0.0005, "step": 14616 }, { "epoch": 0.28, "learning_rate": 4.527266140041018e-05, "loss": 0.0028, "step": 14618 }, { "epoch": 0.28, "learning_rate": 4.527201444015294e-05, "loss": 0.0047, "step": 14620 }, { "epoch": 0.28, "learning_rate": 4.527136747989571e-05, "loss": 0.0015, "step": 14622 }, { "epoch": 0.28, "learning_rate": 4.527072051963848e-05, "loss": 0.0006, "step": 14624 }, { "epoch": 0.28, "learning_rate": 4.527007355938125e-05, "loss": 0.007, "step": 14626 }, { "epoch": 0.28, "learning_rate": 4.5269426599124024e-05, "loss": 0.0117, "step": 14628 }, { "epoch": 0.28, "learning_rate": 4.5268779638866786e-05, "loss": 0.0024, "step": 14630 }, { "epoch": 0.28, "learning_rate": 4.5268132678609555e-05, "loss": 0.0089, "step": 14632 }, { "epoch": 0.28, "learning_rate": 4.5267485718352324e-05, "loss": 0.0036, "step": 14634 }, { "epoch": 0.28, "learning_rate": 4.526683875809509e-05, "loss": 0.0003, "step": 14636 }, { "epoch": 0.28, "learning_rate": 4.526619179783786e-05, "loss": 0.0263, "step": 14638 }, { "epoch": 0.28, "learning_rate": 4.526554483758063e-05, "loss": 0.0038, "step": 14640 }, { "epoch": 0.28, "learning_rate": 4.52648978773234e-05, "loss": 0.0095, "step": 14642 }, { "epoch": 0.28, "learning_rate": 4.526425091706616e-05, "loss": 0.0004, "step": 14644 }, { "epoch": 0.28, "learning_rate": 4.526360395680894e-05, "loss": 0.0023, "step": 14646 }, { "epoch": 0.28, "learning_rate": 4.52629569965517e-05, "loss": 0.0004, "step": 14648 }, { "epoch": 0.28, "learning_rate": 4.526231003629448e-05, "loss": 0.0186, "step": 14650 }, { "epoch": 0.28, "learning_rate": 4.526166307603724e-05, "loss": 0.0008, "step": 14652 }, { "epoch": 0.28, "learning_rate": 4.526101611578001e-05, "loss": 0.0046, "step": 14654 }, { "epoch": 0.28, "learning_rate": 4.526036915552278e-05, "loss": 0.0113, "step": 14656 }, { "epoch": 0.28, "learning_rate": 4.5259722195265546e-05, "loss": 0.003, "step": 14658 }, { "epoch": 0.28, "learning_rate": 4.5259075235008315e-05, "loss": 0.017, "step": 14660 }, { "epoch": 0.28, "learning_rate": 4.5258428274751085e-05, "loss": 0.0004, "step": 14662 }, { "epoch": 0.28, "learning_rate": 4.5257781314493854e-05, "loss": 0.0027, "step": 14664 }, { "epoch": 0.28, "learning_rate": 4.5257134354236616e-05, "loss": 0.0117, "step": 14666 }, { "epoch": 0.28, "learning_rate": 4.525648739397939e-05, "loss": 0.0011, "step": 14668 }, { "epoch": 0.28, "learning_rate": 4.525584043372216e-05, "loss": 0.0117, "step": 14670 }, { "epoch": 0.28, "learning_rate": 4.525519347346493e-05, "loss": 0.0036, "step": 14672 }, { "epoch": 0.28, "learning_rate": 4.52545465132077e-05, "loss": 0.0162, "step": 14674 }, { "epoch": 0.28, "learning_rate": 4.525389955295046e-05, "loss": 0.0055, "step": 14676 }, { "epoch": 0.28, "learning_rate": 4.525325259269324e-05, "loss": 0.0077, "step": 14678 }, { "epoch": 0.28, "learning_rate": 4.5252605632436e-05, "loss": 0.0028, "step": 14680 }, { "epoch": 0.28, "learning_rate": 4.525195867217877e-05, "loss": 0.0157, "step": 14682 }, { "epoch": 0.28, "learning_rate": 4.525131171192154e-05, "loss": 0.0117, "step": 14684 }, { "epoch": 0.29, "learning_rate": 4.525066475166431e-05, "loss": 0.0006, "step": 14686 }, { "epoch": 0.29, "learning_rate": 4.5250017791407076e-05, "loss": 0.015, "step": 14688 }, { "epoch": 0.29, "learning_rate": 4.5249370831149845e-05, "loss": 0.0028, "step": 14690 }, { "epoch": 0.29, "learning_rate": 4.5248723870892614e-05, "loss": 0.0078, "step": 14692 }, { "epoch": 0.29, "learning_rate": 4.524807691063538e-05, "loss": 0.0014, "step": 14694 }, { "epoch": 0.29, "learning_rate": 4.524742995037815e-05, "loss": 0.0021, "step": 14696 }, { "epoch": 0.29, "learning_rate": 4.5246782990120914e-05, "loss": 0.0044, "step": 14698 }, { "epoch": 0.29, "learning_rate": 4.524613602986369e-05, "loss": 0.0022, "step": 14700 }, { "epoch": 0.29, "learning_rate": 4.524548906960645e-05, "loss": 0.0004, "step": 14702 }, { "epoch": 0.29, "learning_rate": 4.524484210934922e-05, "loss": 0.0149, "step": 14704 }, { "epoch": 0.29, "learning_rate": 4.5244195149092e-05, "loss": 0.0059, "step": 14706 }, { "epoch": 0.29, "learning_rate": 4.524354818883476e-05, "loss": 0.0006, "step": 14708 }, { "epoch": 0.29, "learning_rate": 4.5242901228577536e-05, "loss": 0.0008, "step": 14710 }, { "epoch": 0.29, "learning_rate": 4.52422542683203e-05, "loss": 0.0013, "step": 14712 }, { "epoch": 0.29, "learning_rate": 4.524160730806307e-05, "loss": 0.0118, "step": 14714 }, { "epoch": 0.29, "learning_rate": 4.5240960347805836e-05, "loss": 0.0058, "step": 14716 }, { "epoch": 0.29, "learning_rate": 4.5240313387548605e-05, "loss": 0.0028, "step": 14718 }, { "epoch": 0.29, "learning_rate": 4.5239666427291374e-05, "loss": 0.0144, "step": 14720 }, { "epoch": 0.29, "learning_rate": 4.5239019467034144e-05, "loss": 0.0047, "step": 14722 }, { "epoch": 0.29, "learning_rate": 4.523837250677691e-05, "loss": 0.0176, "step": 14724 }, { "epoch": 0.29, "learning_rate": 4.5237725546519675e-05, "loss": 0.0051, "step": 14726 }, { "epoch": 0.29, "learning_rate": 4.523707858626245e-05, "loss": 0.0011, "step": 14728 }, { "epoch": 0.29, "learning_rate": 4.523643162600521e-05, "loss": 0.0008, "step": 14730 }, { "epoch": 0.29, "learning_rate": 4.523578466574799e-05, "loss": 0.0015, "step": 14732 }, { "epoch": 0.29, "learning_rate": 4.523513770549075e-05, "loss": 0.0059, "step": 14734 }, { "epoch": 0.29, "learning_rate": 4.523449074523352e-05, "loss": 0.0003, "step": 14736 }, { "epoch": 0.29, "learning_rate": 4.5233843784976296e-05, "loss": 0.0136, "step": 14738 }, { "epoch": 0.29, "learning_rate": 4.523319682471906e-05, "loss": 0.0112, "step": 14740 }, { "epoch": 0.29, "learning_rate": 4.523254986446183e-05, "loss": 0.0101, "step": 14742 }, { "epoch": 0.29, "learning_rate": 4.52319029042046e-05, "loss": 0.0034, "step": 14744 }, { "epoch": 0.29, "learning_rate": 4.5231255943947366e-05, "loss": 0.0005, "step": 14746 }, { "epoch": 0.29, "learning_rate": 4.5230608983690135e-05, "loss": 0.0005, "step": 14748 }, { "epoch": 0.29, "learning_rate": 4.5229962023432904e-05, "loss": 0.0016, "step": 14750 }, { "epoch": 0.29, "learning_rate": 4.522931506317567e-05, "loss": 0.0015, "step": 14752 }, { "epoch": 0.29, "learning_rate": 4.522866810291844e-05, "loss": 0.0146, "step": 14754 }, { "epoch": 0.29, "learning_rate": 4.522802114266121e-05, "loss": 0.0021, "step": 14756 }, { "epoch": 0.29, "learning_rate": 4.5227374182403973e-05, "loss": 0.0141, "step": 14758 }, { "epoch": 0.29, "learning_rate": 4.522672722214675e-05, "loss": 0.0007, "step": 14760 }, { "epoch": 0.29, "learning_rate": 4.522608026188951e-05, "loss": 0.0086, "step": 14762 }, { "epoch": 0.29, "learning_rate": 4.522543330163228e-05, "loss": 0.0071, "step": 14764 }, { "epoch": 0.29, "learning_rate": 4.522478634137505e-05, "loss": 0.027, "step": 14766 }, { "epoch": 0.29, "learning_rate": 4.522413938111782e-05, "loss": 0.0159, "step": 14768 }, { "epoch": 0.29, "learning_rate": 4.5223492420860595e-05, "loss": 0.0004, "step": 14770 }, { "epoch": 0.29, "learning_rate": 4.522284546060336e-05, "loss": 0.0025, "step": 14772 }, { "epoch": 0.29, "learning_rate": 4.5222198500346126e-05, "loss": 0.0004, "step": 14774 }, { "epoch": 0.29, "learning_rate": 4.5221551540088895e-05, "loss": 0.0012, "step": 14776 }, { "epoch": 0.29, "learning_rate": 4.5220904579831664e-05, "loss": 0.0014, "step": 14778 }, { "epoch": 0.29, "learning_rate": 4.522025761957443e-05, "loss": 0.0038, "step": 14780 }, { "epoch": 0.29, "learning_rate": 4.52196106593172e-05, "loss": 0.0103, "step": 14782 }, { "epoch": 0.29, "learning_rate": 4.521896369905997e-05, "loss": 0.0175, "step": 14784 }, { "epoch": 0.29, "learning_rate": 4.5218316738802734e-05, "loss": 0.0037, "step": 14786 }, { "epoch": 0.29, "learning_rate": 4.521766977854551e-05, "loss": 0.0055, "step": 14788 }, { "epoch": 0.29, "learning_rate": 4.521702281828827e-05, "loss": 0.0014, "step": 14790 }, { "epoch": 0.29, "learning_rate": 4.521637585803105e-05, "loss": 0.0036, "step": 14792 }, { "epoch": 0.29, "learning_rate": 4.521572889777381e-05, "loss": 0.0027, "step": 14794 }, { "epoch": 0.29, "learning_rate": 4.521508193751658e-05, "loss": 0.0219, "step": 14796 }, { "epoch": 0.29, "learning_rate": 4.521443497725935e-05, "loss": 0.0043, "step": 14798 }, { "epoch": 0.29, "learning_rate": 4.521378801700212e-05, "loss": 0.0061, "step": 14800 }, { "epoch": 0.29, "learning_rate": 4.5213141056744887e-05, "loss": 0.0158, "step": 14802 }, { "epoch": 0.29, "learning_rate": 4.5212494096487656e-05, "loss": 0.0166, "step": 14804 }, { "epoch": 0.29, "learning_rate": 4.5211847136230425e-05, "loss": 0.122, "step": 14806 }, { "epoch": 0.29, "learning_rate": 4.521120017597319e-05, "loss": 0.0005, "step": 14808 }, { "epoch": 0.29, "learning_rate": 4.521055321571596e-05, "loss": 0.0482, "step": 14810 }, { "epoch": 0.29, "learning_rate": 4.5209906255458725e-05, "loss": 0.0003, "step": 14812 }, { "epoch": 0.29, "learning_rate": 4.52092592952015e-05, "loss": 0.0131, "step": 14814 }, { "epoch": 0.29, "learning_rate": 4.520861233494427e-05, "loss": 0.0115, "step": 14816 }, { "epoch": 0.29, "learning_rate": 4.520796537468703e-05, "loss": 0.001, "step": 14818 }, { "epoch": 0.29, "learning_rate": 4.520731841442981e-05, "loss": 0.0077, "step": 14820 }, { "epoch": 0.29, "learning_rate": 4.520667145417257e-05, "loss": 0.0039, "step": 14822 }, { "epoch": 0.29, "learning_rate": 4.520602449391534e-05, "loss": 0.005, "step": 14824 }, { "epoch": 0.29, "learning_rate": 4.520537753365811e-05, "loss": 0.0005, "step": 14826 }, { "epoch": 0.29, "learning_rate": 4.520473057340088e-05, "loss": 0.0016, "step": 14828 }, { "epoch": 0.29, "learning_rate": 4.520408361314365e-05, "loss": 0.0022, "step": 14830 }, { "epoch": 0.29, "learning_rate": 4.5203436652886416e-05, "loss": 0.0194, "step": 14832 }, { "epoch": 0.29, "learning_rate": 4.5202789692629185e-05, "loss": 0.0003, "step": 14834 }, { "epoch": 0.29, "learning_rate": 4.5202142732371954e-05, "loss": 0.003, "step": 14836 }, { "epoch": 0.29, "learning_rate": 4.520149577211472e-05, "loss": 0.0013, "step": 14838 }, { "epoch": 0.29, "learning_rate": 4.5200848811857486e-05, "loss": 0.0004, "step": 14840 }, { "epoch": 0.29, "learning_rate": 4.520020185160026e-05, "loss": 0.0005, "step": 14842 }, { "epoch": 0.29, "learning_rate": 4.5199554891343024e-05, "loss": 0.0132, "step": 14844 }, { "epoch": 0.29, "learning_rate": 4.519890793108579e-05, "loss": 0.0008, "step": 14846 }, { "epoch": 0.29, "learning_rate": 4.519826097082857e-05, "loss": 0.0095, "step": 14848 }, { "epoch": 0.29, "learning_rate": 4.519761401057133e-05, "loss": 0.0049, "step": 14850 }, { "epoch": 0.29, "learning_rate": 4.519696705031411e-05, "loss": 0.0006, "step": 14852 }, { "epoch": 0.29, "learning_rate": 4.519632009005687e-05, "loss": 0.0083, "step": 14854 }, { "epoch": 0.29, "learning_rate": 4.519567312979964e-05, "loss": 0.0003, "step": 14856 }, { "epoch": 0.29, "learning_rate": 4.519502616954241e-05, "loss": 0.0057, "step": 14858 }, { "epoch": 0.29, "learning_rate": 4.5194379209285176e-05, "loss": 0.0042, "step": 14860 }, { "epoch": 0.29, "learning_rate": 4.5193732249027946e-05, "loss": 0.001, "step": 14862 }, { "epoch": 0.29, "learning_rate": 4.5193085288770715e-05, "loss": 0.0073, "step": 14864 }, { "epoch": 0.29, "learning_rate": 4.5192438328513484e-05, "loss": 0.0052, "step": 14866 }, { "epoch": 0.29, "learning_rate": 4.5191791368256246e-05, "loss": 0.0009, "step": 14868 }, { "epoch": 0.29, "learning_rate": 4.519114440799902e-05, "loss": 0.0066, "step": 14870 }, { "epoch": 0.29, "learning_rate": 4.5190497447741784e-05, "loss": 0.0197, "step": 14872 }, { "epoch": 0.29, "learning_rate": 4.518985048748456e-05, "loss": 0.0046, "step": 14874 }, { "epoch": 0.29, "learning_rate": 4.518920352722732e-05, "loss": 0.0103, "step": 14876 }, { "epoch": 0.29, "learning_rate": 4.518855656697009e-05, "loss": 0.0068, "step": 14878 }, { "epoch": 0.29, "learning_rate": 4.518790960671286e-05, "loss": 0.0008, "step": 14880 }, { "epoch": 0.29, "learning_rate": 4.518726264645563e-05, "loss": 0.003, "step": 14882 }, { "epoch": 0.29, "learning_rate": 4.51866156861984e-05, "loss": 0.0058, "step": 14884 }, { "epoch": 0.29, "learning_rate": 4.518596872594117e-05, "loss": 0.0228, "step": 14886 }, { "epoch": 0.29, "learning_rate": 4.518532176568394e-05, "loss": 0.0027, "step": 14888 }, { "epoch": 0.29, "learning_rate": 4.51846748054267e-05, "loss": 0.0008, "step": 14890 }, { "epoch": 0.29, "learning_rate": 4.5184027845169475e-05, "loss": 0.0013, "step": 14892 }, { "epoch": 0.29, "learning_rate": 4.5183380884912244e-05, "loss": 0.0013, "step": 14894 }, { "epoch": 0.29, "learning_rate": 4.518273392465501e-05, "loss": 0.0005, "step": 14896 }, { "epoch": 0.29, "learning_rate": 4.518208696439778e-05, "loss": 0.027, "step": 14898 }, { "epoch": 0.29, "learning_rate": 4.5181440004140545e-05, "loss": 0.0073, "step": 14900 }, { "epoch": 0.29, "learning_rate": 4.518079304388332e-05, "loss": 0.0023, "step": 14902 }, { "epoch": 0.29, "learning_rate": 4.518014608362608e-05, "loss": 0.0004, "step": 14904 }, { "epoch": 0.29, "learning_rate": 4.517949912336885e-05, "loss": 0.0014, "step": 14906 }, { "epoch": 0.29, "learning_rate": 4.517885216311162e-05, "loss": 0.0012, "step": 14908 }, { "epoch": 0.29, "learning_rate": 4.517820520285439e-05, "loss": 0.0007, "step": 14910 }, { "epoch": 0.29, "learning_rate": 4.517755824259716e-05, "loss": 0.0004, "step": 14912 }, { "epoch": 0.29, "learning_rate": 4.517691128233993e-05, "loss": 0.0016, "step": 14914 }, { "epoch": 0.29, "learning_rate": 4.51762643220827e-05, "loss": 0.0047, "step": 14916 }, { "epoch": 0.29, "learning_rate": 4.5175617361825466e-05, "loss": 0.0021, "step": 14918 }, { "epoch": 0.29, "learning_rate": 4.5174970401568235e-05, "loss": 0.0032, "step": 14920 }, { "epoch": 0.29, "learning_rate": 4.5174323441311e-05, "loss": 0.0005, "step": 14922 }, { "epoch": 0.29, "learning_rate": 4.5173676481053774e-05, "loss": 0.0161, "step": 14924 }, { "epoch": 0.29, "learning_rate": 4.5173029520796536e-05, "loss": 0.006, "step": 14926 }, { "epoch": 0.29, "learning_rate": 4.5172382560539305e-05, "loss": 0.0067, "step": 14928 }, { "epoch": 0.29, "learning_rate": 4.517173560028208e-05, "loss": 0.0063, "step": 14930 }, { "epoch": 0.29, "learning_rate": 4.517108864002484e-05, "loss": 0.0002, "step": 14932 }, { "epoch": 0.29, "learning_rate": 4.517044167976762e-05, "loss": 0.0013, "step": 14934 }, { "epoch": 0.29, "learning_rate": 4.516979471951038e-05, "loss": 0.0113, "step": 14936 }, { "epoch": 0.29, "learning_rate": 4.516914775925315e-05, "loss": 0.0019, "step": 14938 }, { "epoch": 0.29, "learning_rate": 4.516850079899592e-05, "loss": 0.0005, "step": 14940 }, { "epoch": 0.29, "learning_rate": 4.516785383873869e-05, "loss": 0.0067, "step": 14942 }, { "epoch": 0.29, "learning_rate": 4.516720687848146e-05, "loss": 0.0031, "step": 14944 }, { "epoch": 0.29, "learning_rate": 4.516655991822423e-05, "loss": 0.0056, "step": 14946 }, { "epoch": 0.29, "learning_rate": 4.5165912957966996e-05, "loss": 0.0004, "step": 14948 }, { "epoch": 0.29, "learning_rate": 4.516526599770976e-05, "loss": 0.0003, "step": 14950 }, { "epoch": 0.29, "learning_rate": 4.5164619037452534e-05, "loss": 0.0011, "step": 14952 }, { "epoch": 0.29, "learning_rate": 4.5163972077195296e-05, "loss": 0.0012, "step": 14954 }, { "epoch": 0.29, "learning_rate": 4.516332511693807e-05, "loss": 0.0161, "step": 14956 }, { "epoch": 0.29, "learning_rate": 4.5162678156680834e-05, "loss": 0.0232, "step": 14958 }, { "epoch": 0.29, "learning_rate": 4.5162031196423604e-05, "loss": 0.0074, "step": 14960 }, { "epoch": 0.29, "learning_rate": 4.516138423616638e-05, "loss": 0.0005, "step": 14962 }, { "epoch": 0.29, "learning_rate": 4.516073727590914e-05, "loss": 0.0039, "step": 14964 }, { "epoch": 0.29, "learning_rate": 4.516009031565191e-05, "loss": 0.0002, "step": 14966 }, { "epoch": 0.29, "learning_rate": 4.515944335539468e-05, "loss": 0.1598, "step": 14968 }, { "epoch": 0.29, "learning_rate": 4.515879639513745e-05, "loss": 0.0078, "step": 14970 }, { "epoch": 0.29, "learning_rate": 4.515814943488022e-05, "loss": 0.0071, "step": 14972 }, { "epoch": 0.29, "learning_rate": 4.515750247462299e-05, "loss": 0.0006, "step": 14974 }, { "epoch": 0.29, "learning_rate": 4.5156855514365756e-05, "loss": 0.0003, "step": 14976 }, { "epoch": 0.29, "learning_rate": 4.5156208554108525e-05, "loss": 0.0016, "step": 14978 }, { "epoch": 0.29, "learning_rate": 4.5155561593851294e-05, "loss": 0.0082, "step": 14980 }, { "epoch": 0.29, "learning_rate": 4.515491463359406e-05, "loss": 0.0133, "step": 14982 }, { "epoch": 0.29, "learning_rate": 4.515426767333683e-05, "loss": 0.0023, "step": 14984 }, { "epoch": 0.29, "learning_rate": 4.5153620713079595e-05, "loss": 0.0035, "step": 14986 }, { "epoch": 0.29, "learning_rate": 4.5152973752822364e-05, "loss": 0.0014, "step": 14988 }, { "epoch": 0.29, "learning_rate": 4.515232679256513e-05, "loss": 0.0003, "step": 14990 }, { "epoch": 0.29, "learning_rate": 4.51516798323079e-05, "loss": 0.0181, "step": 14992 }, { "epoch": 0.29, "learning_rate": 4.515103287205068e-05, "loss": 0.0225, "step": 14994 }, { "epoch": 0.29, "learning_rate": 4.515038591179344e-05, "loss": 0.0023, "step": 14996 }, { "epoch": 0.29, "learning_rate": 4.514973895153621e-05, "loss": 0.0111, "step": 14998 }, { "epoch": 0.29, "learning_rate": 4.514909199127898e-05, "loss": 0.0169, "step": 15000 }, { "epoch": 0.29, "learning_rate": 4.514844503102175e-05, "loss": 0.0006, "step": 15002 }, { "epoch": 0.29, "learning_rate": 4.514779807076451e-05, "loss": 0.0003, "step": 15004 }, { "epoch": 0.29, "learning_rate": 4.5147151110507286e-05, "loss": 0.0084, "step": 15006 }, { "epoch": 0.29, "learning_rate": 4.5146504150250055e-05, "loss": 0.0024, "step": 15008 }, { "epoch": 0.29, "learning_rate": 4.514585718999282e-05, "loss": 0.0049, "step": 15010 }, { "epoch": 0.29, "learning_rate": 4.514521022973559e-05, "loss": 0.02, "step": 15012 }, { "epoch": 0.29, "learning_rate": 4.5144563269478355e-05, "loss": 0.003, "step": 15014 }, { "epoch": 0.29, "learning_rate": 4.514391630922113e-05, "loss": 0.0055, "step": 15016 }, { "epoch": 0.29, "learning_rate": 4.5143269348963893e-05, "loss": 0.0054, "step": 15018 }, { "epoch": 0.29, "learning_rate": 4.514262238870666e-05, "loss": 0.0045, "step": 15020 }, { "epoch": 0.29, "learning_rate": 4.514197542844943e-05, "loss": 0.0089, "step": 15022 }, { "epoch": 0.29, "learning_rate": 4.51413284681922e-05, "loss": 0.0006, "step": 15024 }, { "epoch": 0.29, "learning_rate": 4.514068150793497e-05, "loss": 0.0019, "step": 15026 }, { "epoch": 0.29, "learning_rate": 4.514003454767774e-05, "loss": 0.0037, "step": 15028 }, { "epoch": 0.29, "learning_rate": 4.513938758742051e-05, "loss": 0.0008, "step": 15030 }, { "epoch": 0.29, "learning_rate": 4.513874062716327e-05, "loss": 0.0335, "step": 15032 }, { "epoch": 0.29, "learning_rate": 4.5138093666906046e-05, "loss": 0.0037, "step": 15034 }, { "epoch": 0.29, "learning_rate": 4.513744670664881e-05, "loss": 0.0011, "step": 15036 }, { "epoch": 0.29, "learning_rate": 4.5136799746391584e-05, "loss": 0.0007, "step": 15038 }, { "epoch": 0.29, "learning_rate": 4.513615278613435e-05, "loss": 0.0218, "step": 15040 }, { "epoch": 0.29, "learning_rate": 4.5135505825877116e-05, "loss": 0.0193, "step": 15042 }, { "epoch": 0.29, "learning_rate": 4.513485886561989e-05, "loss": 0.0015, "step": 15044 }, { "epoch": 0.29, "learning_rate": 4.5134211905362654e-05, "loss": 0.0022, "step": 15046 }, { "epoch": 0.29, "learning_rate": 4.513356494510542e-05, "loss": 0.0004, "step": 15048 }, { "epoch": 0.29, "learning_rate": 4.513291798484819e-05, "loss": 0.0029, "step": 15050 }, { "epoch": 0.29, "learning_rate": 4.513227102459096e-05, "loss": 0.014, "step": 15052 }, { "epoch": 0.29, "learning_rate": 4.513162406433373e-05, "loss": 0.0013, "step": 15054 }, { "epoch": 0.29, "learning_rate": 4.51309771040765e-05, "loss": 0.0076, "step": 15056 }, { "epoch": 0.29, "learning_rate": 4.513033014381927e-05, "loss": 0.0161, "step": 15058 }, { "epoch": 0.29, "learning_rate": 4.512968318356204e-05, "loss": 0.0177, "step": 15060 }, { "epoch": 0.29, "learning_rate": 4.5129036223304807e-05, "loss": 0.005, "step": 15062 }, { "epoch": 0.29, "learning_rate": 4.512838926304757e-05, "loss": 0.0011, "step": 15064 }, { "epoch": 0.29, "learning_rate": 4.5127742302790345e-05, "loss": 0.0123, "step": 15066 }, { "epoch": 0.29, "learning_rate": 4.512709534253311e-05, "loss": 0.0005, "step": 15068 }, { "epoch": 0.29, "learning_rate": 4.5126448382275876e-05, "loss": 0.0126, "step": 15070 }, { "epoch": 0.29, "learning_rate": 4.512580142201865e-05, "loss": 0.0007, "step": 15072 }, { "epoch": 0.29, "learning_rate": 4.5125154461761414e-05, "loss": 0.0114, "step": 15074 }, { "epoch": 0.29, "learning_rate": 4.512450750150419e-05, "loss": 0.0018, "step": 15076 }, { "epoch": 0.29, "learning_rate": 4.512386054124695e-05, "loss": 0.0007, "step": 15078 }, { "epoch": 0.29, "learning_rate": 4.512321358098972e-05, "loss": 0.0062, "step": 15080 }, { "epoch": 0.29, "learning_rate": 4.512256662073249e-05, "loss": 0.0075, "step": 15082 }, { "epoch": 0.29, "learning_rate": 4.512191966047526e-05, "loss": 0.0011, "step": 15084 }, { "epoch": 0.29, "learning_rate": 4.512127270021803e-05, "loss": 0.0161, "step": 15086 }, { "epoch": 0.29, "learning_rate": 4.51206257399608e-05, "loss": 0.0026, "step": 15088 }, { "epoch": 0.29, "learning_rate": 4.511997877970357e-05, "loss": 0.0002, "step": 15090 }, { "epoch": 0.29, "learning_rate": 4.511933181944633e-05, "loss": 0.0036, "step": 15092 }, { "epoch": 0.29, "learning_rate": 4.5118684859189105e-05, "loss": 0.0007, "step": 15094 }, { "epoch": 0.29, "learning_rate": 4.511803789893187e-05, "loss": 0.0066, "step": 15096 }, { "epoch": 0.29, "learning_rate": 4.511739093867464e-05, "loss": 0.0002, "step": 15098 }, { "epoch": 0.29, "learning_rate": 4.5116743978417406e-05, "loss": 0.0101, "step": 15100 }, { "epoch": 0.29, "learning_rate": 4.5116097018160175e-05, "loss": 0.0138, "step": 15102 }, { "epoch": 0.29, "learning_rate": 4.5115450057902944e-05, "loss": 0.0009, "step": 15104 }, { "epoch": 0.29, "learning_rate": 4.511480309764571e-05, "loss": 0.0092, "step": 15106 }, { "epoch": 0.29, "learning_rate": 4.511415613738848e-05, "loss": 0.0009, "step": 15108 }, { "epoch": 0.29, "learning_rate": 4.511350917713125e-05, "loss": 0.0044, "step": 15110 }, { "epoch": 0.29, "learning_rate": 4.511286221687402e-05, "loss": 0.0084, "step": 15112 }, { "epoch": 0.29, "learning_rate": 4.511221525661678e-05, "loss": 0.0023, "step": 15114 }, { "epoch": 0.29, "learning_rate": 4.511156829635956e-05, "loss": 0.0162, "step": 15116 }, { "epoch": 0.29, "learning_rate": 4.511092133610233e-05, "loss": 0.0009, "step": 15118 }, { "epoch": 0.29, "learning_rate": 4.5110274375845096e-05, "loss": 0.0022, "step": 15120 }, { "epoch": 0.29, "learning_rate": 4.5109627415587866e-05, "loss": 0.0345, "step": 15122 }, { "epoch": 0.29, "learning_rate": 4.510898045533063e-05, "loss": 0.0081, "step": 15124 }, { "epoch": 0.29, "learning_rate": 4.5108333495073404e-05, "loss": 0.0008, "step": 15126 }, { "epoch": 0.29, "learning_rate": 4.5107686534816166e-05, "loss": 0.0038, "step": 15128 }, { "epoch": 0.29, "learning_rate": 4.5107039574558935e-05, "loss": 0.0936, "step": 15130 }, { "epoch": 0.29, "learning_rate": 4.5106392614301704e-05, "loss": 0.0013, "step": 15132 }, { "epoch": 0.29, "learning_rate": 4.510574565404447e-05, "loss": 0.011, "step": 15134 }, { "epoch": 0.29, "learning_rate": 4.510509869378724e-05, "loss": 0.0009, "step": 15136 }, { "epoch": 0.29, "learning_rate": 4.510445173353001e-05, "loss": 0.001, "step": 15138 }, { "epoch": 0.29, "learning_rate": 4.510380477327278e-05, "loss": 0.0067, "step": 15140 }, { "epoch": 0.29, "learning_rate": 4.510315781301555e-05, "loss": 0.005, "step": 15142 }, { "epoch": 0.29, "learning_rate": 4.510251085275832e-05, "loss": 0.0112, "step": 15144 }, { "epoch": 0.29, "learning_rate": 4.510186389250108e-05, "loss": 0.0012, "step": 15146 }, { "epoch": 0.29, "learning_rate": 4.510121693224386e-05, "loss": 0.0005, "step": 15148 }, { "epoch": 0.29, "learning_rate": 4.5100569971986626e-05, "loss": 0.0032, "step": 15150 }, { "epoch": 0.29, "learning_rate": 4.509992301172939e-05, "loss": 0.0001, "step": 15152 }, { "epoch": 0.29, "learning_rate": 4.5099276051472164e-05, "loss": 0.0002, "step": 15154 }, { "epoch": 0.29, "learning_rate": 4.5098629091214926e-05, "loss": 0.0077, "step": 15156 }, { "epoch": 0.29, "learning_rate": 4.50979821309577e-05, "loss": 0.0319, "step": 15158 }, { "epoch": 0.29, "learning_rate": 4.5097335170700465e-05, "loss": 0.0093, "step": 15160 }, { "epoch": 0.29, "learning_rate": 4.5096688210443234e-05, "loss": 0.0232, "step": 15162 }, { "epoch": 0.29, "learning_rate": 4.5096041250186e-05, "loss": 0.0004, "step": 15164 }, { "epoch": 0.29, "learning_rate": 4.509539428992877e-05, "loss": 0.002, "step": 15166 }, { "epoch": 0.29, "learning_rate": 4.509474732967154e-05, "loss": 0.0044, "step": 15168 }, { "epoch": 0.29, "learning_rate": 4.509410036941431e-05, "loss": 0.0047, "step": 15170 }, { "epoch": 0.29, "learning_rate": 4.509345340915708e-05, "loss": 0.0005, "step": 15172 }, { "epoch": 0.29, "learning_rate": 4.509280644889984e-05, "loss": 0.0063, "step": 15174 }, { "epoch": 0.29, "learning_rate": 4.509215948864262e-05, "loss": 0.0139, "step": 15176 }, { "epoch": 0.29, "learning_rate": 4.509151252838538e-05, "loss": 0.0351, "step": 15178 }, { "epoch": 0.29, "learning_rate": 4.5090865568128155e-05, "loss": 0.0032, "step": 15180 }, { "epoch": 0.29, "learning_rate": 4.509021860787092e-05, "loss": 0.001, "step": 15182 }, { "epoch": 0.29, "learning_rate": 4.508957164761369e-05, "loss": 0.0152, "step": 15184 }, { "epoch": 0.29, "learning_rate": 4.508892468735646e-05, "loss": 0.0019, "step": 15186 }, { "epoch": 0.29, "learning_rate": 4.5088277727099225e-05, "loss": 0.0071, "step": 15188 }, { "epoch": 0.29, "learning_rate": 4.5087630766841994e-05, "loss": 0.0003, "step": 15190 }, { "epoch": 0.29, "learning_rate": 4.508698380658476e-05, "loss": 0.002, "step": 15192 }, { "epoch": 0.29, "learning_rate": 4.508633684632753e-05, "loss": 0.0061, "step": 15194 }, { "epoch": 0.29, "learning_rate": 4.50856898860703e-05, "loss": 0.0114, "step": 15196 }, { "epoch": 0.29, "learning_rate": 4.508504292581307e-05, "loss": 0.0009, "step": 15198 }, { "epoch": 0.3, "learning_rate": 4.508439596555584e-05, "loss": 0.0035, "step": 15200 }, { "epoch": 0.3, "learning_rate": 4.508374900529861e-05, "loss": 0.0004, "step": 15202 }, { "epoch": 0.3, "learning_rate": 4.508310204504138e-05, "loss": 0.0005, "step": 15204 }, { "epoch": 0.3, "learning_rate": 4.508245508478414e-05, "loss": 0.0021, "step": 15206 }, { "epoch": 0.3, "learning_rate": 4.5081808124526916e-05, "loss": 0.0005, "step": 15208 }, { "epoch": 0.3, "learning_rate": 4.508116116426968e-05, "loss": 0.0035, "step": 15210 }, { "epoch": 0.3, "learning_rate": 4.508051420401245e-05, "loss": 0.0353, "step": 15212 }, { "epoch": 0.3, "learning_rate": 4.5079867243755216e-05, "loss": 0.0033, "step": 15214 }, { "epoch": 0.3, "learning_rate": 4.5079220283497985e-05, "loss": 0.0002, "step": 15216 }, { "epoch": 0.3, "learning_rate": 4.507857332324076e-05, "loss": 0.0062, "step": 15218 }, { "epoch": 0.3, "learning_rate": 4.5077926362983524e-05, "loss": 0.0011, "step": 15220 }, { "epoch": 0.3, "learning_rate": 4.507727940272629e-05, "loss": 0.0044, "step": 15222 }, { "epoch": 0.3, "learning_rate": 4.507663244246906e-05, "loss": 0.0029, "step": 15224 }, { "epoch": 0.3, "learning_rate": 4.507598548221183e-05, "loss": 0.0089, "step": 15226 }, { "epoch": 0.3, "learning_rate": 4.50753385219546e-05, "loss": 0.0008, "step": 15228 }, { "epoch": 0.3, "learning_rate": 4.507469156169737e-05, "loss": 0.0022, "step": 15230 }, { "epoch": 0.3, "learning_rate": 4.507404460144014e-05, "loss": 0.0003, "step": 15232 }, { "epoch": 0.3, "learning_rate": 4.50733976411829e-05, "loss": 0.003, "step": 15234 }, { "epoch": 0.3, "learning_rate": 4.5072750680925676e-05, "loss": 0.0011, "step": 15236 }, { "epoch": 0.3, "learning_rate": 4.507210372066844e-05, "loss": 0.0073, "step": 15238 }, { "epoch": 0.3, "learning_rate": 4.5071456760411214e-05, "loss": 0.0062, "step": 15240 }, { "epoch": 0.3, "learning_rate": 4.507080980015398e-05, "loss": 0.0004, "step": 15242 }, { "epoch": 0.3, "learning_rate": 4.5070162839896746e-05, "loss": 0.0198, "step": 15244 }, { "epoch": 0.3, "learning_rate": 4.5069515879639515e-05, "loss": 0.0006, "step": 15246 }, { "epoch": 0.3, "learning_rate": 4.5068868919382284e-05, "loss": 0.0003, "step": 15248 }, { "epoch": 0.3, "learning_rate": 4.506822195912505e-05, "loss": 0.0003, "step": 15250 }, { "epoch": 0.3, "learning_rate": 4.506757499886782e-05, "loss": 0.0039, "step": 15252 }, { "epoch": 0.3, "learning_rate": 4.506692803861059e-05, "loss": 0.0134, "step": 15254 }, { "epoch": 0.3, "learning_rate": 4.5066281078353354e-05, "loss": 0.0078, "step": 15256 }, { "epoch": 0.3, "learning_rate": 4.506563411809613e-05, "loss": 0.0002, "step": 15258 }, { "epoch": 0.3, "learning_rate": 4.506498715783889e-05, "loss": 0.0098, "step": 15260 }, { "epoch": 0.3, "learning_rate": 4.506434019758167e-05, "loss": 0.0002, "step": 15262 }, { "epoch": 0.3, "learning_rate": 4.506369323732444e-05, "loss": 0.0004, "step": 15264 }, { "epoch": 0.3, "learning_rate": 4.50630462770672e-05, "loss": 0.0028, "step": 15266 }, { "epoch": 0.3, "learning_rate": 4.5062399316809975e-05, "loss": 0.0027, "step": 15268 }, { "epoch": 0.3, "learning_rate": 4.506175235655274e-05, "loss": 0.0015, "step": 15270 }, { "epoch": 0.3, "learning_rate": 4.5061105396295506e-05, "loss": 0.0113, "step": 15272 }, { "epoch": 0.3, "learning_rate": 4.5060458436038275e-05, "loss": 0.0021, "step": 15274 }, { "epoch": 0.3, "learning_rate": 4.5059811475781044e-05, "loss": 0.0002, "step": 15276 }, { "epoch": 0.3, "learning_rate": 4.5059164515523813e-05, "loss": 0.0186, "step": 15278 }, { "epoch": 0.3, "learning_rate": 4.505851755526658e-05, "loss": 0.001, "step": 15280 }, { "epoch": 0.3, "learning_rate": 4.505787059500935e-05, "loss": 0.0002, "step": 15282 }, { "epoch": 0.3, "learning_rate": 4.505722363475212e-05, "loss": 0.0056, "step": 15284 }, { "epoch": 0.3, "learning_rate": 4.505657667449489e-05, "loss": 0.0036, "step": 15286 }, { "epoch": 0.3, "learning_rate": 4.505592971423765e-05, "loss": 0.0007, "step": 15288 }, { "epoch": 0.3, "learning_rate": 4.505528275398043e-05, "loss": 0.0117, "step": 15290 }, { "epoch": 0.3, "learning_rate": 4.505463579372319e-05, "loss": 0.0043, "step": 15292 }, { "epoch": 0.3, "learning_rate": 4.505398883346596e-05, "loss": 0.0003, "step": 15294 }, { "epoch": 0.3, "learning_rate": 4.5053341873208735e-05, "loss": 0.0021, "step": 15296 }, { "epoch": 0.3, "learning_rate": 4.50526949129515e-05, "loss": 0.0004, "step": 15298 }, { "epoch": 0.3, "learning_rate": 4.505204795269427e-05, "loss": 0.0006, "step": 15300 }, { "epoch": 0.3, "learning_rate": 4.5051400992437036e-05, "loss": 0.0004, "step": 15302 }, { "epoch": 0.3, "learning_rate": 4.5050754032179805e-05, "loss": 0.0036, "step": 15304 }, { "epoch": 0.3, "learning_rate": 4.5050107071922574e-05, "loss": 0.0004, "step": 15306 }, { "epoch": 0.3, "learning_rate": 4.504946011166534e-05, "loss": 0.0003, "step": 15308 }, { "epoch": 0.3, "learning_rate": 4.504881315140811e-05, "loss": 0.0031, "step": 15310 }, { "epoch": 0.3, "learning_rate": 4.504816619115088e-05, "loss": 0.0103, "step": 15312 }, { "epoch": 0.3, "learning_rate": 4.504751923089365e-05, "loss": 0.0004, "step": 15314 }, { "epoch": 0.3, "learning_rate": 4.504687227063641e-05, "loss": 0.0152, "step": 15316 }, { "epoch": 0.3, "learning_rate": 4.504622531037919e-05, "loss": 0.0028, "step": 15318 }, { "epoch": 0.3, "learning_rate": 4.504557835012195e-05, "loss": 0.0042, "step": 15320 }, { "epoch": 0.3, "learning_rate": 4.5044931389864727e-05, "loss": 0.0237, "step": 15322 }, { "epoch": 0.3, "learning_rate": 4.504428442960749e-05, "loss": 0.0003, "step": 15324 }, { "epoch": 0.3, "learning_rate": 4.504363746935026e-05, "loss": 0.0011, "step": 15326 }, { "epoch": 0.3, "learning_rate": 4.5042990509093034e-05, "loss": 0.0002, "step": 15328 }, { "epoch": 0.3, "learning_rate": 4.5042343548835796e-05, "loss": 0.0011, "step": 15330 }, { "epoch": 0.3, "learning_rate": 4.5041696588578565e-05, "loss": 0.0007, "step": 15332 }, { "epoch": 0.3, "learning_rate": 4.5041049628321334e-05, "loss": 0.0208, "step": 15334 }, { "epoch": 0.3, "learning_rate": 4.50404026680641e-05, "loss": 0.0004, "step": 15336 }, { "epoch": 0.3, "learning_rate": 4.5039755707806866e-05, "loss": 0.004, "step": 15338 }, { "epoch": 0.3, "learning_rate": 4.503910874754964e-05, "loss": 0.0001, "step": 15340 }, { "epoch": 0.3, "learning_rate": 4.503846178729241e-05, "loss": 0.0012, "step": 15342 }, { "epoch": 0.3, "learning_rate": 4.503781482703518e-05, "loss": 0.0017, "step": 15344 }, { "epoch": 0.3, "learning_rate": 4.503716786677795e-05, "loss": 0.0008, "step": 15346 }, { "epoch": 0.3, "learning_rate": 4.503652090652071e-05, "loss": 0.0002, "step": 15348 }, { "epoch": 0.3, "learning_rate": 4.503587394626349e-05, "loss": 0.0062, "step": 15350 }, { "epoch": 0.3, "learning_rate": 4.503522698600625e-05, "loss": 0.0024, "step": 15352 }, { "epoch": 0.3, "learning_rate": 4.503458002574902e-05, "loss": 0.0014, "step": 15354 }, { "epoch": 0.3, "learning_rate": 4.503393306549179e-05, "loss": 0.0069, "step": 15356 }, { "epoch": 0.3, "learning_rate": 4.5033286105234556e-05, "loss": 0.004, "step": 15358 }, { "epoch": 0.3, "learning_rate": 4.5032639144977326e-05, "loss": 0.0016, "step": 15360 }, { "epoch": 0.3, "learning_rate": 4.5031992184720095e-05, "loss": 0.0027, "step": 15362 }, { "epoch": 0.3, "learning_rate": 4.5031345224462864e-05, "loss": 0.0015, "step": 15364 }, { "epoch": 0.3, "learning_rate": 4.503069826420563e-05, "loss": 0.0051, "step": 15366 }, { "epoch": 0.3, "learning_rate": 4.50300513039484e-05, "loss": 0.0032, "step": 15368 }, { "epoch": 0.3, "learning_rate": 4.5029404343691164e-05, "loss": 0.0145, "step": 15370 }, { "epoch": 0.3, "learning_rate": 4.502875738343394e-05, "loss": 0.0037, "step": 15372 }, { "epoch": 0.3, "learning_rate": 4.502811042317671e-05, "loss": 0.0059, "step": 15374 }, { "epoch": 0.3, "learning_rate": 4.502746346291947e-05, "loss": 0.0015, "step": 15376 }, { "epoch": 0.3, "learning_rate": 4.502681650266225e-05, "loss": 0.0025, "step": 15378 }, { "epoch": 0.3, "learning_rate": 4.502616954240501e-05, "loss": 0.001, "step": 15380 }, { "epoch": 0.3, "learning_rate": 4.5025522582147785e-05, "loss": 0.0196, "step": 15382 }, { "epoch": 0.3, "learning_rate": 4.502487562189055e-05, "loss": 0.001, "step": 15384 }, { "epoch": 0.3, "learning_rate": 4.502422866163332e-05, "loss": 0.002, "step": 15386 }, { "epoch": 0.3, "learning_rate": 4.5023581701376086e-05, "loss": 0.0204, "step": 15388 }, { "epoch": 0.3, "learning_rate": 4.5022934741118855e-05, "loss": 0.0015, "step": 15390 }, { "epoch": 0.3, "learning_rate": 4.5022287780861624e-05, "loss": 0.0072, "step": 15392 }, { "epoch": 0.3, "learning_rate": 4.502164082060439e-05, "loss": 0.0009, "step": 15394 }, { "epoch": 0.3, "learning_rate": 4.502099386034716e-05, "loss": 0.0194, "step": 15396 }, { "epoch": 0.3, "learning_rate": 4.5020346900089925e-05, "loss": 0.0028, "step": 15398 }, { "epoch": 0.3, "learning_rate": 4.50196999398327e-05, "loss": 0.0018, "step": 15400 }, { "epoch": 0.3, "learning_rate": 4.501905297957546e-05, "loss": 0.0009, "step": 15402 }, { "epoch": 0.3, "learning_rate": 4.501840601931824e-05, "loss": 0.0232, "step": 15404 }, { "epoch": 0.3, "learning_rate": 4.5017759059061e-05, "loss": 0.0004, "step": 15406 }, { "epoch": 0.3, "learning_rate": 4.501711209880377e-05, "loss": 0.0006, "step": 15408 }, { "epoch": 0.3, "learning_rate": 4.5016465138546546e-05, "loss": 0.0476, "step": 15410 }, { "epoch": 0.3, "learning_rate": 4.501581817828931e-05, "loss": 0.0027, "step": 15412 }, { "epoch": 0.3, "learning_rate": 4.501517121803208e-05, "loss": 0.0232, "step": 15414 }, { "epoch": 0.3, "learning_rate": 4.5014524257774846e-05, "loss": 0.0012, "step": 15416 }, { "epoch": 0.3, "learning_rate": 4.5013877297517615e-05, "loss": 0.0178, "step": 15418 }, { "epoch": 0.3, "learning_rate": 4.5013230337260385e-05, "loss": 0.0094, "step": 15420 }, { "epoch": 0.3, "learning_rate": 4.5012583377003154e-05, "loss": 0.0003, "step": 15422 }, { "epoch": 0.3, "learning_rate": 4.501193641674592e-05, "loss": 0.001, "step": 15424 }, { "epoch": 0.3, "learning_rate": 4.501128945648869e-05, "loss": 0.0012, "step": 15426 }, { "epoch": 0.3, "learning_rate": 4.501064249623146e-05, "loss": 0.0009, "step": 15428 }, { "epoch": 0.3, "learning_rate": 4.500999553597422e-05, "loss": 0.0008, "step": 15430 }, { "epoch": 0.3, "learning_rate": 4.5009348575717e-05, "loss": 0.0004, "step": 15432 }, { "epoch": 0.3, "learning_rate": 4.500870161545976e-05, "loss": 0.0131, "step": 15434 }, { "epoch": 0.3, "learning_rate": 4.500805465520253e-05, "loss": 0.0276, "step": 15436 }, { "epoch": 0.3, "learning_rate": 4.50074076949453e-05, "loss": 0.0405, "step": 15438 }, { "epoch": 0.3, "learning_rate": 4.500676073468807e-05, "loss": 0.0003, "step": 15440 }, { "epoch": 0.3, "learning_rate": 4.5006113774430844e-05, "loss": 0.0143, "step": 15442 }, { "epoch": 0.3, "learning_rate": 4.500546681417361e-05, "loss": 0.0008, "step": 15444 }, { "epoch": 0.3, "learning_rate": 4.5004819853916376e-05, "loss": 0.0004, "step": 15446 }, { "epoch": 0.3, "learning_rate": 4.5004172893659145e-05, "loss": 0.0064, "step": 15448 }, { "epoch": 0.3, "learning_rate": 4.5003525933401914e-05, "loss": 0.0383, "step": 15450 }, { "epoch": 0.3, "learning_rate": 4.500287897314468e-05, "loss": 0.0001, "step": 15452 }, { "epoch": 0.3, "learning_rate": 4.500223201288745e-05, "loss": 0.0319, "step": 15454 }, { "epoch": 0.3, "learning_rate": 4.500158505263022e-05, "loss": 0.0018, "step": 15456 }, { "epoch": 0.3, "learning_rate": 4.5000938092372984e-05, "loss": 0.0002, "step": 15458 }, { "epoch": 0.3, "learning_rate": 4.500029113211576e-05, "loss": 0.0131, "step": 15460 }, { "epoch": 0.3, "learning_rate": 4.499964417185852e-05, "loss": 0.0013, "step": 15462 }, { "epoch": 0.3, "learning_rate": 4.49989972116013e-05, "loss": 0.0012, "step": 15464 }, { "epoch": 0.3, "learning_rate": 4.499835025134406e-05, "loss": 0.0003, "step": 15466 }, { "epoch": 0.3, "learning_rate": 4.499770329108683e-05, "loss": 0.0186, "step": 15468 }, { "epoch": 0.3, "learning_rate": 4.49970563308296e-05, "loss": 0.0102, "step": 15470 }, { "epoch": 0.3, "learning_rate": 4.499640937057237e-05, "loss": 0.0086, "step": 15472 }, { "epoch": 0.3, "learning_rate": 4.4995762410315136e-05, "loss": 0.0035, "step": 15474 }, { "epoch": 0.3, "learning_rate": 4.4995115450057905e-05, "loss": 0.0305, "step": 15476 }, { "epoch": 0.3, "learning_rate": 4.4994468489800674e-05, "loss": 0.0053, "step": 15478 }, { "epoch": 0.3, "learning_rate": 4.499382152954344e-05, "loss": 0.0266, "step": 15480 }, { "epoch": 0.3, "learning_rate": 4.499317456928621e-05, "loss": 0.0003, "step": 15482 }, { "epoch": 0.3, "learning_rate": 4.4992527609028975e-05, "loss": 0.0006, "step": 15484 }, { "epoch": 0.3, "learning_rate": 4.499188064877175e-05, "loss": 0.0071, "step": 15486 }, { "epoch": 0.3, "learning_rate": 4.499123368851452e-05, "loss": 0.0021, "step": 15488 }, { "epoch": 0.3, "learning_rate": 4.499058672825728e-05, "loss": 0.0016, "step": 15490 }, { "epoch": 0.3, "learning_rate": 4.498993976800006e-05, "loss": 0.0158, "step": 15492 }, { "epoch": 0.3, "learning_rate": 4.498929280774282e-05, "loss": 0.0027, "step": 15494 }, { "epoch": 0.3, "learning_rate": 4.498864584748559e-05, "loss": 0.0013, "step": 15496 }, { "epoch": 0.3, "learning_rate": 4.498799888722836e-05, "loss": 0.0003, "step": 15498 }, { "epoch": 0.3, "learning_rate": 4.498735192697113e-05, "loss": 0.0025, "step": 15500 }, { "epoch": 0.3, "learning_rate": 4.49867049667139e-05, "loss": 0.0037, "step": 15502 }, { "epoch": 0.3, "learning_rate": 4.4986058006456666e-05, "loss": 0.0009, "step": 15504 }, { "epoch": 0.3, "learning_rate": 4.4985411046199435e-05, "loss": 0.009, "step": 15506 }, { "epoch": 0.3, "learning_rate": 4.4984764085942204e-05, "loss": 0.0014, "step": 15508 }, { "epoch": 0.3, "learning_rate": 4.498411712568497e-05, "loss": 0.0264, "step": 15510 }, { "epoch": 0.3, "learning_rate": 4.4983470165427735e-05, "loss": 0.0177, "step": 15512 }, { "epoch": 0.3, "learning_rate": 4.498282320517051e-05, "loss": 0.0114, "step": 15514 }, { "epoch": 0.3, "learning_rate": 4.4982176244913273e-05, "loss": 0.0014, "step": 15516 }, { "epoch": 0.3, "learning_rate": 4.498152928465604e-05, "loss": 0.0078, "step": 15518 }, { "epoch": 0.3, "learning_rate": 4.498088232439882e-05, "loss": 0.0004, "step": 15520 }, { "epoch": 0.3, "learning_rate": 4.498023536414158e-05, "loss": 0.0071, "step": 15522 }, { "epoch": 0.3, "learning_rate": 4.4979588403884357e-05, "loss": 0.0006, "step": 15524 }, { "epoch": 0.3, "learning_rate": 4.497894144362712e-05, "loss": 0.0008, "step": 15526 }, { "epoch": 0.3, "learning_rate": 4.497829448336989e-05, "loss": 0.0166, "step": 15528 }, { "epoch": 0.3, "learning_rate": 4.497764752311266e-05, "loss": 0.0074, "step": 15530 }, { "epoch": 0.3, "learning_rate": 4.4977000562855426e-05, "loss": 0.0024, "step": 15532 }, { "epoch": 0.3, "learning_rate": 4.4976353602598195e-05, "loss": 0.0002, "step": 15534 }, { "epoch": 0.3, "learning_rate": 4.4975706642340964e-05, "loss": 0.0062, "step": 15536 }, { "epoch": 0.3, "learning_rate": 4.4975059682083733e-05, "loss": 0.0037, "step": 15538 }, { "epoch": 0.3, "learning_rate": 4.4974412721826496e-05, "loss": 0.0055, "step": 15540 }, { "epoch": 0.3, "learning_rate": 4.497376576156927e-05, "loss": 0.0004, "step": 15542 }, { "epoch": 0.3, "learning_rate": 4.4973118801312034e-05, "loss": 0.017, "step": 15544 }, { "epoch": 0.3, "learning_rate": 4.497247184105481e-05, "loss": 0.0171, "step": 15546 }, { "epoch": 0.3, "learning_rate": 4.497182488079757e-05, "loss": 0.0013, "step": 15548 }, { "epoch": 0.3, "learning_rate": 4.497117792054034e-05, "loss": 0.0001, "step": 15550 }, { "epoch": 0.3, "learning_rate": 4.497053096028312e-05, "loss": 0.0134, "step": 15552 }, { "epoch": 0.3, "learning_rate": 4.496988400002588e-05, "loss": 0.0002, "step": 15554 }, { "epoch": 0.3, "learning_rate": 4.496923703976865e-05, "loss": 0.0027, "step": 15556 }, { "epoch": 0.3, "learning_rate": 4.496859007951142e-05, "loss": 0.0005, "step": 15558 }, { "epoch": 0.3, "learning_rate": 4.4967943119254187e-05, "loss": 0.0006, "step": 15560 }, { "epoch": 0.3, "learning_rate": 4.4967296158996956e-05, "loss": 0.0045, "step": 15562 }, { "epoch": 0.3, "learning_rate": 4.4966649198739725e-05, "loss": 0.0024, "step": 15564 }, { "epoch": 0.3, "learning_rate": 4.4966002238482494e-05, "loss": 0.0103, "step": 15566 }, { "epoch": 0.3, "learning_rate": 4.496535527822526e-05, "loss": 0.0039, "step": 15568 }, { "epoch": 0.3, "learning_rate": 4.496470831796803e-05, "loss": 0.0096, "step": 15570 }, { "epoch": 0.3, "learning_rate": 4.4964061357710794e-05, "loss": 0.0001, "step": 15572 }, { "epoch": 0.3, "learning_rate": 4.496341439745357e-05, "loss": 0.0017, "step": 15574 }, { "epoch": 0.3, "learning_rate": 4.496276743719633e-05, "loss": 0.0227, "step": 15576 }, { "epoch": 0.3, "learning_rate": 4.49621204769391e-05, "loss": 0.0005, "step": 15578 }, { "epoch": 0.3, "learning_rate": 4.496147351668187e-05, "loss": 0.001, "step": 15580 }, { "epoch": 0.3, "learning_rate": 4.496082655642464e-05, "loss": 0.0007, "step": 15582 }, { "epoch": 0.3, "learning_rate": 4.496017959616741e-05, "loss": 0.0129, "step": 15584 }, { "epoch": 0.3, "learning_rate": 4.495953263591018e-05, "loss": 0.0093, "step": 15586 }, { "epoch": 0.3, "learning_rate": 4.495888567565295e-05, "loss": 0.0076, "step": 15588 }, { "epoch": 0.3, "learning_rate": 4.4958238715395716e-05, "loss": 0.0007, "step": 15590 }, { "epoch": 0.3, "learning_rate": 4.4957591755138485e-05, "loss": 0.0005, "step": 15592 }, { "epoch": 0.3, "learning_rate": 4.495694479488125e-05, "loss": 0.0227, "step": 15594 }, { "epoch": 0.3, "learning_rate": 4.495629783462402e-05, "loss": 0.0301, "step": 15596 }, { "epoch": 0.3, "learning_rate": 4.495565087436679e-05, "loss": 0.0178, "step": 15598 }, { "epoch": 0.3, "learning_rate": 4.4955003914109555e-05, "loss": 0.0011, "step": 15600 }, { "epoch": 0.3, "learning_rate": 4.495435695385233e-05, "loss": 0.0054, "step": 15602 }, { "epoch": 0.3, "learning_rate": 4.495370999359509e-05, "loss": 0.0019, "step": 15604 }, { "epoch": 0.3, "learning_rate": 4.495306303333787e-05, "loss": 0.0096, "step": 15606 }, { "epoch": 0.3, "learning_rate": 4.495241607308063e-05, "loss": 0.0069, "step": 15608 }, { "epoch": 0.3, "learning_rate": 4.49517691128234e-05, "loss": 0.0018, "step": 15610 }, { "epoch": 0.3, "learning_rate": 4.495112215256617e-05, "loss": 0.0046, "step": 15612 }, { "epoch": 0.3, "learning_rate": 4.495047519230894e-05, "loss": 0.0035, "step": 15614 }, { "epoch": 0.3, "learning_rate": 4.494982823205171e-05, "loss": 0.0001, "step": 15616 }, { "epoch": 0.3, "learning_rate": 4.4949181271794476e-05, "loss": 0.0064, "step": 15618 }, { "epoch": 0.3, "learning_rate": 4.4948534311537246e-05, "loss": 0.0014, "step": 15620 }, { "epoch": 0.3, "learning_rate": 4.494788735128001e-05, "loss": 0.0007, "step": 15622 }, { "epoch": 0.3, "learning_rate": 4.4947240391022784e-05, "loss": 0.001, "step": 15624 }, { "epoch": 0.3, "learning_rate": 4.4946593430765546e-05, "loss": 0.0011, "step": 15626 }, { "epoch": 0.3, "learning_rate": 4.494594647050832e-05, "loss": 0.0082, "step": 15628 }, { "epoch": 0.3, "learning_rate": 4.494529951025109e-05, "loss": 0.0033, "step": 15630 }, { "epoch": 0.3, "learning_rate": 4.494465254999385e-05, "loss": 0.0311, "step": 15632 }, { "epoch": 0.3, "learning_rate": 4.494400558973663e-05, "loss": 0.0303, "step": 15634 }, { "epoch": 0.3, "learning_rate": 4.494335862947939e-05, "loss": 0.0028, "step": 15636 }, { "epoch": 0.3, "learning_rate": 4.494271166922216e-05, "loss": 0.0011, "step": 15638 }, { "epoch": 0.3, "learning_rate": 4.494206470896493e-05, "loss": 0.0038, "step": 15640 }, { "epoch": 0.3, "learning_rate": 4.49414177487077e-05, "loss": 0.0018, "step": 15642 }, { "epoch": 0.3, "learning_rate": 4.494077078845047e-05, "loss": 0.0007, "step": 15644 }, { "epoch": 0.3, "learning_rate": 4.494012382819324e-05, "loss": 0.0071, "step": 15646 }, { "epoch": 0.3, "learning_rate": 4.4939476867936006e-05, "loss": 0.0008, "step": 15648 }, { "epoch": 0.3, "learning_rate": 4.4938829907678775e-05, "loss": 0.0006, "step": 15650 }, { "epoch": 0.3, "learning_rate": 4.4938182947421544e-05, "loss": 0.0142, "step": 15652 }, { "epoch": 0.3, "learning_rate": 4.4937535987164306e-05, "loss": 0.001, "step": 15654 }, { "epoch": 0.3, "learning_rate": 4.493688902690708e-05, "loss": 0.003, "step": 15656 }, { "epoch": 0.3, "learning_rate": 4.4936242066649845e-05, "loss": 0.0007, "step": 15658 }, { "epoch": 0.3, "learning_rate": 4.4935595106392614e-05, "loss": 0.0001, "step": 15660 }, { "epoch": 0.3, "learning_rate": 4.493494814613538e-05, "loss": 0.0049, "step": 15662 }, { "epoch": 0.3, "learning_rate": 4.493430118587815e-05, "loss": 0.017, "step": 15664 }, { "epoch": 0.3, "learning_rate": 4.493365422562093e-05, "loss": 0.0002, "step": 15666 }, { "epoch": 0.3, "learning_rate": 4.493300726536369e-05, "loss": 0.058, "step": 15668 }, { "epoch": 0.3, "learning_rate": 4.493236030510646e-05, "loss": 0.0385, "step": 15670 }, { "epoch": 0.3, "learning_rate": 4.493171334484923e-05, "loss": 0.004, "step": 15672 }, { "epoch": 0.3, "learning_rate": 4.4931066384592e-05, "loss": 0.0134, "step": 15674 }, { "epoch": 0.3, "learning_rate": 4.4930419424334766e-05, "loss": 0.0027, "step": 15676 }, { "epoch": 0.3, "learning_rate": 4.4929772464077535e-05, "loss": 0.0004, "step": 15678 }, { "epoch": 0.3, "learning_rate": 4.4929125503820305e-05, "loss": 0.0164, "step": 15680 }, { "epoch": 0.3, "learning_rate": 4.492847854356307e-05, "loss": 0.0018, "step": 15682 }, { "epoch": 0.3, "learning_rate": 4.492783158330584e-05, "loss": 0.0007, "step": 15684 }, { "epoch": 0.3, "learning_rate": 4.4927184623048605e-05, "loss": 0.0006, "step": 15686 }, { "epoch": 0.3, "learning_rate": 4.492653766279138e-05, "loss": 0.0113, "step": 15688 }, { "epoch": 0.3, "learning_rate": 4.492589070253414e-05, "loss": 0.0023, "step": 15690 }, { "epoch": 0.3, "learning_rate": 4.492524374227691e-05, "loss": 0.0083, "step": 15692 }, { "epoch": 0.3, "learning_rate": 4.492459678201968e-05, "loss": 0.0063, "step": 15694 }, { "epoch": 0.3, "learning_rate": 4.492394982176245e-05, "loss": 0.0124, "step": 15696 }, { "epoch": 0.3, "learning_rate": 4.492330286150522e-05, "loss": 0.0402, "step": 15698 }, { "epoch": 0.3, "learning_rate": 4.492265590124799e-05, "loss": 0.0067, "step": 15700 }, { "epoch": 0.3, "learning_rate": 4.492200894099076e-05, "loss": 0.0002, "step": 15702 }, { "epoch": 0.3, "learning_rate": 4.492136198073353e-05, "loss": 0.0008, "step": 15704 }, { "epoch": 0.3, "learning_rate": 4.4920715020476296e-05, "loss": 0.0017, "step": 15706 }, { "epoch": 0.3, "learning_rate": 4.4920068060219065e-05, "loss": 0.0133, "step": 15708 }, { "epoch": 0.3, "learning_rate": 4.4919421099961834e-05, "loss": 0.0084, "step": 15710 }, { "epoch": 0.3, "learning_rate": 4.49187741397046e-05, "loss": 0.01, "step": 15712 }, { "epoch": 0.3, "learning_rate": 4.4918127179447365e-05, "loss": 0.0017, "step": 15714 }, { "epoch": 0.31, "learning_rate": 4.491748021919014e-05, "loss": 0.011, "step": 15716 }, { "epoch": 0.31, "learning_rate": 4.4916833258932904e-05, "loss": 0.0006, "step": 15718 }, { "epoch": 0.31, "learning_rate": 4.491618629867567e-05, "loss": 0.0135, "step": 15720 }, { "epoch": 0.31, "learning_rate": 4.491553933841844e-05, "loss": 0.0002, "step": 15722 }, { "epoch": 0.31, "learning_rate": 4.491489237816121e-05, "loss": 0.0008, "step": 15724 }, { "epoch": 0.31, "learning_rate": 4.491424541790398e-05, "loss": 0.0009, "step": 15726 }, { "epoch": 0.31, "learning_rate": 4.491359845764675e-05, "loss": 0.0017, "step": 15728 }, { "epoch": 0.31, "learning_rate": 4.491295149738952e-05, "loss": 0.0001, "step": 15730 }, { "epoch": 0.31, "learning_rate": 4.491230453713229e-05, "loss": 0.0123, "step": 15732 }, { "epoch": 0.31, "learning_rate": 4.4911657576875056e-05, "loss": 0.0002, "step": 15734 }, { "epoch": 0.31, "learning_rate": 4.491101061661782e-05, "loss": 0.0005, "step": 15736 }, { "epoch": 0.31, "learning_rate": 4.4910363656360594e-05, "loss": 0.0043, "step": 15738 }, { "epoch": 0.31, "learning_rate": 4.490971669610336e-05, "loss": 0.0047, "step": 15740 }, { "epoch": 0.31, "learning_rate": 4.4909069735846126e-05, "loss": 0.0476, "step": 15742 }, { "epoch": 0.31, "learning_rate": 4.49084227755889e-05, "loss": 0.0128, "step": 15744 }, { "epoch": 0.31, "learning_rate": 4.4907775815331664e-05, "loss": 0.0017, "step": 15746 }, { "epoch": 0.31, "learning_rate": 4.490712885507444e-05, "loss": 0.0044, "step": 15748 }, { "epoch": 0.31, "learning_rate": 4.49064818948172e-05, "loss": 0.0086, "step": 15750 }, { "epoch": 0.31, "learning_rate": 4.490583493455997e-05, "loss": 0.0137, "step": 15752 }, { "epoch": 0.31, "learning_rate": 4.490518797430274e-05, "loss": 0.008, "step": 15754 }, { "epoch": 0.31, "learning_rate": 4.490454101404551e-05, "loss": 0.0015, "step": 15756 }, { "epoch": 0.31, "learning_rate": 4.490389405378828e-05, "loss": 0.0207, "step": 15758 }, { "epoch": 0.31, "learning_rate": 4.490324709353105e-05, "loss": 0.0003, "step": 15760 }, { "epoch": 0.31, "learning_rate": 4.490260013327382e-05, "loss": 0.0004, "step": 15762 }, { "epoch": 0.31, "learning_rate": 4.490195317301658e-05, "loss": 0.0008, "step": 15764 }, { "epoch": 0.31, "learning_rate": 4.4901306212759355e-05, "loss": 0.0032, "step": 15766 }, { "epoch": 0.31, "learning_rate": 4.490065925250212e-05, "loss": 0.0006, "step": 15768 }, { "epoch": 0.31, "learning_rate": 4.490001229224489e-05, "loss": 0.0062, "step": 15770 }, { "epoch": 0.31, "learning_rate": 4.4899365331987655e-05, "loss": 0.0003, "step": 15772 }, { "epoch": 0.31, "learning_rate": 4.4898718371730424e-05, "loss": 0.0004, "step": 15774 }, { "epoch": 0.31, "learning_rate": 4.48980714114732e-05, "loss": 0.0151, "step": 15776 }, { "epoch": 0.31, "learning_rate": 4.489742445121596e-05, "loss": 0.0005, "step": 15778 }, { "epoch": 0.31, "learning_rate": 4.489677749095873e-05, "loss": 0.0009, "step": 15780 }, { "epoch": 0.31, "learning_rate": 4.48961305307015e-05, "loss": 0.0013, "step": 15782 }, { "epoch": 0.31, "learning_rate": 4.489548357044427e-05, "loss": 0.0174, "step": 15784 }, { "epoch": 0.31, "learning_rate": 4.489483661018704e-05, "loss": 0.0012, "step": 15786 }, { "epoch": 0.31, "learning_rate": 4.489418964992981e-05, "loss": 0.0003, "step": 15788 }, { "epoch": 0.31, "learning_rate": 4.489354268967258e-05, "loss": 0.0176, "step": 15790 }, { "epoch": 0.31, "learning_rate": 4.4892895729415346e-05, "loss": 0.0005, "step": 15792 }, { "epoch": 0.31, "learning_rate": 4.4892248769158115e-05, "loss": 0.0191, "step": 15794 }, { "epoch": 0.31, "learning_rate": 4.489160180890088e-05, "loss": 0.0127, "step": 15796 }, { "epoch": 0.31, "learning_rate": 4.4890954848643653e-05, "loss": 0.0016, "step": 15798 }, { "epoch": 0.31, "learning_rate": 4.4890307888386416e-05, "loss": 0.0081, "step": 15800 }, { "epoch": 0.31, "learning_rate": 4.4889660928129185e-05, "loss": 0.0017, "step": 15802 }, { "epoch": 0.31, "learning_rate": 4.4889013967871954e-05, "loss": 0.0004, "step": 15804 }, { "epoch": 0.31, "learning_rate": 4.488836700761472e-05, "loss": 0.0003, "step": 15806 }, { "epoch": 0.31, "learning_rate": 4.48877200473575e-05, "loss": 0.001, "step": 15808 }, { "epoch": 0.31, "learning_rate": 4.488707308710026e-05, "loss": 0.0133, "step": 15810 }, { "epoch": 0.31, "learning_rate": 4.488642612684303e-05, "loss": 0.0142, "step": 15812 }, { "epoch": 0.31, "learning_rate": 4.48857791665858e-05, "loss": 0.0002, "step": 15814 }, { "epoch": 0.31, "learning_rate": 4.488513220632857e-05, "loss": 0.0146, "step": 15816 }, { "epoch": 0.31, "learning_rate": 4.488448524607133e-05, "loss": 0.0173, "step": 15818 }, { "epoch": 0.31, "learning_rate": 4.4883838285814107e-05, "loss": 0.0092, "step": 15820 }, { "epoch": 0.31, "learning_rate": 4.4883191325556876e-05, "loss": 0.0001, "step": 15822 }, { "epoch": 0.31, "learning_rate": 4.488254436529964e-05, "loss": 0.0117, "step": 15824 }, { "epoch": 0.31, "learning_rate": 4.4881897405042414e-05, "loss": 0.0171, "step": 15826 }, { "epoch": 0.31, "learning_rate": 4.4881250444785176e-05, "loss": 0.0002, "step": 15828 }, { "epoch": 0.31, "learning_rate": 4.488060348452795e-05, "loss": 0.0044, "step": 15830 }, { "epoch": 0.31, "learning_rate": 4.4879956524270714e-05, "loss": 0.0127, "step": 15832 }, { "epoch": 0.31, "learning_rate": 4.487930956401348e-05, "loss": 0.0325, "step": 15834 }, { "epoch": 0.31, "learning_rate": 4.487866260375625e-05, "loss": 0.0004, "step": 15836 }, { "epoch": 0.31, "learning_rate": 4.487801564349902e-05, "loss": 0.0006, "step": 15838 }, { "epoch": 0.31, "learning_rate": 4.487736868324179e-05, "loss": 0.0024, "step": 15840 }, { "epoch": 0.31, "learning_rate": 4.487672172298456e-05, "loss": 0.0049, "step": 15842 }, { "epoch": 0.31, "learning_rate": 4.487607476272733e-05, "loss": 0.0034, "step": 15844 }, { "epoch": 0.31, "learning_rate": 4.48754278024701e-05, "loss": 0.0004, "step": 15846 }, { "epoch": 0.31, "learning_rate": 4.487478084221287e-05, "loss": 0.0003, "step": 15848 }, { "epoch": 0.31, "learning_rate": 4.487413388195563e-05, "loss": 0.0008, "step": 15850 }, { "epoch": 0.31, "learning_rate": 4.4873486921698405e-05, "loss": 0.0097, "step": 15852 }, { "epoch": 0.31, "learning_rate": 4.4872839961441174e-05, "loss": 0.0108, "step": 15854 }, { "epoch": 0.31, "learning_rate": 4.4872193001183937e-05, "loss": 0.0006, "step": 15856 }, { "epoch": 0.31, "learning_rate": 4.487154604092671e-05, "loss": 0.0423, "step": 15858 }, { "epoch": 0.31, "learning_rate": 4.4870899080669475e-05, "loss": 0.0001, "step": 15860 }, { "epoch": 0.31, "learning_rate": 4.4870252120412244e-05, "loss": 0.0004, "step": 15862 }, { "epoch": 0.31, "learning_rate": 4.486960516015501e-05, "loss": 0.0004, "step": 15864 }, { "epoch": 0.31, "learning_rate": 4.486895819989778e-05, "loss": 0.0075, "step": 15866 }, { "epoch": 0.31, "learning_rate": 4.486831123964055e-05, "loss": 0.0003, "step": 15868 }, { "epoch": 0.31, "learning_rate": 4.486766427938332e-05, "loss": 0.0161, "step": 15870 }, { "epoch": 0.31, "learning_rate": 4.486701731912609e-05, "loss": 0.0056, "step": 15872 }, { "epoch": 0.31, "learning_rate": 4.486637035886886e-05, "loss": 0.0213, "step": 15874 }, { "epoch": 0.31, "learning_rate": 4.486572339861163e-05, "loss": 0.0235, "step": 15876 }, { "epoch": 0.31, "learning_rate": 4.486507643835439e-05, "loss": 0.0003, "step": 15878 }, { "epoch": 0.31, "learning_rate": 4.4864429478097166e-05, "loss": 0.0005, "step": 15880 }, { "epoch": 0.31, "learning_rate": 4.486378251783993e-05, "loss": 0.0072, "step": 15882 }, { "epoch": 0.31, "learning_rate": 4.48631355575827e-05, "loss": 0.0006, "step": 15884 }, { "epoch": 0.31, "learning_rate": 4.4862488597325466e-05, "loss": 0.0011, "step": 15886 }, { "epoch": 0.31, "learning_rate": 4.4861841637068235e-05, "loss": 0.0073, "step": 15888 }, { "epoch": 0.31, "learning_rate": 4.486119467681101e-05, "loss": 0.0003, "step": 15890 }, { "epoch": 0.31, "learning_rate": 4.486054771655377e-05, "loss": 0.0014, "step": 15892 }, { "epoch": 0.31, "learning_rate": 4.485990075629654e-05, "loss": 0.0331, "step": 15894 }, { "epoch": 0.31, "learning_rate": 4.485925379603931e-05, "loss": 0.0003, "step": 15896 }, { "epoch": 0.31, "learning_rate": 4.485860683578208e-05, "loss": 0.0002, "step": 15898 }, { "epoch": 0.31, "learning_rate": 4.485795987552485e-05, "loss": 0.0047, "step": 15900 }, { "epoch": 0.31, "learning_rate": 4.485731291526762e-05, "loss": 0.0305, "step": 15902 }, { "epoch": 0.31, "learning_rate": 4.485666595501039e-05, "loss": 0.0007, "step": 15904 }, { "epoch": 0.31, "learning_rate": 4.485601899475315e-05, "loss": 0.0157, "step": 15906 }, { "epoch": 0.31, "learning_rate": 4.4855372034495926e-05, "loss": 0.0006, "step": 15908 }, { "epoch": 0.31, "learning_rate": 4.485472507423869e-05, "loss": 0.0009, "step": 15910 }, { "epoch": 0.31, "learning_rate": 4.4854078113981464e-05, "loss": 0.0009, "step": 15912 }, { "epoch": 0.31, "learning_rate": 4.4853431153724226e-05, "loss": 0.0234, "step": 15914 }, { "epoch": 0.31, "learning_rate": 4.4852784193466995e-05, "loss": 0.0019, "step": 15916 }, { "epoch": 0.31, "learning_rate": 4.4852137233209765e-05, "loss": 0.0047, "step": 15918 }, { "epoch": 0.31, "learning_rate": 4.4851490272952534e-05, "loss": 0.0207, "step": 15920 }, { "epoch": 0.31, "learning_rate": 4.48508433126953e-05, "loss": 0.0058, "step": 15922 }, { "epoch": 0.31, "learning_rate": 4.485019635243807e-05, "loss": 0.0001, "step": 15924 }, { "epoch": 0.31, "learning_rate": 4.484954939218084e-05, "loss": 0.0015, "step": 15926 }, { "epoch": 0.31, "learning_rate": 4.484890243192361e-05, "loss": 0.0049, "step": 15928 }, { "epoch": 0.31, "learning_rate": 4.484825547166638e-05, "loss": 0.0059, "step": 15930 }, { "epoch": 0.31, "learning_rate": 4.484760851140915e-05, "loss": 0.0149, "step": 15932 }, { "epoch": 0.31, "learning_rate": 4.484696155115192e-05, "loss": 0.0003, "step": 15934 }, { "epoch": 0.31, "learning_rate": 4.4846314590894686e-05, "loss": 0.0016, "step": 15936 }, { "epoch": 0.31, "learning_rate": 4.484566763063745e-05, "loss": 0.0003, "step": 15938 }, { "epoch": 0.31, "learning_rate": 4.4845020670380225e-05, "loss": 0.0035, "step": 15940 }, { "epoch": 0.31, "learning_rate": 4.484437371012299e-05, "loss": 0.0015, "step": 15942 }, { "epoch": 0.31, "learning_rate": 4.4843726749865756e-05, "loss": 0.0015, "step": 15944 }, { "epoch": 0.31, "learning_rate": 4.4843079789608525e-05, "loss": 0.0144, "step": 15946 }, { "epoch": 0.31, "learning_rate": 4.4842432829351294e-05, "loss": 0.0193, "step": 15948 }, { "epoch": 0.31, "learning_rate": 4.484178586909406e-05, "loss": 0.0046, "step": 15950 }, { "epoch": 0.31, "learning_rate": 4.484113890883683e-05, "loss": 0.0013, "step": 15952 }, { "epoch": 0.31, "learning_rate": 4.48404919485796e-05, "loss": 0.0013, "step": 15954 }, { "epoch": 0.31, "learning_rate": 4.483984498832237e-05, "loss": 0.0002, "step": 15956 }, { "epoch": 0.31, "learning_rate": 4.483919802806514e-05, "loss": 0.0138, "step": 15958 }, { "epoch": 0.31, "learning_rate": 4.48385510678079e-05, "loss": 0.0059, "step": 15960 }, { "epoch": 0.31, "learning_rate": 4.483790410755068e-05, "loss": 0.0032, "step": 15962 }, { "epoch": 0.31, "learning_rate": 4.483725714729344e-05, "loss": 0.0364, "step": 15964 }, { "epoch": 0.31, "learning_rate": 4.483661018703621e-05, "loss": 0.0017, "step": 15966 }, { "epoch": 0.31, "learning_rate": 4.4835963226778985e-05, "loss": 0.0074, "step": 15968 }, { "epoch": 0.31, "learning_rate": 4.483531626652175e-05, "loss": 0.0001, "step": 15970 }, { "epoch": 0.31, "learning_rate": 4.483466930626452e-05, "loss": 0.0003, "step": 15972 }, { "epoch": 0.31, "learning_rate": 4.4834022346007285e-05, "loss": 0.0001, "step": 15974 }, { "epoch": 0.31, "learning_rate": 4.4833375385750054e-05, "loss": 0.001, "step": 15976 }, { "epoch": 0.31, "learning_rate": 4.4832728425492824e-05, "loss": 0.0055, "step": 15978 }, { "epoch": 0.31, "learning_rate": 4.483208146523559e-05, "loss": 0.0064, "step": 15980 }, { "epoch": 0.31, "learning_rate": 4.483143450497836e-05, "loss": 0.0102, "step": 15982 }, { "epoch": 0.31, "learning_rate": 4.483078754472113e-05, "loss": 0.0005, "step": 15984 }, { "epoch": 0.31, "learning_rate": 4.48301405844639e-05, "loss": 0.0009, "step": 15986 }, { "epoch": 0.31, "learning_rate": 4.482949362420667e-05, "loss": 0.0136, "step": 15988 }, { "epoch": 0.31, "learning_rate": 4.482884666394944e-05, "loss": 0.0061, "step": 15990 }, { "epoch": 0.31, "learning_rate": 4.48281997036922e-05, "loss": 0.0002, "step": 15992 }, { "epoch": 0.31, "learning_rate": 4.4827552743434976e-05, "loss": 0.0256, "step": 15994 }, { "epoch": 0.31, "learning_rate": 4.482690578317774e-05, "loss": 0.0013, "step": 15996 }, { "epoch": 0.31, "learning_rate": 4.482625882292051e-05, "loss": 0.0133, "step": 15998 }, { "epoch": 0.31, "learning_rate": 4.4825611862663283e-05, "loss": 0.0005, "step": 16000 }, { "epoch": 0.31, "learning_rate": 4.4824964902406046e-05, "loss": 0.0023, "step": 16002 }, { "epoch": 0.31, "learning_rate": 4.4824317942148815e-05, "loss": 0.0041, "step": 16004 }, { "epoch": 0.31, "learning_rate": 4.4823670981891584e-05, "loss": 0.003, "step": 16006 }, { "epoch": 0.31, "learning_rate": 4.482302402163435e-05, "loss": 0.001, "step": 16008 }, { "epoch": 0.31, "learning_rate": 4.482237706137712e-05, "loss": 0.0006, "step": 16010 }, { "epoch": 0.31, "learning_rate": 4.482173010111989e-05, "loss": 0.0077, "step": 16012 }, { "epoch": 0.31, "learning_rate": 4.482108314086266e-05, "loss": 0.0021, "step": 16014 }, { "epoch": 0.31, "learning_rate": 4.482043618060543e-05, "loss": 0.0006, "step": 16016 }, { "epoch": 0.31, "learning_rate": 4.48197892203482e-05, "loss": 0.0053, "step": 16018 }, { "epoch": 0.31, "learning_rate": 4.481914226009096e-05, "loss": 0.0146, "step": 16020 }, { "epoch": 0.31, "learning_rate": 4.481849529983374e-05, "loss": 0.0132, "step": 16022 }, { "epoch": 0.31, "learning_rate": 4.48178483395765e-05, "loss": 0.0097, "step": 16024 }, { "epoch": 0.31, "learning_rate": 4.481720137931927e-05, "loss": 0.0357, "step": 16026 }, { "epoch": 0.31, "learning_rate": 4.481655441906204e-05, "loss": 0.0053, "step": 16028 }, { "epoch": 0.31, "learning_rate": 4.4815907458804806e-05, "loss": 0.0025, "step": 16030 }, { "epoch": 0.31, "learning_rate": 4.481526049854758e-05, "loss": 0.0099, "step": 16032 }, { "epoch": 0.31, "learning_rate": 4.4814613538290344e-05, "loss": 0.0167, "step": 16034 }, { "epoch": 0.31, "learning_rate": 4.4813966578033113e-05, "loss": 0.0016, "step": 16036 }, { "epoch": 0.31, "learning_rate": 4.481331961777588e-05, "loss": 0.0135, "step": 16038 }, { "epoch": 0.31, "learning_rate": 4.481267265751865e-05, "loss": 0.1584, "step": 16040 }, { "epoch": 0.31, "learning_rate": 4.4812025697261414e-05, "loss": 0.0019, "step": 16042 }, { "epoch": 0.31, "learning_rate": 4.481137873700419e-05, "loss": 0.0039, "step": 16044 }, { "epoch": 0.31, "learning_rate": 4.481073177674696e-05, "loss": 0.0061, "step": 16046 }, { "epoch": 0.31, "learning_rate": 4.481008481648972e-05, "loss": 0.0044, "step": 16048 }, { "epoch": 0.31, "learning_rate": 4.48094378562325e-05, "loss": 0.0012, "step": 16050 }, { "epoch": 0.31, "learning_rate": 4.480879089597526e-05, "loss": 0.0149, "step": 16052 }, { "epoch": 0.31, "learning_rate": 4.4808143935718035e-05, "loss": 0.0006, "step": 16054 }, { "epoch": 0.31, "learning_rate": 4.48074969754608e-05, "loss": 0.0006, "step": 16056 }, { "epoch": 0.31, "learning_rate": 4.4806850015203567e-05, "loss": 0.0208, "step": 16058 }, { "epoch": 0.31, "learning_rate": 4.4806203054946336e-05, "loss": 0.1147, "step": 16060 }, { "epoch": 0.31, "learning_rate": 4.4805556094689105e-05, "loss": 0.0001, "step": 16062 }, { "epoch": 0.31, "learning_rate": 4.4804909134431874e-05, "loss": 0.005, "step": 16064 }, { "epoch": 0.31, "learning_rate": 4.480426217417464e-05, "loss": 0.0012, "step": 16066 }, { "epoch": 0.31, "learning_rate": 4.480361521391741e-05, "loss": 0.0011, "step": 16068 }, { "epoch": 0.31, "learning_rate": 4.480296825366018e-05, "loss": 0.031, "step": 16070 }, { "epoch": 0.31, "learning_rate": 4.480232129340295e-05, "loss": 0.0038, "step": 16072 }, { "epoch": 0.31, "learning_rate": 4.480167433314571e-05, "loss": 0.0047, "step": 16074 }, { "epoch": 0.31, "learning_rate": 4.480102737288849e-05, "loss": 0.0071, "step": 16076 }, { "epoch": 0.31, "learning_rate": 4.480038041263126e-05, "loss": 0.0085, "step": 16078 }, { "epoch": 0.31, "learning_rate": 4.479973345237402e-05, "loss": 0.0026, "step": 16080 }, { "epoch": 0.31, "learning_rate": 4.4799086492116796e-05, "loss": 0.0082, "step": 16082 }, { "epoch": 0.31, "learning_rate": 4.479843953185956e-05, "loss": 0.0006, "step": 16084 }, { "epoch": 0.31, "learning_rate": 4.479779257160233e-05, "loss": 0.0002, "step": 16086 }, { "epoch": 0.31, "learning_rate": 4.4797145611345096e-05, "loss": 0.0004, "step": 16088 }, { "epoch": 0.31, "learning_rate": 4.4796498651087865e-05, "loss": 0.0007, "step": 16090 }, { "epoch": 0.31, "learning_rate": 4.4795851690830634e-05, "loss": 0.0001, "step": 16092 }, { "epoch": 0.31, "learning_rate": 4.47952047305734e-05, "loss": 0.0008, "step": 16094 }, { "epoch": 0.31, "learning_rate": 4.479455777031617e-05, "loss": 0.0116, "step": 16096 }, { "epoch": 0.31, "learning_rate": 4.479391081005894e-05, "loss": 0.0018, "step": 16098 }, { "epoch": 0.31, "learning_rate": 4.479326384980171e-05, "loss": 0.0026, "step": 16100 }, { "epoch": 0.31, "learning_rate": 4.479261688954447e-05, "loss": 0.0002, "step": 16102 }, { "epoch": 0.31, "learning_rate": 4.479196992928725e-05, "loss": 0.0002, "step": 16104 }, { "epoch": 0.31, "learning_rate": 4.479132296903001e-05, "loss": 0.015, "step": 16106 }, { "epoch": 0.31, "learning_rate": 4.479067600877278e-05, "loss": 0.0028, "step": 16108 }, { "epoch": 0.31, "learning_rate": 4.4790029048515556e-05, "loss": 0.0168, "step": 16110 }, { "epoch": 0.31, "learning_rate": 4.478938208825832e-05, "loss": 0.026, "step": 16112 }, { "epoch": 0.31, "learning_rate": 4.4788735128001094e-05, "loss": 0.0011, "step": 16114 }, { "epoch": 0.31, "learning_rate": 4.4788088167743856e-05, "loss": 0.0128, "step": 16116 }, { "epoch": 0.31, "learning_rate": 4.4787441207486626e-05, "loss": 0.0113, "step": 16118 }, { "epoch": 0.31, "learning_rate": 4.4786794247229395e-05, "loss": 0.0007, "step": 16120 }, { "epoch": 0.31, "learning_rate": 4.4786147286972164e-05, "loss": 0.0052, "step": 16122 }, { "epoch": 0.31, "learning_rate": 4.478550032671493e-05, "loss": 0.0004, "step": 16124 }, { "epoch": 0.31, "learning_rate": 4.47848533664577e-05, "loss": 0.0009, "step": 16126 }, { "epoch": 0.31, "learning_rate": 4.478420640620047e-05, "loss": 0.0202, "step": 16128 }, { "epoch": 0.31, "learning_rate": 4.478355944594324e-05, "loss": 0.0067, "step": 16130 }, { "epoch": 0.31, "learning_rate": 4.478291248568601e-05, "loss": 0.0055, "step": 16132 }, { "epoch": 0.31, "learning_rate": 4.478226552542877e-05, "loss": 0.0071, "step": 16134 }, { "epoch": 0.31, "learning_rate": 4.478161856517155e-05, "loss": 0.0011, "step": 16136 }, { "epoch": 0.31, "learning_rate": 4.478097160491431e-05, "loss": 0.002, "step": 16138 }, { "epoch": 0.31, "learning_rate": 4.478032464465708e-05, "loss": 0.009, "step": 16140 }, { "epoch": 0.31, "learning_rate": 4.477967768439985e-05, "loss": 0.0072, "step": 16142 }, { "epoch": 0.31, "learning_rate": 4.477903072414262e-05, "loss": 0.0047, "step": 16144 }, { "epoch": 0.31, "learning_rate": 4.4778383763885386e-05, "loss": 0.0002, "step": 16146 }, { "epoch": 0.31, "learning_rate": 4.4777736803628155e-05, "loss": 0.0039, "step": 16148 }, { "epoch": 0.31, "learning_rate": 4.4777089843370924e-05, "loss": 0.0099, "step": 16150 }, { "epoch": 0.31, "learning_rate": 4.477644288311369e-05, "loss": 0.0121, "step": 16152 }, { "epoch": 0.31, "learning_rate": 4.477579592285646e-05, "loss": 0.0189, "step": 16154 }, { "epoch": 0.31, "learning_rate": 4.477514896259923e-05, "loss": 0.0001, "step": 16156 }, { "epoch": 0.31, "learning_rate": 4.4774502002342e-05, "loss": 0.0037, "step": 16158 }, { "epoch": 0.31, "learning_rate": 4.477385504208477e-05, "loss": 0.0094, "step": 16160 }, { "epoch": 0.31, "learning_rate": 4.477320808182753e-05, "loss": 0.0018, "step": 16162 }, { "epoch": 0.31, "learning_rate": 4.477256112157031e-05, "loss": 0.0037, "step": 16164 }, { "epoch": 0.31, "learning_rate": 4.477191416131307e-05, "loss": 0.0059, "step": 16166 }, { "epoch": 0.31, "learning_rate": 4.477126720105584e-05, "loss": 0.0004, "step": 16168 }, { "epoch": 0.31, "learning_rate": 4.477062024079861e-05, "loss": 0.0004, "step": 16170 }, { "epoch": 0.31, "learning_rate": 4.476997328054138e-05, "loss": 0.0002, "step": 16172 }, { "epoch": 0.31, "learning_rate": 4.4769326320284146e-05, "loss": 0.0007, "step": 16174 }, { "epoch": 0.31, "learning_rate": 4.4768679360026915e-05, "loss": 0.0143, "step": 16176 }, { "epoch": 0.31, "learning_rate": 4.4768032399769685e-05, "loss": 0.0002, "step": 16178 }, { "epoch": 0.31, "learning_rate": 4.4767385439512454e-05, "loss": 0.0105, "step": 16180 }, { "epoch": 0.31, "learning_rate": 4.476673847925522e-05, "loss": 0.0016, "step": 16182 }, { "epoch": 0.31, "learning_rate": 4.4766091518997985e-05, "loss": 0.0003, "step": 16184 }, { "epoch": 0.31, "learning_rate": 4.476544455874076e-05, "loss": 0.0055, "step": 16186 }, { "epoch": 0.31, "learning_rate": 4.476479759848353e-05, "loss": 0.0027, "step": 16188 }, { "epoch": 0.31, "learning_rate": 4.476415063822629e-05, "loss": 0.0001, "step": 16190 }, { "epoch": 0.31, "learning_rate": 4.476350367796907e-05, "loss": 0.0002, "step": 16192 }, { "epoch": 0.31, "learning_rate": 4.476285671771183e-05, "loss": 0.0014, "step": 16194 }, { "epoch": 0.31, "learning_rate": 4.4762209757454606e-05, "loss": 0.0002, "step": 16196 }, { "epoch": 0.31, "learning_rate": 4.476156279719737e-05, "loss": 0.0044, "step": 16198 }, { "epoch": 0.31, "learning_rate": 4.476091583694014e-05, "loss": 0.0091, "step": 16200 }, { "epoch": 0.31, "learning_rate": 4.476026887668291e-05, "loss": 0.0037, "step": 16202 }, { "epoch": 0.31, "learning_rate": 4.4759621916425676e-05, "loss": 0.0003, "step": 16204 }, { "epoch": 0.31, "learning_rate": 4.4758974956168445e-05, "loss": 0.0009, "step": 16206 }, { "epoch": 0.31, "learning_rate": 4.4758327995911214e-05, "loss": 0.0002, "step": 16208 }, { "epoch": 0.31, "learning_rate": 4.475768103565398e-05, "loss": 0.0112, "step": 16210 }, { "epoch": 0.31, "learning_rate": 4.475703407539675e-05, "loss": 0.0085, "step": 16212 }, { "epoch": 0.31, "learning_rate": 4.475638711513952e-05, "loss": 0.0002, "step": 16214 }, { "epoch": 0.31, "learning_rate": 4.4755740154882284e-05, "loss": 0.0059, "step": 16216 }, { "epoch": 0.31, "learning_rate": 4.475509319462506e-05, "loss": 0.0002, "step": 16218 }, { "epoch": 0.31, "learning_rate": 4.475444623436782e-05, "loss": 0.0002, "step": 16220 }, { "epoch": 0.31, "learning_rate": 4.475379927411059e-05, "loss": 0.0008, "step": 16222 }, { "epoch": 0.31, "learning_rate": 4.475315231385337e-05, "loss": 0.0006, "step": 16224 }, { "epoch": 0.31, "learning_rate": 4.475250535359613e-05, "loss": 0.0031, "step": 16226 }, { "epoch": 0.31, "learning_rate": 4.47518583933389e-05, "loss": 0.0004, "step": 16228 }, { "epoch": 0.32, "learning_rate": 4.475121143308167e-05, "loss": 0.0002, "step": 16230 }, { "epoch": 0.32, "learning_rate": 4.4750564472824436e-05, "loss": 0.0005, "step": 16232 }, { "epoch": 0.32, "learning_rate": 4.4749917512567205e-05, "loss": 0.0044, "step": 16234 }, { "epoch": 0.32, "learning_rate": 4.4749270552309974e-05, "loss": 0.0009, "step": 16236 }, { "epoch": 0.32, "learning_rate": 4.4748623592052744e-05, "loss": 0.0069, "step": 16238 }, { "epoch": 0.32, "learning_rate": 4.474797663179551e-05, "loss": 0.0006, "step": 16240 }, { "epoch": 0.32, "learning_rate": 4.474732967153828e-05, "loss": 0.0191, "step": 16242 }, { "epoch": 0.32, "learning_rate": 4.4746682711281044e-05, "loss": 0.0057, "step": 16244 }, { "epoch": 0.32, "learning_rate": 4.474603575102382e-05, "loss": 0.008, "step": 16246 }, { "epoch": 0.32, "learning_rate": 4.474538879076658e-05, "loss": 0.0037, "step": 16248 }, { "epoch": 0.32, "learning_rate": 4.474474183050935e-05, "loss": 0.0003, "step": 16250 }, { "epoch": 0.32, "learning_rate": 4.474409487025212e-05, "loss": 0.0083, "step": 16252 }, { "epoch": 0.32, "learning_rate": 4.474344790999489e-05, "loss": 0.0005, "step": 16254 }, { "epoch": 0.32, "learning_rate": 4.4742800949737665e-05, "loss": 0.0005, "step": 16256 }, { "epoch": 0.32, "learning_rate": 4.474215398948043e-05, "loss": 0.0002, "step": 16258 }, { "epoch": 0.32, "learning_rate": 4.47415070292232e-05, "loss": 0.0238, "step": 16260 }, { "epoch": 0.32, "learning_rate": 4.4740860068965966e-05, "loss": 0.0005, "step": 16262 }, { "epoch": 0.32, "learning_rate": 4.4740213108708735e-05, "loss": 0.0045, "step": 16264 }, { "epoch": 0.32, "learning_rate": 4.47395661484515e-05, "loss": 0.002, "step": 16266 }, { "epoch": 0.32, "learning_rate": 4.473891918819427e-05, "loss": 0.0179, "step": 16268 }, { "epoch": 0.32, "learning_rate": 4.473827222793704e-05, "loss": 0.0076, "step": 16270 }, { "epoch": 0.32, "learning_rate": 4.473762526767981e-05, "loss": 0.0019, "step": 16272 }, { "epoch": 0.32, "learning_rate": 4.473697830742258e-05, "loss": 0.0363, "step": 16274 }, { "epoch": 0.32, "learning_rate": 4.473633134716534e-05, "loss": 0.0043, "step": 16276 }, { "epoch": 0.32, "learning_rate": 4.473568438690812e-05, "loss": 0.0009, "step": 16278 }, { "epoch": 0.32, "learning_rate": 4.473503742665088e-05, "loss": 0.0012, "step": 16280 }, { "epoch": 0.32, "learning_rate": 4.473439046639365e-05, "loss": 0.002, "step": 16282 }, { "epoch": 0.32, "learning_rate": 4.473374350613642e-05, "loss": 0.0208, "step": 16284 }, { "epoch": 0.32, "learning_rate": 4.473309654587919e-05, "loss": 0.0002, "step": 16286 }, { "epoch": 0.32, "learning_rate": 4.473244958562196e-05, "loss": 0.0002, "step": 16288 }, { "epoch": 0.32, "learning_rate": 4.4731802625364726e-05, "loss": 0.0062, "step": 16290 }, { "epoch": 0.32, "learning_rate": 4.4731155665107495e-05, "loss": 0.0039, "step": 16292 }, { "epoch": 0.32, "learning_rate": 4.4730508704850264e-05, "loss": 0.0061, "step": 16294 }, { "epoch": 0.32, "learning_rate": 4.4729861744593033e-05, "loss": 0.0018, "step": 16296 }, { "epoch": 0.32, "learning_rate": 4.4729214784335796e-05, "loss": 0.0093, "step": 16298 }, { "epoch": 0.32, "learning_rate": 4.472856782407857e-05, "loss": 0.0117, "step": 16300 }, { "epoch": 0.32, "learning_rate": 4.472792086382134e-05, "loss": 0.036, "step": 16302 }, { "epoch": 0.32, "learning_rate": 4.47272739035641e-05, "loss": 0.0053, "step": 16304 }, { "epoch": 0.32, "learning_rate": 4.472662694330688e-05, "loss": 0.0212, "step": 16306 }, { "epoch": 0.32, "learning_rate": 4.472597998304964e-05, "loss": 0.0001, "step": 16308 }, { "epoch": 0.32, "learning_rate": 4.472533302279241e-05, "loss": 0.0268, "step": 16310 }, { "epoch": 0.32, "learning_rate": 4.472468606253518e-05, "loss": 0.0002, "step": 16312 }, { "epoch": 0.32, "learning_rate": 4.472403910227795e-05, "loss": 0.0096, "step": 16314 }, { "epoch": 0.32, "learning_rate": 4.472339214202072e-05, "loss": 0.0107, "step": 16316 }, { "epoch": 0.32, "learning_rate": 4.4722745181763487e-05, "loss": 0.001, "step": 16318 }, { "epoch": 0.32, "learning_rate": 4.4722098221506256e-05, "loss": 0.0011, "step": 16320 }, { "epoch": 0.32, "learning_rate": 4.4721451261249025e-05, "loss": 0.0016, "step": 16322 }, { "epoch": 0.32, "learning_rate": 4.4720804300991794e-05, "loss": 0.0007, "step": 16324 }, { "epoch": 0.32, "learning_rate": 4.4720157340734556e-05, "loss": 0.0053, "step": 16326 }, { "epoch": 0.32, "learning_rate": 4.471951038047733e-05, "loss": 0.0004, "step": 16328 }, { "epoch": 0.32, "learning_rate": 4.4718863420220094e-05, "loss": 0.0087, "step": 16330 }, { "epoch": 0.32, "learning_rate": 4.4718216459962863e-05, "loss": 0.0168, "step": 16332 }, { "epoch": 0.32, "learning_rate": 4.471756949970564e-05, "loss": 0.0006, "step": 16334 }, { "epoch": 0.32, "learning_rate": 4.47169225394484e-05, "loss": 0.0053, "step": 16336 }, { "epoch": 0.32, "learning_rate": 4.471627557919118e-05, "loss": 0.0096, "step": 16338 }, { "epoch": 0.32, "learning_rate": 4.471562861893394e-05, "loss": 0.0156, "step": 16340 }, { "epoch": 0.32, "learning_rate": 4.471498165867671e-05, "loss": 0.0023, "step": 16342 }, { "epoch": 0.32, "learning_rate": 4.471433469841948e-05, "loss": 0.0004, "step": 16344 }, { "epoch": 0.32, "learning_rate": 4.471368773816225e-05, "loss": 0.0178, "step": 16346 }, { "epoch": 0.32, "learning_rate": 4.4713040777905016e-05, "loss": 0.0394, "step": 16348 }, { "epoch": 0.32, "learning_rate": 4.4712393817647785e-05, "loss": 0.0017, "step": 16350 }, { "epoch": 0.32, "learning_rate": 4.4711746857390554e-05, "loss": 0.0058, "step": 16352 }, { "epoch": 0.32, "learning_rate": 4.471109989713332e-05, "loss": 0.0101, "step": 16354 }, { "epoch": 0.32, "learning_rate": 4.471045293687609e-05, "loss": 0.0038, "step": 16356 }, { "epoch": 0.32, "learning_rate": 4.4709805976618855e-05, "loss": 0.0018, "step": 16358 }, { "epoch": 0.32, "learning_rate": 4.470915901636163e-05, "loss": 0.0088, "step": 16360 }, { "epoch": 0.32, "learning_rate": 4.470851205610439e-05, "loss": 0.0019, "step": 16362 }, { "epoch": 0.32, "learning_rate": 4.470786509584716e-05, "loss": 0.0001, "step": 16364 }, { "epoch": 0.32, "learning_rate": 4.470721813558993e-05, "loss": 0.0006, "step": 16366 }, { "epoch": 0.32, "learning_rate": 4.47065711753327e-05, "loss": 0.0002, "step": 16368 }, { "epoch": 0.32, "learning_rate": 4.470592421507547e-05, "loss": 0.0068, "step": 16370 }, { "epoch": 0.32, "learning_rate": 4.470527725481824e-05, "loss": 0.0004, "step": 16372 }, { "epoch": 0.32, "learning_rate": 4.470463029456101e-05, "loss": 0.0009, "step": 16374 }, { "epoch": 0.32, "learning_rate": 4.4703983334303776e-05, "loss": 0.0028, "step": 16376 }, { "epoch": 0.32, "learning_rate": 4.4703336374046546e-05, "loss": 0.0122, "step": 16378 }, { "epoch": 0.32, "learning_rate": 4.4702689413789315e-05, "loss": 0.0099, "step": 16380 }, { "epoch": 0.32, "learning_rate": 4.4702042453532084e-05, "loss": 0.0058, "step": 16382 }, { "epoch": 0.32, "learning_rate": 4.470139549327485e-05, "loss": 0.0012, "step": 16384 }, { "epoch": 0.32, "learning_rate": 4.4700748533017615e-05, "loss": 0.0007, "step": 16386 }, { "epoch": 0.32, "learning_rate": 4.470010157276039e-05, "loss": 0.0002, "step": 16388 }, { "epoch": 0.32, "learning_rate": 4.469945461250315e-05, "loss": 0.0098, "step": 16390 }, { "epoch": 0.32, "learning_rate": 4.469880765224592e-05, "loss": 0.0008, "step": 16392 }, { "epoch": 0.32, "learning_rate": 4.469816069198869e-05, "loss": 0.0048, "step": 16394 }, { "epoch": 0.32, "learning_rate": 4.469751373173146e-05, "loss": 0.0008, "step": 16396 }, { "epoch": 0.32, "learning_rate": 4.469686677147423e-05, "loss": 0.0001, "step": 16398 }, { "epoch": 0.32, "learning_rate": 4.4696219811217e-05, "loss": 0.0159, "step": 16400 }, { "epoch": 0.32, "learning_rate": 4.469557285095977e-05, "loss": 0.0404, "step": 16402 }, { "epoch": 0.32, "learning_rate": 4.469492589070254e-05, "loss": 0.0004, "step": 16404 }, { "epoch": 0.32, "learning_rate": 4.4694278930445306e-05, "loss": 0.0003, "step": 16406 }, { "epoch": 0.32, "learning_rate": 4.469363197018807e-05, "loss": 0.0003, "step": 16408 }, { "epoch": 0.32, "learning_rate": 4.4692985009930844e-05, "loss": 0.004, "step": 16410 }, { "epoch": 0.32, "learning_rate": 4.469233804967361e-05, "loss": 0.0004, "step": 16412 }, { "epoch": 0.32, "learning_rate": 4.469169108941638e-05, "loss": 0.0082, "step": 16414 }, { "epoch": 0.32, "learning_rate": 4.469104412915915e-05, "loss": 0.019, "step": 16416 }, { "epoch": 0.32, "learning_rate": 4.4690397168901914e-05, "loss": 0.002, "step": 16418 }, { "epoch": 0.32, "learning_rate": 4.468975020864469e-05, "loss": 0.0005, "step": 16420 }, { "epoch": 0.32, "learning_rate": 4.468910324838745e-05, "loss": 0.0117, "step": 16422 }, { "epoch": 0.32, "learning_rate": 4.468845628813022e-05, "loss": 0.0003, "step": 16424 }, { "epoch": 0.32, "learning_rate": 4.468780932787299e-05, "loss": 0.005, "step": 16426 }, { "epoch": 0.32, "learning_rate": 4.468716236761576e-05, "loss": 0.0013, "step": 16428 }, { "epoch": 0.32, "learning_rate": 4.468651540735853e-05, "loss": 0.0044, "step": 16430 }, { "epoch": 0.32, "learning_rate": 4.46858684471013e-05, "loss": 0.0066, "step": 16432 }, { "epoch": 0.32, "learning_rate": 4.4685221486844066e-05, "loss": 0.003, "step": 16434 }, { "epoch": 0.32, "learning_rate": 4.4684574526586835e-05, "loss": 0.0052, "step": 16436 }, { "epoch": 0.32, "learning_rate": 4.4683927566329605e-05, "loss": 0.0002, "step": 16438 }, { "epoch": 0.32, "learning_rate": 4.468328060607237e-05, "loss": 0.0055, "step": 16440 }, { "epoch": 0.32, "learning_rate": 4.468263364581514e-05, "loss": 0.0007, "step": 16442 }, { "epoch": 0.32, "learning_rate": 4.4681986685557905e-05, "loss": 0.0004, "step": 16444 }, { "epoch": 0.32, "learning_rate": 4.4681339725300674e-05, "loss": 0.0063, "step": 16446 }, { "epoch": 0.32, "learning_rate": 4.468069276504345e-05, "loss": 0.0179, "step": 16448 }, { "epoch": 0.32, "learning_rate": 4.468004580478621e-05, "loss": 0.0009, "step": 16450 }, { "epoch": 0.32, "learning_rate": 4.467939884452898e-05, "loss": 0.0004, "step": 16452 }, { "epoch": 0.32, "learning_rate": 4.467875188427175e-05, "loss": 0.0003, "step": 16454 }, { "epoch": 0.32, "learning_rate": 4.467810492401452e-05, "loss": 0.0009, "step": 16456 }, { "epoch": 0.32, "learning_rate": 4.467745796375729e-05, "loss": 0.0006, "step": 16458 }, { "epoch": 0.32, "learning_rate": 4.467681100350006e-05, "loss": 0.017, "step": 16460 }, { "epoch": 0.32, "learning_rate": 4.467616404324283e-05, "loss": 0.0008, "step": 16462 }, { "epoch": 0.32, "learning_rate": 4.4675517082985596e-05, "loss": 0.0073, "step": 16464 }, { "epoch": 0.32, "learning_rate": 4.4674870122728365e-05, "loss": 0.0016, "step": 16466 }, { "epoch": 0.32, "learning_rate": 4.467422316247113e-05, "loss": 0.0007, "step": 16468 }, { "epoch": 0.32, "learning_rate": 4.46735762022139e-05, "loss": 0.0042, "step": 16470 }, { "epoch": 0.32, "learning_rate": 4.4672929241956665e-05, "loss": 0.0199, "step": 16472 }, { "epoch": 0.32, "learning_rate": 4.4672282281699435e-05, "loss": 0.0045, "step": 16474 }, { "epoch": 0.32, "learning_rate": 4.4671635321442204e-05, "loss": 0.0002, "step": 16476 }, { "epoch": 0.32, "learning_rate": 4.467098836118497e-05, "loss": 0.0002, "step": 16478 }, { "epoch": 0.32, "learning_rate": 4.467034140092775e-05, "loss": 0.0132, "step": 16480 }, { "epoch": 0.32, "learning_rate": 4.466969444067051e-05, "loss": 0.0246, "step": 16482 }, { "epoch": 0.32, "learning_rate": 4.466904748041328e-05, "loss": 0.003, "step": 16484 }, { "epoch": 0.32, "learning_rate": 4.466840052015605e-05, "loss": 0.0033, "step": 16486 }, { "epoch": 0.32, "learning_rate": 4.466775355989882e-05, "loss": 0.0005, "step": 16488 }, { "epoch": 0.32, "learning_rate": 4.466710659964159e-05, "loss": 0.0015, "step": 16490 }, { "epoch": 0.32, "learning_rate": 4.4666459639384356e-05, "loss": 0.0001, "step": 16492 }, { "epoch": 0.32, "learning_rate": 4.4665812679127125e-05, "loss": 0.0003, "step": 16494 }, { "epoch": 0.32, "learning_rate": 4.4665165718869894e-05, "loss": 0.0104, "step": 16496 }, { "epoch": 0.32, "learning_rate": 4.4664518758612664e-05, "loss": 0.0084, "step": 16498 }, { "epoch": 0.32, "learning_rate": 4.4663871798355426e-05, "loss": 0.0168, "step": 16500 }, { "epoch": 0.32, "learning_rate": 4.46632248380982e-05, "loss": 0.0018, "step": 16502 }, { "epoch": 0.32, "learning_rate": 4.4662577877840964e-05, "loss": 0.0005, "step": 16504 }, { "epoch": 0.32, "learning_rate": 4.466193091758373e-05, "loss": 0.0101, "step": 16506 }, { "epoch": 0.32, "learning_rate": 4.46612839573265e-05, "loss": 0.0003, "step": 16508 }, { "epoch": 0.32, "learning_rate": 4.466063699706927e-05, "loss": 0.0003, "step": 16510 }, { "epoch": 0.32, "learning_rate": 4.465999003681204e-05, "loss": 0.0002, "step": 16512 }, { "epoch": 0.32, "learning_rate": 4.465934307655481e-05, "loss": 0.0005, "step": 16514 }, { "epoch": 0.32, "learning_rate": 4.465869611629758e-05, "loss": 0.0002, "step": 16516 }, { "epoch": 0.32, "learning_rate": 4.465804915604035e-05, "loss": 0.0002, "step": 16518 }, { "epoch": 0.32, "learning_rate": 4.465740219578312e-05, "loss": 0.0107, "step": 16520 }, { "epoch": 0.32, "learning_rate": 4.465675523552588e-05, "loss": 0.0008, "step": 16522 }, { "epoch": 0.32, "learning_rate": 4.4656108275268655e-05, "loss": 0.0068, "step": 16524 }, { "epoch": 0.32, "learning_rate": 4.4655461315011424e-05, "loss": 0.0578, "step": 16526 }, { "epoch": 0.32, "learning_rate": 4.4654814354754186e-05, "loss": 0.0002, "step": 16528 }, { "epoch": 0.32, "learning_rate": 4.465416739449696e-05, "loss": 0.0135, "step": 16530 }, { "epoch": 0.32, "learning_rate": 4.4653520434239724e-05, "loss": 0.0005, "step": 16532 }, { "epoch": 0.32, "learning_rate": 4.4652873473982493e-05, "loss": 0.0001, "step": 16534 }, { "epoch": 0.32, "learning_rate": 4.465222651372526e-05, "loss": 0.0088, "step": 16536 }, { "epoch": 0.32, "learning_rate": 4.465157955346803e-05, "loss": 0.0497, "step": 16538 }, { "epoch": 0.32, "learning_rate": 4.46509325932108e-05, "loss": 0.0012, "step": 16540 }, { "epoch": 0.32, "learning_rate": 4.465028563295357e-05, "loss": 0.0399, "step": 16542 }, { "epoch": 0.32, "learning_rate": 4.464963867269634e-05, "loss": 0.011, "step": 16544 }, { "epoch": 0.32, "learning_rate": 4.464899171243911e-05, "loss": 0.0017, "step": 16546 }, { "epoch": 0.32, "learning_rate": 4.464834475218188e-05, "loss": 0.0058, "step": 16548 }, { "epoch": 0.32, "learning_rate": 4.464769779192464e-05, "loss": 0.0125, "step": 16550 }, { "epoch": 0.32, "learning_rate": 4.4647050831667415e-05, "loss": 0.0252, "step": 16552 }, { "epoch": 0.32, "learning_rate": 4.464640387141018e-05, "loss": 0.0005, "step": 16554 }, { "epoch": 0.32, "learning_rate": 4.4645756911152953e-05, "loss": 0.0107, "step": 16556 }, { "epoch": 0.32, "learning_rate": 4.464510995089572e-05, "loss": 0.0114, "step": 16558 }, { "epoch": 0.32, "learning_rate": 4.4644462990638485e-05, "loss": 0.0019, "step": 16560 }, { "epoch": 0.32, "learning_rate": 4.464381603038126e-05, "loss": 0.0022, "step": 16562 }, { "epoch": 0.32, "learning_rate": 4.464316907012402e-05, "loss": 0.0235, "step": 16564 }, { "epoch": 0.32, "learning_rate": 4.464252210986679e-05, "loss": 0.0004, "step": 16566 }, { "epoch": 0.32, "learning_rate": 4.464187514960956e-05, "loss": 0.023, "step": 16568 }, { "epoch": 0.32, "learning_rate": 4.464122818935233e-05, "loss": 0.0002, "step": 16570 }, { "epoch": 0.32, "learning_rate": 4.46405812290951e-05, "loss": 0.0191, "step": 16572 }, { "epoch": 0.32, "learning_rate": 4.463993426883787e-05, "loss": 0.0023, "step": 16574 }, { "epoch": 0.32, "learning_rate": 4.463928730858064e-05, "loss": 0.018, "step": 16576 }, { "epoch": 0.32, "learning_rate": 4.4638640348323407e-05, "loss": 0.0016, "step": 16578 }, { "epoch": 0.32, "learning_rate": 4.4637993388066176e-05, "loss": 0.0059, "step": 16580 }, { "epoch": 0.32, "learning_rate": 4.463734642780894e-05, "loss": 0.0069, "step": 16582 }, { "epoch": 0.32, "learning_rate": 4.4636699467551714e-05, "loss": 0.0202, "step": 16584 }, { "epoch": 0.32, "learning_rate": 4.4636052507294476e-05, "loss": 0.001, "step": 16586 }, { "epoch": 0.32, "learning_rate": 4.4635405547037245e-05, "loss": 0.0036, "step": 16588 }, { "epoch": 0.32, "learning_rate": 4.463475858678002e-05, "loss": 0.0061, "step": 16590 }, { "epoch": 0.32, "learning_rate": 4.463411162652278e-05, "loss": 0.0052, "step": 16592 }, { "epoch": 0.32, "learning_rate": 4.463346466626555e-05, "loss": 0.0001, "step": 16594 }, { "epoch": 0.32, "learning_rate": 4.463281770600832e-05, "loss": 0.0019, "step": 16596 }, { "epoch": 0.32, "learning_rate": 4.463217074575109e-05, "loss": 0.0064, "step": 16598 }, { "epoch": 0.32, "learning_rate": 4.463152378549386e-05, "loss": 0.0004, "step": 16600 }, { "epoch": 0.32, "learning_rate": 4.463087682523663e-05, "loss": 0.0017, "step": 16602 }, { "epoch": 0.32, "learning_rate": 4.46302298649794e-05, "loss": 0.0584, "step": 16604 }, { "epoch": 0.32, "learning_rate": 4.462958290472217e-05, "loss": 0.0006, "step": 16606 }, { "epoch": 0.32, "learning_rate": 4.4628935944464936e-05, "loss": 0.0375, "step": 16608 }, { "epoch": 0.32, "learning_rate": 4.46282889842077e-05, "loss": 0.0001, "step": 16610 }, { "epoch": 0.32, "learning_rate": 4.4627642023950474e-05, "loss": 0.0208, "step": 16612 }, { "epoch": 0.32, "learning_rate": 4.4626995063693237e-05, "loss": 0.0041, "step": 16614 }, { "epoch": 0.32, "learning_rate": 4.4626348103436006e-05, "loss": 0.0039, "step": 16616 }, { "epoch": 0.32, "learning_rate": 4.4625701143178775e-05, "loss": 0.0002, "step": 16618 }, { "epoch": 0.32, "learning_rate": 4.4625054182921544e-05, "loss": 0.0003, "step": 16620 }, { "epoch": 0.32, "learning_rate": 4.462440722266431e-05, "loss": 0.0147, "step": 16622 }, { "epoch": 0.32, "learning_rate": 4.462376026240708e-05, "loss": 0.0003, "step": 16624 }, { "epoch": 0.32, "learning_rate": 4.462311330214985e-05, "loss": 0.0037, "step": 16626 }, { "epoch": 0.32, "learning_rate": 4.462246634189262e-05, "loss": 0.0165, "step": 16628 }, { "epoch": 0.32, "learning_rate": 4.462181938163539e-05, "loss": 0.006, "step": 16630 }, { "epoch": 0.32, "learning_rate": 4.462117242137815e-05, "loss": 0.0011, "step": 16632 }, { "epoch": 0.32, "learning_rate": 4.462052546112093e-05, "loss": 0.0034, "step": 16634 }, { "epoch": 0.32, "learning_rate": 4.4619878500863696e-05, "loss": 0.009, "step": 16636 }, { "epoch": 0.32, "learning_rate": 4.4619231540606466e-05, "loss": 0.0032, "step": 16638 }, { "epoch": 0.32, "learning_rate": 4.4618584580349235e-05, "loss": 0.0015, "step": 16640 }, { "epoch": 0.32, "learning_rate": 4.4617937620092e-05, "loss": 0.0004, "step": 16642 }, { "epoch": 0.32, "learning_rate": 4.461729065983477e-05, "loss": 0.0005, "step": 16644 }, { "epoch": 0.32, "learning_rate": 4.4616643699577535e-05, "loss": 0.0087, "step": 16646 }, { "epoch": 0.32, "learning_rate": 4.4615996739320304e-05, "loss": 0.0021, "step": 16648 }, { "epoch": 0.32, "learning_rate": 4.461534977906307e-05, "loss": 0.0008, "step": 16650 }, { "epoch": 0.32, "learning_rate": 4.461470281880584e-05, "loss": 0.0028, "step": 16652 }, { "epoch": 0.32, "learning_rate": 4.461405585854861e-05, "loss": 0.0008, "step": 16654 }, { "epoch": 0.32, "learning_rate": 4.461340889829138e-05, "loss": 0.0005, "step": 16656 }, { "epoch": 0.32, "learning_rate": 4.461276193803415e-05, "loss": 0.0098, "step": 16658 }, { "epoch": 0.32, "learning_rate": 4.461211497777692e-05, "loss": 0.0003, "step": 16660 }, { "epoch": 0.32, "learning_rate": 4.461146801751969e-05, "loss": 0.0006, "step": 16662 }, { "epoch": 0.32, "learning_rate": 4.461082105726245e-05, "loss": 0.0016, "step": 16664 }, { "epoch": 0.32, "learning_rate": 4.4610174097005226e-05, "loss": 0.0038, "step": 16666 }, { "epoch": 0.32, "learning_rate": 4.4609527136747995e-05, "loss": 0.01, "step": 16668 }, { "epoch": 0.32, "learning_rate": 4.460888017649076e-05, "loss": 0.0002, "step": 16670 }, { "epoch": 0.32, "learning_rate": 4.460823321623353e-05, "loss": 0.0266, "step": 16672 }, { "epoch": 0.32, "learning_rate": 4.4607586255976296e-05, "loss": 0.0009, "step": 16674 }, { "epoch": 0.32, "learning_rate": 4.4606939295719065e-05, "loss": 0.0002, "step": 16676 }, { "epoch": 0.32, "learning_rate": 4.4606292335461834e-05, "loss": 0.0003, "step": 16678 }, { "epoch": 0.32, "learning_rate": 4.46056453752046e-05, "loss": 0.0044, "step": 16680 }, { "epoch": 0.32, "learning_rate": 4.460499841494737e-05, "loss": 0.0021, "step": 16682 }, { "epoch": 0.32, "learning_rate": 4.460435145469014e-05, "loss": 0.0018, "step": 16684 }, { "epoch": 0.32, "learning_rate": 4.460370449443291e-05, "loss": 0.0092, "step": 16686 }, { "epoch": 0.32, "learning_rate": 4.460305753417568e-05, "loss": 0.013, "step": 16688 }, { "epoch": 0.32, "learning_rate": 4.460241057391845e-05, "loss": 0.0129, "step": 16690 }, { "epoch": 0.32, "learning_rate": 4.460176361366121e-05, "loss": 0.0163, "step": 16692 }, { "epoch": 0.32, "learning_rate": 4.4601116653403986e-05, "loss": 0.0056, "step": 16694 }, { "epoch": 0.32, "learning_rate": 4.460046969314675e-05, "loss": 0.0022, "step": 16696 }, { "epoch": 0.32, "learning_rate": 4.4599822732889525e-05, "loss": 0.0009, "step": 16698 }, { "epoch": 0.32, "learning_rate": 4.459917577263229e-05, "loss": 0.0088, "step": 16700 }, { "epoch": 0.32, "learning_rate": 4.4598528812375056e-05, "loss": 0.0008, "step": 16702 }, { "epoch": 0.32, "learning_rate": 4.459788185211783e-05, "loss": 0.0018, "step": 16704 }, { "epoch": 0.32, "learning_rate": 4.4597234891860594e-05, "loss": 0.0092, "step": 16706 }, { "epoch": 0.32, "learning_rate": 4.459658793160336e-05, "loss": 0.0047, "step": 16708 }, { "epoch": 0.32, "learning_rate": 4.459594097134613e-05, "loss": 0.0046, "step": 16710 }, { "epoch": 0.32, "learning_rate": 4.45952940110889e-05, "loss": 0.0062, "step": 16712 }, { "epoch": 0.32, "learning_rate": 4.459464705083167e-05, "loss": 0.0002, "step": 16714 }, { "epoch": 0.32, "learning_rate": 4.459400009057444e-05, "loss": 0.0219, "step": 16716 }, { "epoch": 0.32, "learning_rate": 4.459335313031721e-05, "loss": 0.0001, "step": 16718 }, { "epoch": 0.32, "learning_rate": 4.459270617005998e-05, "loss": 0.0007, "step": 16720 }, { "epoch": 0.32, "learning_rate": 4.459205920980275e-05, "loss": 0.0077, "step": 16722 }, { "epoch": 0.32, "learning_rate": 4.459141224954551e-05, "loss": 0.0101, "step": 16724 }, { "epoch": 0.32, "learning_rate": 4.4590765289288285e-05, "loss": 0.0001, "step": 16726 }, { "epoch": 0.32, "learning_rate": 4.459011832903105e-05, "loss": 0.0426, "step": 16728 }, { "epoch": 0.32, "learning_rate": 4.4589471368773816e-05, "loss": 0.0177, "step": 16730 }, { "epoch": 0.32, "learning_rate": 4.4588824408516585e-05, "loss": 0.0077, "step": 16732 }, { "epoch": 0.32, "learning_rate": 4.4588177448259354e-05, "loss": 0.0005, "step": 16734 }, { "epoch": 0.32, "learning_rate": 4.4587530488002124e-05, "loss": 0.0011, "step": 16736 }, { "epoch": 0.32, "learning_rate": 4.458688352774489e-05, "loss": 0.0003, "step": 16738 }, { "epoch": 0.32, "learning_rate": 4.458623656748766e-05, "loss": 0.0075, "step": 16740 }, { "epoch": 0.32, "learning_rate": 4.458558960723043e-05, "loss": 0.0032, "step": 16742 }, { "epoch": 0.32, "learning_rate": 4.45849426469732e-05, "loss": 0.0223, "step": 16744 }, { "epoch": 0.33, "learning_rate": 4.458429568671596e-05, "loss": 0.0015, "step": 16746 }, { "epoch": 0.33, "learning_rate": 4.458364872645874e-05, "loss": 0.0145, "step": 16748 }, { "epoch": 0.33, "learning_rate": 4.458300176620151e-05, "loss": 0.0119, "step": 16750 }, { "epoch": 0.33, "learning_rate": 4.458235480594427e-05, "loss": 0.01, "step": 16752 }, { "epoch": 0.33, "learning_rate": 4.4581707845687045e-05, "loss": 0.0014, "step": 16754 }, { "epoch": 0.33, "learning_rate": 4.458106088542981e-05, "loss": 0.0158, "step": 16756 }, { "epoch": 0.33, "learning_rate": 4.458041392517258e-05, "loss": 0.0002, "step": 16758 }, { "epoch": 0.33, "learning_rate": 4.4579766964915346e-05, "loss": 0.0128, "step": 16760 }, { "epoch": 0.33, "learning_rate": 4.4579120004658115e-05, "loss": 0.0002, "step": 16762 }, { "epoch": 0.33, "learning_rate": 4.4578473044400884e-05, "loss": 0.0077, "step": 16764 }, { "epoch": 0.33, "learning_rate": 4.457782608414365e-05, "loss": 0.0001, "step": 16766 }, { "epoch": 0.33, "learning_rate": 4.457717912388642e-05, "loss": 0.0163, "step": 16768 }, { "epoch": 0.33, "learning_rate": 4.457653216362919e-05, "loss": 0.0111, "step": 16770 }, { "epoch": 0.33, "learning_rate": 4.457588520337196e-05, "loss": 0.0182, "step": 16772 }, { "epoch": 0.33, "learning_rate": 4.457523824311472e-05, "loss": 0.0029, "step": 16774 }, { "epoch": 0.33, "learning_rate": 4.45745912828575e-05, "loss": 0.0168, "step": 16776 }, { "epoch": 0.33, "learning_rate": 4.457394432260026e-05, "loss": 0.0036, "step": 16778 }, { "epoch": 0.33, "learning_rate": 4.457329736234304e-05, "loss": 0.0079, "step": 16780 }, { "epoch": 0.33, "learning_rate": 4.4572650402085806e-05, "loss": 0.0022, "step": 16782 }, { "epoch": 0.33, "learning_rate": 4.457200344182857e-05, "loss": 0.0002, "step": 16784 }, { "epoch": 0.33, "learning_rate": 4.4571356481571344e-05, "loss": 0.0132, "step": 16786 }, { "epoch": 0.33, "learning_rate": 4.4570709521314106e-05, "loss": 0.0069, "step": 16788 }, { "epoch": 0.33, "learning_rate": 4.4570062561056875e-05, "loss": 0.0045, "step": 16790 }, { "epoch": 0.33, "learning_rate": 4.4569415600799644e-05, "loss": 0.0111, "step": 16792 }, { "epoch": 0.33, "learning_rate": 4.4568768640542413e-05, "loss": 0.0033, "step": 16794 }, { "epoch": 0.33, "learning_rate": 4.456812168028518e-05, "loss": 0.0005, "step": 16796 }, { "epoch": 0.33, "learning_rate": 4.456747472002795e-05, "loss": 0.0043, "step": 16798 }, { "epoch": 0.33, "learning_rate": 4.456682775977072e-05, "loss": 0.0005, "step": 16800 }, { "epoch": 0.33, "learning_rate": 4.456618079951349e-05, "loss": 0.0003, "step": 16802 }, { "epoch": 0.33, "learning_rate": 4.456553383925626e-05, "loss": 0.0017, "step": 16804 }, { "epoch": 0.33, "learning_rate": 4.456488687899902e-05, "loss": 0.0083, "step": 16806 }, { "epoch": 0.33, "learning_rate": 4.45642399187418e-05, "loss": 0.006, "step": 16808 }, { "epoch": 0.33, "learning_rate": 4.456359295848456e-05, "loss": 0.0138, "step": 16810 }, { "epoch": 0.33, "learning_rate": 4.456294599822733e-05, "loss": 0.0096, "step": 16812 }, { "epoch": 0.33, "learning_rate": 4.4562299037970104e-05, "loss": 0.0031, "step": 16814 }, { "epoch": 0.33, "learning_rate": 4.456165207771287e-05, "loss": 0.0009, "step": 16816 }, { "epoch": 0.33, "learning_rate": 4.4561005117455636e-05, "loss": 0.0078, "step": 16818 }, { "epoch": 0.33, "learning_rate": 4.4560358157198405e-05, "loss": 0.0027, "step": 16820 }, { "epoch": 0.33, "learning_rate": 4.4559711196941174e-05, "loss": 0.001, "step": 16822 }, { "epoch": 0.33, "learning_rate": 4.455906423668394e-05, "loss": 0.0002, "step": 16824 }, { "epoch": 0.33, "learning_rate": 4.455841727642671e-05, "loss": 0.0009, "step": 16826 }, { "epoch": 0.33, "learning_rate": 4.455777031616948e-05, "loss": 0.0001, "step": 16828 }, { "epoch": 0.33, "learning_rate": 4.455712335591225e-05, "loss": 0.0034, "step": 16830 }, { "epoch": 0.33, "learning_rate": 4.455647639565502e-05, "loss": 0.004, "step": 16832 }, { "epoch": 0.33, "learning_rate": 4.455582943539778e-05, "loss": 0.0002, "step": 16834 }, { "epoch": 0.33, "learning_rate": 4.455518247514056e-05, "loss": 0.0126, "step": 16836 }, { "epoch": 0.33, "learning_rate": 4.455453551488332e-05, "loss": 0.0282, "step": 16838 }, { "epoch": 0.33, "learning_rate": 4.4554212034754704e-05, "loss": 0.1832, "step": 16840 }, { "epoch": 0.33, "learning_rate": 4.455356507449748e-05, "loss": 0.0052, "step": 16842 }, { "epoch": 0.33, "learning_rate": 4.455291811424024e-05, "loss": 0.0004, "step": 16844 }, { "epoch": 0.33, "learning_rate": 4.455227115398301e-05, "loss": 0.0004, "step": 16846 }, { "epoch": 0.33, "learning_rate": 4.455162419372578e-05, "loss": 0.0003, "step": 16848 }, { "epoch": 0.33, "learning_rate": 4.455097723346855e-05, "loss": 0.0027, "step": 16850 }, { "epoch": 0.33, "learning_rate": 4.455033027321132e-05, "loss": 0.0119, "step": 16852 }, { "epoch": 0.33, "learning_rate": 4.454968331295409e-05, "loss": 0.0132, "step": 16854 }, { "epoch": 0.33, "learning_rate": 4.454903635269686e-05, "loss": 0.0031, "step": 16856 }, { "epoch": 0.33, "learning_rate": 4.4548389392439626e-05, "loss": 0.0075, "step": 16858 }, { "epoch": 0.33, "learning_rate": 4.4547742432182395e-05, "loss": 0.0172, "step": 16860 }, { "epoch": 0.33, "learning_rate": 4.454709547192516e-05, "loss": 0.0048, "step": 16862 }, { "epoch": 0.33, "learning_rate": 4.454644851166793e-05, "loss": 0.0002, "step": 16864 }, { "epoch": 0.33, "learning_rate": 4.4545801551410696e-05, "loss": 0.0138, "step": 16866 }, { "epoch": 0.33, "learning_rate": 4.4545154591153465e-05, "loss": 0.0051, "step": 16868 }, { "epoch": 0.33, "learning_rate": 4.454450763089624e-05, "loss": 0.0163, "step": 16870 }, { "epoch": 0.33, "learning_rate": 4.4543860670639e-05, "loss": 0.0061, "step": 16872 }, { "epoch": 0.33, "learning_rate": 4.454321371038178e-05, "loss": 0.0002, "step": 16874 }, { "epoch": 0.33, "learning_rate": 4.454256675012454e-05, "loss": 0.0081, "step": 16876 }, { "epoch": 0.33, "learning_rate": 4.454191978986731e-05, "loss": 0.0341, "step": 16878 }, { "epoch": 0.33, "learning_rate": 4.454127282961008e-05, "loss": 0.0009, "step": 16880 }, { "epoch": 0.33, "learning_rate": 4.454062586935285e-05, "loss": 0.0009, "step": 16882 }, { "epoch": 0.33, "learning_rate": 4.453997890909562e-05, "loss": 0.0007, "step": 16884 }, { "epoch": 0.33, "learning_rate": 4.4539331948838386e-05, "loss": 0.0007, "step": 16886 }, { "epoch": 0.33, "learning_rate": 4.4538684988581156e-05, "loss": 0.0105, "step": 16888 }, { "epoch": 0.33, "learning_rate": 4.453803802832392e-05, "loss": 0.008, "step": 16890 }, { "epoch": 0.33, "learning_rate": 4.4537391068066694e-05, "loss": 0.0184, "step": 16892 }, { "epoch": 0.33, "learning_rate": 4.4536744107809456e-05, "loss": 0.0082, "step": 16894 }, { "epoch": 0.33, "learning_rate": 4.453609714755223e-05, "loss": 0.0012, "step": 16896 }, { "epoch": 0.33, "learning_rate": 4.4535450187294994e-05, "loss": 0.0032, "step": 16898 }, { "epoch": 0.33, "learning_rate": 4.453480322703776e-05, "loss": 0.0005, "step": 16900 }, { "epoch": 0.33, "learning_rate": 4.453415626678053e-05, "loss": 0.0001, "step": 16902 }, { "epoch": 0.33, "learning_rate": 4.45335093065233e-05, "loss": 0.0005, "step": 16904 }, { "epoch": 0.33, "learning_rate": 4.453286234626607e-05, "loss": 0.0014, "step": 16906 }, { "epoch": 0.33, "learning_rate": 4.453221538600884e-05, "loss": 0.0076, "step": 16908 }, { "epoch": 0.33, "learning_rate": 4.453156842575161e-05, "loss": 0.0008, "step": 16910 }, { "epoch": 0.33, "learning_rate": 4.453092146549437e-05, "loss": 0.003, "step": 16912 }, { "epoch": 0.33, "learning_rate": 4.453027450523715e-05, "loss": 0.0027, "step": 16914 }, { "epoch": 0.33, "learning_rate": 4.4529627544979916e-05, "loss": 0.0007, "step": 16916 }, { "epoch": 0.33, "learning_rate": 4.4528980584722685e-05, "loss": 0.0006, "step": 16918 }, { "epoch": 0.33, "learning_rate": 4.4528333624465454e-05, "loss": 0.0004, "step": 16920 }, { "epoch": 0.33, "learning_rate": 4.4527686664208216e-05, "loss": 0.001, "step": 16922 }, { "epoch": 0.33, "learning_rate": 4.452703970395099e-05, "loss": 0.0003, "step": 16924 }, { "epoch": 0.33, "learning_rate": 4.4526392743693755e-05, "loss": 0.0013, "step": 16926 }, { "epoch": 0.33, "learning_rate": 4.4525745783436524e-05, "loss": 0.0216, "step": 16928 }, { "epoch": 0.33, "learning_rate": 4.452509882317929e-05, "loss": 0.0011, "step": 16930 }, { "epoch": 0.33, "learning_rate": 4.452445186292206e-05, "loss": 0.0001, "step": 16932 }, { "epoch": 0.33, "learning_rate": 4.452380490266483e-05, "loss": 0.0086, "step": 16934 }, { "epoch": 0.33, "learning_rate": 4.45231579424076e-05, "loss": 0.0053, "step": 16936 }, { "epoch": 0.33, "learning_rate": 4.452251098215037e-05, "loss": 0.0008, "step": 16938 }, { "epoch": 0.33, "learning_rate": 4.452186402189314e-05, "loss": 0.0087, "step": 16940 }, { "epoch": 0.33, "learning_rate": 4.452121706163591e-05, "loss": 0.0093, "step": 16942 }, { "epoch": 0.33, "learning_rate": 4.452057010137867e-05, "loss": 0.003, "step": 16944 }, { "epoch": 0.33, "learning_rate": 4.4519923141121445e-05, "loss": 0.0105, "step": 16946 }, { "epoch": 0.33, "learning_rate": 4.451927618086421e-05, "loss": 0.0054, "step": 16948 }, { "epoch": 0.33, "learning_rate": 4.451862922060698e-05, "loss": 0.0001, "step": 16950 }, { "epoch": 0.33, "learning_rate": 4.451798226034975e-05, "loss": 0.0002, "step": 16952 }, { "epoch": 0.33, "learning_rate": 4.4517335300092515e-05, "loss": 0.001, "step": 16954 }, { "epoch": 0.33, "learning_rate": 4.451668833983529e-05, "loss": 0.0007, "step": 16956 }, { "epoch": 0.33, "learning_rate": 4.451604137957805e-05, "loss": 0.0221, "step": 16958 }, { "epoch": 0.33, "learning_rate": 4.451539441932082e-05, "loss": 0.0035, "step": 16960 }, { "epoch": 0.33, "learning_rate": 4.451474745906359e-05, "loss": 0.0063, "step": 16962 }, { "epoch": 0.33, "learning_rate": 4.451410049880636e-05, "loss": 0.0161, "step": 16964 }, { "epoch": 0.33, "learning_rate": 4.451345353854913e-05, "loss": 0.0029, "step": 16966 }, { "epoch": 0.33, "learning_rate": 4.45128065782919e-05, "loss": 0.0024, "step": 16968 }, { "epoch": 0.33, "learning_rate": 4.451215961803467e-05, "loss": 0.0004, "step": 16970 }, { "epoch": 0.33, "learning_rate": 4.451151265777743e-05, "loss": 0.0047, "step": 16972 }, { "epoch": 0.33, "learning_rate": 4.4510865697520206e-05, "loss": 0.0164, "step": 16974 }, { "epoch": 0.33, "learning_rate": 4.451021873726297e-05, "loss": 0.0004, "step": 16976 }, { "epoch": 0.33, "learning_rate": 4.4509571777005744e-05, "loss": 0.0019, "step": 16978 }, { "epoch": 0.33, "learning_rate": 4.4508924816748506e-05, "loss": 0.0026, "step": 16980 }, { "epoch": 0.33, "learning_rate": 4.4508277856491275e-05, "loss": 0.0001, "step": 16982 }, { "epoch": 0.33, "learning_rate": 4.450763089623405e-05, "loss": 0.0007, "step": 16984 }, { "epoch": 0.33, "learning_rate": 4.4506983935976814e-05, "loss": 0.0018, "step": 16986 }, { "epoch": 0.33, "learning_rate": 4.450633697571958e-05, "loss": 0.0261, "step": 16988 }, { "epoch": 0.33, "learning_rate": 4.450569001546235e-05, "loss": 0.0007, "step": 16990 }, { "epoch": 0.33, "learning_rate": 4.450504305520512e-05, "loss": 0.0003, "step": 16992 }, { "epoch": 0.33, "learning_rate": 4.450439609494789e-05, "loss": 0.0048, "step": 16994 }, { "epoch": 0.33, "learning_rate": 4.450374913469066e-05, "loss": 0.0052, "step": 16996 }, { "epoch": 0.33, "learning_rate": 4.450310217443343e-05, "loss": 0.0004, "step": 16998 }, { "epoch": 0.33, "learning_rate": 4.45024552141762e-05, "loss": 0.001, "step": 17000 }, { "epoch": 0.33, "learning_rate": 4.4501808253918966e-05, "loss": 0.0063, "step": 17002 }, { "epoch": 0.33, "learning_rate": 4.450116129366173e-05, "loss": 0.0015, "step": 17004 }, { "epoch": 0.33, "learning_rate": 4.4500514333404504e-05, "loss": 0.0074, "step": 17006 }, { "epoch": 0.33, "learning_rate": 4.449986737314727e-05, "loss": 0.0018, "step": 17008 }, { "epoch": 0.33, "learning_rate": 4.4499220412890036e-05, "loss": 0.0019, "step": 17010 }, { "epoch": 0.33, "learning_rate": 4.4498573452632805e-05, "loss": 0.0005, "step": 17012 }, { "epoch": 0.33, "learning_rate": 4.4497926492375574e-05, "loss": 0.0072, "step": 17014 }, { "epoch": 0.33, "learning_rate": 4.449727953211835e-05, "loss": 0.0234, "step": 17016 }, { "epoch": 0.33, "learning_rate": 4.449663257186111e-05, "loss": 0.0002, "step": 17018 }, { "epoch": 0.33, "learning_rate": 4.449598561160388e-05, "loss": 0.0004, "step": 17020 }, { "epoch": 0.33, "learning_rate": 4.449533865134665e-05, "loss": 0.024, "step": 17022 }, { "epoch": 0.33, "learning_rate": 4.449469169108942e-05, "loss": 0.0001, "step": 17024 }, { "epoch": 0.33, "learning_rate": 4.449404473083218e-05, "loss": 0.0001, "step": 17026 }, { "epoch": 0.33, "learning_rate": 4.449339777057496e-05, "loss": 0.0131, "step": 17028 }, { "epoch": 0.33, "learning_rate": 4.449275081031773e-05, "loss": 0.0055, "step": 17030 }, { "epoch": 0.33, "learning_rate": 4.449210385006049e-05, "loss": 0.0003, "step": 17032 }, { "epoch": 0.33, "learning_rate": 4.4491456889803265e-05, "loss": 0.0004, "step": 17034 }, { "epoch": 0.33, "learning_rate": 4.449080992954603e-05, "loss": 0.0029, "step": 17036 }, { "epoch": 0.33, "learning_rate": 4.44901629692888e-05, "loss": 0.0014, "step": 17038 }, { "epoch": 0.33, "learning_rate": 4.4489516009031565e-05, "loss": 0.0003, "step": 17040 }, { "epoch": 0.33, "learning_rate": 4.4488869048774334e-05, "loss": 0.0145, "step": 17042 }, { "epoch": 0.33, "learning_rate": 4.4488222088517103e-05, "loss": 0.0008, "step": 17044 }, { "epoch": 0.33, "learning_rate": 4.448757512825987e-05, "loss": 0.0004, "step": 17046 }, { "epoch": 0.33, "learning_rate": 4.448692816800264e-05, "loss": 0.0283, "step": 17048 }, { "epoch": 0.33, "learning_rate": 4.448628120774541e-05, "loss": 0.0003, "step": 17050 }, { "epoch": 0.33, "learning_rate": 4.448563424748818e-05, "loss": 0.0005, "step": 17052 }, { "epoch": 0.33, "learning_rate": 4.448498728723094e-05, "loss": 0.0015, "step": 17054 }, { "epoch": 0.33, "learning_rate": 4.448434032697372e-05, "loss": 0.0089, "step": 17056 }, { "epoch": 0.33, "learning_rate": 4.448369336671648e-05, "loss": 0.0013, "step": 17058 }, { "epoch": 0.33, "learning_rate": 4.4483046406459256e-05, "loss": 0.0127, "step": 17060 }, { "epoch": 0.33, "learning_rate": 4.4482399446202025e-05, "loss": 0.0001, "step": 17062 }, { "epoch": 0.33, "learning_rate": 4.448175248594479e-05, "loss": 0.002, "step": 17064 }, { "epoch": 0.33, "learning_rate": 4.4481105525687563e-05, "loss": 0.0097, "step": 17066 }, { "epoch": 0.33, "learning_rate": 4.4480458565430326e-05, "loss": 0.0011, "step": 17068 }, { "epoch": 0.33, "learning_rate": 4.4479811605173095e-05, "loss": 0.0339, "step": 17070 }, { "epoch": 0.33, "learning_rate": 4.4479164644915864e-05, "loss": 0.0001, "step": 17072 }, { "epoch": 0.33, "learning_rate": 4.447851768465863e-05, "loss": 0.004, "step": 17074 }, { "epoch": 0.33, "learning_rate": 4.44778707244014e-05, "loss": 0.0002, "step": 17076 }, { "epoch": 0.33, "learning_rate": 4.447722376414417e-05, "loss": 0.0007, "step": 17078 }, { "epoch": 0.33, "learning_rate": 4.447657680388694e-05, "loss": 0.0023, "step": 17080 }, { "epoch": 0.33, "learning_rate": 4.447592984362971e-05, "loss": 0.0008, "step": 17082 }, { "epoch": 0.33, "learning_rate": 4.447528288337248e-05, "loss": 0.0002, "step": 17084 }, { "epoch": 0.33, "learning_rate": 4.447463592311524e-05, "loss": 0.0044, "step": 17086 }, { "epoch": 0.33, "learning_rate": 4.4473988962858017e-05, "loss": 0.0005, "step": 17088 }, { "epoch": 0.33, "learning_rate": 4.447334200260078e-05, "loss": 0.0025, "step": 17090 }, { "epoch": 0.33, "learning_rate": 4.447269504234355e-05, "loss": 0.0181, "step": 17092 }, { "epoch": 0.33, "learning_rate": 4.4472048082086324e-05, "loss": 0.0003, "step": 17094 }, { "epoch": 0.33, "learning_rate": 4.4471401121829086e-05, "loss": 0.0004, "step": 17096 }, { "epoch": 0.33, "learning_rate": 4.447075416157186e-05, "loss": 0.0008, "step": 17098 }, { "epoch": 0.33, "learning_rate": 4.4470107201314624e-05, "loss": 0.0208, "step": 17100 }, { "epoch": 0.33, "learning_rate": 4.446946024105739e-05, "loss": 0.0062, "step": 17102 }, { "epoch": 0.33, "learning_rate": 4.446881328080016e-05, "loss": 0.0002, "step": 17104 }, { "epoch": 0.33, "learning_rate": 4.446816632054293e-05, "loss": 0.001, "step": 17106 }, { "epoch": 0.33, "learning_rate": 4.44675193602857e-05, "loss": 0.001, "step": 17108 }, { "epoch": 0.33, "learning_rate": 4.446687240002847e-05, "loss": 0.0037, "step": 17110 }, { "epoch": 0.33, "learning_rate": 4.446622543977124e-05, "loss": 0.0001, "step": 17112 }, { "epoch": 0.33, "learning_rate": 4.4465578479514e-05, "loss": 0.0001, "step": 17114 }, { "epoch": 0.33, "learning_rate": 4.446493151925678e-05, "loss": 0.0035, "step": 17116 }, { "epoch": 0.33, "learning_rate": 4.446428455899954e-05, "loss": 0.0027, "step": 17118 }, { "epoch": 0.33, "learning_rate": 4.4463637598742315e-05, "loss": 0.0032, "step": 17120 }, { "epoch": 0.33, "learning_rate": 4.446299063848508e-05, "loss": 0.0111, "step": 17122 }, { "epoch": 0.33, "learning_rate": 4.4462343678227847e-05, "loss": 0.0018, "step": 17124 }, { "epoch": 0.33, "learning_rate": 4.4461696717970616e-05, "loss": 0.0004, "step": 17126 }, { "epoch": 0.33, "learning_rate": 4.4461049757713385e-05, "loss": 0.0009, "step": 17128 }, { "epoch": 0.33, "learning_rate": 4.4460402797456154e-05, "loss": 0.0029, "step": 17130 }, { "epoch": 0.33, "learning_rate": 4.445975583719892e-05, "loss": 0.0007, "step": 17132 }, { "epoch": 0.33, "learning_rate": 4.445910887694169e-05, "loss": 0.0041, "step": 17134 }, { "epoch": 0.33, "learning_rate": 4.445846191668446e-05, "loss": 0.0147, "step": 17136 }, { "epoch": 0.33, "learning_rate": 4.445781495642723e-05, "loss": 0.0014, "step": 17138 }, { "epoch": 0.33, "learning_rate": 4.445716799617e-05, "loss": 0.0016, "step": 17140 }, { "epoch": 0.33, "learning_rate": 4.445652103591277e-05, "loss": 0.0009, "step": 17142 }, { "epoch": 0.33, "learning_rate": 4.445587407565554e-05, "loss": 0.0002, "step": 17144 }, { "epoch": 0.33, "learning_rate": 4.44552271153983e-05, "loss": 0.0411, "step": 17146 }, { "epoch": 0.33, "learning_rate": 4.4454580155141076e-05, "loss": 0.0036, "step": 17148 }, { "epoch": 0.33, "learning_rate": 4.445393319488384e-05, "loss": 0.0101, "step": 17150 }, { "epoch": 0.33, "learning_rate": 4.445328623462661e-05, "loss": 0.0007, "step": 17152 }, { "epoch": 0.33, "learning_rate": 4.4452639274369376e-05, "loss": 0.0017, "step": 17154 }, { "epoch": 0.33, "learning_rate": 4.4451992314112145e-05, "loss": 0.0116, "step": 17156 }, { "epoch": 0.33, "learning_rate": 4.4451345353854914e-05, "loss": 0.0026, "step": 17158 }, { "epoch": 0.33, "learning_rate": 4.445069839359768e-05, "loss": 0.014, "step": 17160 }, { "epoch": 0.33, "learning_rate": 4.445005143334045e-05, "loss": 0.0001, "step": 17162 }, { "epoch": 0.33, "learning_rate": 4.444940447308322e-05, "loss": 0.0044, "step": 17164 }, { "epoch": 0.33, "learning_rate": 4.444875751282599e-05, "loss": 0.0017, "step": 17166 }, { "epoch": 0.33, "learning_rate": 4.444811055256875e-05, "loss": 0.0121, "step": 17168 }, { "epoch": 0.33, "learning_rate": 4.444746359231153e-05, "loss": 0.0002, "step": 17170 }, { "epoch": 0.33, "learning_rate": 4.44468166320543e-05, "loss": 0.0092, "step": 17172 }, { "epoch": 0.33, "learning_rate": 4.444616967179706e-05, "loss": 0.0045, "step": 17174 }, { "epoch": 0.33, "learning_rate": 4.4445522711539836e-05, "loss": 0.0016, "step": 17176 }, { "epoch": 0.33, "learning_rate": 4.44448757512826e-05, "loss": 0.002, "step": 17178 }, { "epoch": 0.33, "learning_rate": 4.4444228791025374e-05, "loss": 0.0184, "step": 17180 }, { "epoch": 0.33, "learning_rate": 4.4443581830768136e-05, "loss": 0.0002, "step": 17182 }, { "epoch": 0.33, "learning_rate": 4.4442934870510906e-05, "loss": 0.0006, "step": 17184 }, { "epoch": 0.33, "learning_rate": 4.4442287910253675e-05, "loss": 0.0003, "step": 17186 }, { "epoch": 0.33, "learning_rate": 4.4441640949996444e-05, "loss": 0.0008, "step": 17188 }, { "epoch": 0.33, "learning_rate": 4.444099398973921e-05, "loss": 0.0106, "step": 17190 }, { "epoch": 0.33, "learning_rate": 4.444034702948198e-05, "loss": 0.0154, "step": 17192 }, { "epoch": 0.33, "learning_rate": 4.443970006922475e-05, "loss": 0.0011, "step": 17194 }, { "epoch": 0.33, "learning_rate": 4.443905310896751e-05, "loss": 0.0008, "step": 17196 }, { "epoch": 0.33, "learning_rate": 4.443840614871029e-05, "loss": 0.0002, "step": 17198 }, { "epoch": 0.33, "learning_rate": 4.443775918845305e-05, "loss": 0.0023, "step": 17200 }, { "epoch": 0.33, "learning_rate": 4.443711222819583e-05, "loss": 0.0043, "step": 17202 }, { "epoch": 0.33, "learning_rate": 4.443646526793859e-05, "loss": 0.001, "step": 17204 }, { "epoch": 0.33, "learning_rate": 4.443581830768136e-05, "loss": 0.0026, "step": 17206 }, { "epoch": 0.33, "learning_rate": 4.4435171347424135e-05, "loss": 0.009, "step": 17208 }, { "epoch": 0.33, "learning_rate": 4.44345243871669e-05, "loss": 0.0092, "step": 17210 }, { "epoch": 0.33, "learning_rate": 4.4433877426909666e-05, "loss": 0.0263, "step": 17212 }, { "epoch": 0.33, "learning_rate": 4.4433230466652435e-05, "loss": 0.0016, "step": 17214 }, { "epoch": 0.33, "learning_rate": 4.4432583506395204e-05, "loss": 0.0009, "step": 17216 }, { "epoch": 0.33, "learning_rate": 4.443193654613797e-05, "loss": 0.002, "step": 17218 }, { "epoch": 0.33, "learning_rate": 4.443128958588074e-05, "loss": 0.0288, "step": 17220 }, { "epoch": 0.33, "learning_rate": 4.443064262562351e-05, "loss": 0.0051, "step": 17222 }, { "epoch": 0.33, "learning_rate": 4.442999566536628e-05, "loss": 0.0012, "step": 17224 }, { "epoch": 0.33, "learning_rate": 4.442934870510905e-05, "loss": 0.0025, "step": 17226 }, { "epoch": 0.33, "learning_rate": 4.442870174485181e-05, "loss": 0.0002, "step": 17228 }, { "epoch": 0.33, "learning_rate": 4.442805478459459e-05, "loss": 0.0009, "step": 17230 }, { "epoch": 0.33, "learning_rate": 4.442740782433735e-05, "loss": 0.0075, "step": 17232 }, { "epoch": 0.33, "learning_rate": 4.442676086408012e-05, "loss": 0.005, "step": 17234 }, { "epoch": 0.33, "learning_rate": 4.442611390382289e-05, "loss": 0.0006, "step": 17236 }, { "epoch": 0.33, "learning_rate": 4.442546694356566e-05, "loss": 0.0001, "step": 17238 }, { "epoch": 0.33, "learning_rate": 4.442481998330843e-05, "loss": 0.0075, "step": 17240 }, { "epoch": 0.33, "learning_rate": 4.4424173023051195e-05, "loss": 0.0102, "step": 17242 }, { "epoch": 0.33, "learning_rate": 4.4423526062793964e-05, "loss": 0.0012, "step": 17244 }, { "epoch": 0.33, "learning_rate": 4.4422879102536734e-05, "loss": 0.0002, "step": 17246 }, { "epoch": 0.33, "learning_rate": 4.44222321422795e-05, "loss": 0.0166, "step": 17248 }, { "epoch": 0.33, "learning_rate": 4.442158518202227e-05, "loss": 0.0055, "step": 17250 }, { "epoch": 0.33, "learning_rate": 4.442093822176504e-05, "loss": 0.0046, "step": 17252 }, { "epoch": 0.33, "learning_rate": 4.442029126150781e-05, "loss": 0.0096, "step": 17254 }, { "epoch": 0.33, "learning_rate": 4.441964430125057e-05, "loss": 0.0021, "step": 17256 }, { "epoch": 0.33, "learning_rate": 4.441899734099335e-05, "loss": 0.0358, "step": 17258 }, { "epoch": 0.33, "learning_rate": 4.441835038073611e-05, "loss": 0.0032, "step": 17260 }, { "epoch": 0.34, "learning_rate": 4.4417703420478886e-05, "loss": 0.0026, "step": 17262 }, { "epoch": 0.34, "learning_rate": 4.441705646022165e-05, "loss": 0.0007, "step": 17264 }, { "epoch": 0.34, "learning_rate": 4.441640949996442e-05, "loss": 0.006, "step": 17266 }, { "epoch": 0.34, "learning_rate": 4.441576253970719e-05, "loss": 0.0032, "step": 17268 }, { "epoch": 0.34, "learning_rate": 4.4415115579449956e-05, "loss": 0.004, "step": 17270 }, { "epoch": 0.34, "learning_rate": 4.4414468619192725e-05, "loss": 0.0061, "step": 17272 }, { "epoch": 0.34, "learning_rate": 4.4413821658935494e-05, "loss": 0.0032, "step": 17274 }, { "epoch": 0.34, "learning_rate": 4.441317469867826e-05, "loss": 0.0002, "step": 17276 }, { "epoch": 0.34, "learning_rate": 4.441252773842103e-05, "loss": 0.0001, "step": 17278 }, { "epoch": 0.34, "learning_rate": 4.44118807781638e-05, "loss": 0.0016, "step": 17280 }, { "epoch": 0.34, "learning_rate": 4.4411233817906564e-05, "loss": 0.0002, "step": 17282 }, { "epoch": 0.34, "learning_rate": 4.441058685764934e-05, "loss": 0.0244, "step": 17284 }, { "epoch": 0.34, "learning_rate": 4.440993989739211e-05, "loss": 0.0001, "step": 17286 }, { "epoch": 0.34, "learning_rate": 4.440929293713487e-05, "loss": 0.0003, "step": 17288 }, { "epoch": 0.34, "learning_rate": 4.440864597687765e-05, "loss": 0.0033, "step": 17290 }, { "epoch": 0.34, "learning_rate": 4.440799901662041e-05, "loss": 0.0052, "step": 17292 }, { "epoch": 0.34, "learning_rate": 4.440735205636318e-05, "loss": 0.0121, "step": 17294 }, { "epoch": 0.34, "learning_rate": 4.440670509610595e-05, "loss": 0.0026, "step": 17296 }, { "epoch": 0.34, "learning_rate": 4.4406058135848716e-05, "loss": 0.0056, "step": 17298 }, { "epoch": 0.34, "learning_rate": 4.4405411175591485e-05, "loss": 0.0081, "step": 17300 }, { "epoch": 0.34, "learning_rate": 4.4404764215334254e-05, "loss": 0.0214, "step": 17302 }, { "epoch": 0.34, "learning_rate": 4.4404117255077023e-05, "loss": 0.0012, "step": 17304 }, { "epoch": 0.34, "learning_rate": 4.440347029481979e-05, "loss": 0.0002, "step": 17306 }, { "epoch": 0.34, "learning_rate": 4.440282333456256e-05, "loss": 0.0007, "step": 17308 }, { "epoch": 0.34, "learning_rate": 4.4402176374305324e-05, "loss": 0.0084, "step": 17310 }, { "epoch": 0.34, "learning_rate": 4.44015294140481e-05, "loss": 0.0009, "step": 17312 }, { "epoch": 0.34, "learning_rate": 4.440088245379086e-05, "loss": 0.0025, "step": 17314 }, { "epoch": 0.34, "learning_rate": 4.440023549353363e-05, "loss": 0.0001, "step": 17316 }, { "epoch": 0.34, "learning_rate": 4.439958853327641e-05, "loss": 0.0002, "step": 17318 }, { "epoch": 0.34, "learning_rate": 4.439894157301917e-05, "loss": 0.0012, "step": 17320 }, { "epoch": 0.34, "learning_rate": 4.4398294612761945e-05, "loss": 0.0101, "step": 17322 }, { "epoch": 0.34, "learning_rate": 4.439764765250471e-05, "loss": 0.0045, "step": 17324 }, { "epoch": 0.34, "learning_rate": 4.439700069224748e-05, "loss": 0.0072, "step": 17326 }, { "epoch": 0.34, "learning_rate": 4.4396353731990246e-05, "loss": 0.0097, "step": 17328 }, { "epoch": 0.34, "learning_rate": 4.4395706771733015e-05, "loss": 0.0063, "step": 17330 }, { "epoch": 0.34, "learning_rate": 4.4395059811475784e-05, "loss": 0.0082, "step": 17332 }, { "epoch": 0.34, "learning_rate": 4.439441285121855e-05, "loss": 0.0063, "step": 17334 }, { "epoch": 0.34, "learning_rate": 4.439376589096132e-05, "loss": 0.0061, "step": 17336 }, { "epoch": 0.34, "learning_rate": 4.4393118930704084e-05, "loss": 0.0001, "step": 17338 }, { "epoch": 0.34, "learning_rate": 4.439247197044686e-05, "loss": 0.009, "step": 17340 }, { "epoch": 0.34, "learning_rate": 4.439182501018962e-05, "loss": 0.0031, "step": 17342 }, { "epoch": 0.34, "learning_rate": 4.43911780499324e-05, "loss": 0.002, "step": 17344 }, { "epoch": 0.34, "learning_rate": 4.439053108967516e-05, "loss": 0.0004, "step": 17346 }, { "epoch": 0.34, "learning_rate": 4.438988412941793e-05, "loss": 0.0012, "step": 17348 }, { "epoch": 0.34, "learning_rate": 4.4389237169160706e-05, "loss": 0.0064, "step": 17350 }, { "epoch": 0.34, "learning_rate": 4.438859020890347e-05, "loss": 0.0001, "step": 17352 }, { "epoch": 0.34, "learning_rate": 4.438794324864624e-05, "loss": 0.0137, "step": 17354 }, { "epoch": 0.34, "learning_rate": 4.4387296288389006e-05, "loss": 0.0016, "step": 17356 }, { "epoch": 0.34, "learning_rate": 4.4386649328131775e-05, "loss": 0.0005, "step": 17358 }, { "epoch": 0.34, "learning_rate": 4.4386002367874544e-05, "loss": 0.004, "step": 17360 }, { "epoch": 0.34, "learning_rate": 4.438535540761731e-05, "loss": 0.0041, "step": 17362 }, { "epoch": 0.34, "learning_rate": 4.438470844736008e-05, "loss": 0.0014, "step": 17364 }, { "epoch": 0.34, "learning_rate": 4.438406148710285e-05, "loss": 0.0004, "step": 17366 }, { "epoch": 0.34, "learning_rate": 4.438341452684562e-05, "loss": 0.0023, "step": 17368 }, { "epoch": 0.34, "learning_rate": 4.438276756658838e-05, "loss": 0.0081, "step": 17370 }, { "epoch": 0.34, "learning_rate": 4.438212060633116e-05, "loss": 0.0002, "step": 17372 }, { "epoch": 0.34, "learning_rate": 4.438147364607392e-05, "loss": 0.0004, "step": 17374 }, { "epoch": 0.34, "learning_rate": 4.438082668581669e-05, "loss": 0.0006, "step": 17376 }, { "epoch": 0.34, "learning_rate": 4.438017972555946e-05, "loss": 0.0059, "step": 17378 }, { "epoch": 0.34, "learning_rate": 4.437953276530223e-05, "loss": 0.0088, "step": 17380 }, { "epoch": 0.34, "learning_rate": 4.4378885805045e-05, "loss": 0.0007, "step": 17382 }, { "epoch": 0.34, "learning_rate": 4.4378238844787767e-05, "loss": 0.0023, "step": 17384 }, { "epoch": 0.34, "learning_rate": 4.4377591884530536e-05, "loss": 0.0008, "step": 17386 }, { "epoch": 0.34, "learning_rate": 4.4376944924273305e-05, "loss": 0.0007, "step": 17388 }, { "epoch": 0.34, "learning_rate": 4.4376297964016074e-05, "loss": 0.0003, "step": 17390 }, { "epoch": 0.34, "learning_rate": 4.4375651003758836e-05, "loss": 0.0001, "step": 17392 }, { "epoch": 0.34, "learning_rate": 4.437500404350161e-05, "loss": 0.0018, "step": 17394 }, { "epoch": 0.34, "learning_rate": 4.437435708324438e-05, "loss": 0.0127, "step": 17396 }, { "epoch": 0.34, "learning_rate": 4.437371012298714e-05, "loss": 0.0032, "step": 17398 }, { "epoch": 0.34, "learning_rate": 4.437306316272992e-05, "loss": 0.0038, "step": 17400 }, { "epoch": 0.34, "learning_rate": 4.437241620247268e-05, "loss": 0.0018, "step": 17402 }, { "epoch": 0.34, "learning_rate": 4.437176924221546e-05, "loss": 0.0009, "step": 17404 }, { "epoch": 0.34, "learning_rate": 4.437112228195822e-05, "loss": 0.0075, "step": 17406 }, { "epoch": 0.34, "learning_rate": 4.437047532170099e-05, "loss": 0.0004, "step": 17408 }, { "epoch": 0.34, "learning_rate": 4.436982836144376e-05, "loss": 0.0004, "step": 17410 }, { "epoch": 0.34, "learning_rate": 4.436918140118653e-05, "loss": 0.0008, "step": 17412 }, { "epoch": 0.34, "learning_rate": 4.4368534440929296e-05, "loss": 0.0001, "step": 17414 }, { "epoch": 0.34, "learning_rate": 4.4367887480672065e-05, "loss": 0.0009, "step": 17416 }, { "epoch": 0.34, "learning_rate": 4.4367240520414834e-05, "loss": 0.0002, "step": 17418 }, { "epoch": 0.34, "learning_rate": 4.43665935601576e-05, "loss": 0.0071, "step": 17420 }, { "epoch": 0.34, "learning_rate": 4.436594659990037e-05, "loss": 0.0056, "step": 17422 }, { "epoch": 0.34, "learning_rate": 4.4365299639643135e-05, "loss": 0.0036, "step": 17424 }, { "epoch": 0.34, "learning_rate": 4.436465267938591e-05, "loss": 0.0185, "step": 17426 }, { "epoch": 0.34, "learning_rate": 4.436400571912867e-05, "loss": 0.0025, "step": 17428 }, { "epoch": 0.34, "learning_rate": 4.436335875887144e-05, "loss": 0.0043, "step": 17430 }, { "epoch": 0.34, "learning_rate": 4.436271179861422e-05, "loss": 0.0024, "step": 17432 }, { "epoch": 0.34, "learning_rate": 4.436206483835698e-05, "loss": 0.0128, "step": 17434 }, { "epoch": 0.34, "learning_rate": 4.436141787809975e-05, "loss": 0.0002, "step": 17436 }, { "epoch": 0.34, "learning_rate": 4.436077091784252e-05, "loss": 0.0006, "step": 17438 }, { "epoch": 0.34, "learning_rate": 4.436012395758529e-05, "loss": 0.0016, "step": 17440 }, { "epoch": 0.34, "learning_rate": 4.4359476997328056e-05, "loss": 0.0032, "step": 17442 }, { "epoch": 0.34, "learning_rate": 4.4358830037070825e-05, "loss": 0.0014, "step": 17444 }, { "epoch": 0.34, "learning_rate": 4.4358183076813595e-05, "loss": 0.0003, "step": 17446 }, { "epoch": 0.34, "learning_rate": 4.4357536116556364e-05, "loss": 0.0026, "step": 17448 }, { "epoch": 0.34, "learning_rate": 4.435688915629913e-05, "loss": 0.0016, "step": 17450 }, { "epoch": 0.34, "learning_rate": 4.4356242196041895e-05, "loss": 0.0101, "step": 17452 }, { "epoch": 0.34, "learning_rate": 4.435559523578467e-05, "loss": 0.0057, "step": 17454 }, { "epoch": 0.34, "learning_rate": 4.435494827552743e-05, "loss": 0.0125, "step": 17456 }, { "epoch": 0.34, "learning_rate": 4.43543013152702e-05, "loss": 0.0006, "step": 17458 }, { "epoch": 0.34, "learning_rate": 4.435365435501297e-05, "loss": 0.0009, "step": 17460 }, { "epoch": 0.34, "learning_rate": 4.435300739475574e-05, "loss": 0.0003, "step": 17462 }, { "epoch": 0.34, "learning_rate": 4.4352360434498516e-05, "loss": 0.0009, "step": 17464 }, { "epoch": 0.34, "learning_rate": 4.435171347424128e-05, "loss": 0.0065, "step": 17466 }, { "epoch": 0.34, "learning_rate": 4.435106651398405e-05, "loss": 0.017, "step": 17468 }, { "epoch": 0.34, "learning_rate": 4.435041955372682e-05, "loss": 0.0079, "step": 17470 }, { "epoch": 0.34, "learning_rate": 4.4349772593469586e-05, "loss": 0.0015, "step": 17472 }, { "epoch": 0.34, "learning_rate": 4.4349125633212355e-05, "loss": 0.0103, "step": 17474 }, { "epoch": 0.34, "learning_rate": 4.4348478672955124e-05, "loss": 0.0104, "step": 17476 }, { "epoch": 0.34, "learning_rate": 4.434783171269789e-05, "loss": 0.0008, "step": 17478 }, { "epoch": 0.34, "learning_rate": 4.4347184752440655e-05, "loss": 0.0056, "step": 17480 }, { "epoch": 0.34, "learning_rate": 4.434653779218343e-05, "loss": 0.0004, "step": 17482 }, { "epoch": 0.34, "learning_rate": 4.4345890831926194e-05, "loss": 0.0007, "step": 17484 }, { "epoch": 0.34, "learning_rate": 4.434524387166897e-05, "loss": 0.0158, "step": 17486 }, { "epoch": 0.34, "learning_rate": 4.434459691141173e-05, "loss": 0.0003, "step": 17488 }, { "epoch": 0.34, "learning_rate": 4.43439499511545e-05, "loss": 0.0032, "step": 17490 }, { "epoch": 0.34, "learning_rate": 4.434330299089727e-05, "loss": 0.0001, "step": 17492 }, { "epoch": 0.34, "learning_rate": 4.434265603064004e-05, "loss": 0.008, "step": 17494 }, { "epoch": 0.34, "learning_rate": 4.434200907038281e-05, "loss": 0.0017, "step": 17496 }, { "epoch": 0.34, "learning_rate": 4.434136211012558e-05, "loss": 0.0021, "step": 17498 }, { "epoch": 0.34, "learning_rate": 4.4340715149868346e-05, "loss": 0.0009, "step": 17500 }, { "epoch": 0.34, "learning_rate": 4.4340068189611115e-05, "loss": 0.0244, "step": 17502 }, { "epoch": 0.34, "learning_rate": 4.4339421229353884e-05, "loss": 0.0059, "step": 17504 }, { "epoch": 0.34, "learning_rate": 4.433877426909665e-05, "loss": 0.0082, "step": 17506 }, { "epoch": 0.34, "learning_rate": 4.433812730883942e-05, "loss": 0.0001, "step": 17508 }, { "epoch": 0.34, "learning_rate": 4.433748034858219e-05, "loss": 0.0024, "step": 17510 }, { "epoch": 0.34, "learning_rate": 4.4336833388324954e-05, "loss": 0.0005, "step": 17512 }, { "epoch": 0.34, "learning_rate": 4.433618642806773e-05, "loss": 0.0001, "step": 17514 }, { "epoch": 0.34, "learning_rate": 4.433553946781049e-05, "loss": 0.0003, "step": 17516 }, { "epoch": 0.34, "learning_rate": 4.433489250755326e-05, "loss": 0.0003, "step": 17518 }, { "epoch": 0.34, "learning_rate": 4.433424554729603e-05, "loss": 0.0024, "step": 17520 }, { "epoch": 0.34, "learning_rate": 4.43335985870388e-05, "loss": 0.0002, "step": 17522 }, { "epoch": 0.34, "learning_rate": 4.433295162678157e-05, "loss": 0.0002, "step": 17524 }, { "epoch": 0.34, "learning_rate": 4.433230466652434e-05, "loss": 0.0003, "step": 17526 }, { "epoch": 0.34, "learning_rate": 4.433165770626711e-05, "loss": 0.006, "step": 17528 }, { "epoch": 0.34, "learning_rate": 4.4331010746009876e-05, "loss": 0.0345, "step": 17530 }, { "epoch": 0.34, "learning_rate": 4.4330363785752645e-05, "loss": 0.0007, "step": 17532 }, { "epoch": 0.34, "learning_rate": 4.432971682549541e-05, "loss": 0.0049, "step": 17534 }, { "epoch": 0.34, "learning_rate": 4.432906986523818e-05, "loss": 0.0026, "step": 17536 }, { "epoch": 0.34, "learning_rate": 4.4328422904980945e-05, "loss": 0.0263, "step": 17538 }, { "epoch": 0.34, "learning_rate": 4.4327775944723714e-05, "loss": 0.0001, "step": 17540 }, { "epoch": 0.34, "learning_rate": 4.432712898446649e-05, "loss": 0.0205, "step": 17542 }, { "epoch": 0.34, "learning_rate": 4.432648202420925e-05, "loss": 0.0047, "step": 17544 }, { "epoch": 0.34, "learning_rate": 4.432583506395203e-05, "loss": 0.0189, "step": 17546 }, { "epoch": 0.34, "learning_rate": 4.432518810369479e-05, "loss": 0.003, "step": 17548 }, { "epoch": 0.34, "learning_rate": 4.432454114343756e-05, "loss": 0.0002, "step": 17550 }, { "epoch": 0.34, "learning_rate": 4.432389418318033e-05, "loss": 0.0001, "step": 17552 }, { "epoch": 0.34, "learning_rate": 4.43232472229231e-05, "loss": 0.0006, "step": 17554 }, { "epoch": 0.34, "learning_rate": 4.432260026266587e-05, "loss": 0.0134, "step": 17556 }, { "epoch": 0.34, "learning_rate": 4.4321953302408636e-05, "loss": 0.0658, "step": 17558 }, { "epoch": 0.34, "learning_rate": 4.4321306342151405e-05, "loss": 0.0042, "step": 17560 }, { "epoch": 0.34, "learning_rate": 4.4320659381894174e-05, "loss": 0.0068, "step": 17562 }, { "epoch": 0.34, "learning_rate": 4.4320012421636943e-05, "loss": 0.001, "step": 17564 }, { "epoch": 0.34, "learning_rate": 4.4319365461379706e-05, "loss": 0.0006, "step": 17566 }, { "epoch": 0.34, "learning_rate": 4.431871850112248e-05, "loss": 0.0014, "step": 17568 }, { "epoch": 0.34, "learning_rate": 4.4318071540865244e-05, "loss": 0.005, "step": 17570 }, { "epoch": 0.34, "learning_rate": 4.431742458060801e-05, "loss": 0.0029, "step": 17572 }, { "epoch": 0.34, "learning_rate": 4.431677762035079e-05, "loss": 0.0075, "step": 17574 }, { "epoch": 0.34, "learning_rate": 4.431613066009355e-05, "loss": 0.0001, "step": 17576 }, { "epoch": 0.34, "learning_rate": 4.431548369983632e-05, "loss": 0.0253, "step": 17578 }, { "epoch": 0.34, "learning_rate": 4.431483673957909e-05, "loss": 0.0004, "step": 17580 }, { "epoch": 0.34, "learning_rate": 4.431418977932186e-05, "loss": 0.0002, "step": 17582 }, { "epoch": 0.34, "learning_rate": 4.431354281906463e-05, "loss": 0.003, "step": 17584 }, { "epoch": 0.34, "learning_rate": 4.4312895858807397e-05, "loss": 0.0036, "step": 17586 }, { "epoch": 0.34, "learning_rate": 4.4312248898550166e-05, "loss": 0.002, "step": 17588 }, { "epoch": 0.34, "learning_rate": 4.4311601938292935e-05, "loss": 0.0053, "step": 17590 }, { "epoch": 0.34, "learning_rate": 4.4310954978035704e-05, "loss": 0.0001, "step": 17592 }, { "epoch": 0.34, "learning_rate": 4.4310308017778466e-05, "loss": 0.0009, "step": 17594 }, { "epoch": 0.34, "learning_rate": 4.430966105752124e-05, "loss": 0.0032, "step": 17596 }, { "epoch": 0.34, "learning_rate": 4.4309014097264004e-05, "loss": 0.0002, "step": 17598 }, { "epoch": 0.34, "learning_rate": 4.4308367137006773e-05, "loss": 0.0018, "step": 17600 }, { "epoch": 0.34, "learning_rate": 4.430772017674954e-05, "loss": 0.0001, "step": 17602 }, { "epoch": 0.34, "learning_rate": 4.430707321649231e-05, "loss": 0.0001, "step": 17604 }, { "epoch": 0.34, "learning_rate": 4.430642625623508e-05, "loss": 0.0037, "step": 17606 }, { "epoch": 0.34, "learning_rate": 4.430577929597785e-05, "loss": 0.0027, "step": 17608 }, { "epoch": 0.34, "learning_rate": 4.430513233572062e-05, "loss": 0.0032, "step": 17610 }, { "epoch": 0.34, "learning_rate": 4.430448537546339e-05, "loss": 0.0003, "step": 17612 }, { "epoch": 0.34, "learning_rate": 4.430383841520616e-05, "loss": 0.0004, "step": 17614 }, { "epoch": 0.34, "learning_rate": 4.430319145494892e-05, "loss": 0.0002, "step": 17616 }, { "epoch": 0.34, "learning_rate": 4.4302544494691695e-05, "loss": 0.0133, "step": 17618 }, { "epoch": 0.34, "learning_rate": 4.4301897534434464e-05, "loss": 0.0004, "step": 17620 }, { "epoch": 0.34, "learning_rate": 4.4301250574177227e-05, "loss": 0.0156, "step": 17622 }, { "epoch": 0.34, "learning_rate": 4.430060361392e-05, "loss": 0.0003, "step": 17624 }, { "epoch": 0.34, "learning_rate": 4.4299956653662765e-05, "loss": 0.0004, "step": 17626 }, { "epoch": 0.34, "learning_rate": 4.429930969340554e-05, "loss": 0.0292, "step": 17628 }, { "epoch": 0.34, "learning_rate": 4.42986627331483e-05, "loss": 0.0003, "step": 17630 }, { "epoch": 0.34, "learning_rate": 4.429801577289107e-05, "loss": 0.001, "step": 17632 }, { "epoch": 0.34, "learning_rate": 4.429736881263384e-05, "loss": 0.0004, "step": 17634 }, { "epoch": 0.34, "learning_rate": 4.429672185237661e-05, "loss": 0.0008, "step": 17636 }, { "epoch": 0.34, "learning_rate": 4.429607489211938e-05, "loss": 0.0029, "step": 17638 }, { "epoch": 0.34, "learning_rate": 4.429542793186215e-05, "loss": 0.0041, "step": 17640 }, { "epoch": 0.34, "learning_rate": 4.429478097160492e-05, "loss": 0.0011, "step": 17642 }, { "epoch": 0.34, "learning_rate": 4.4294134011347686e-05, "loss": 0.0002, "step": 17644 }, { "epoch": 0.34, "learning_rate": 4.4293487051090456e-05, "loss": 0.0033, "step": 17646 }, { "epoch": 0.34, "learning_rate": 4.429284009083322e-05, "loss": 0.003, "step": 17648 }, { "epoch": 0.34, "learning_rate": 4.4292193130575994e-05, "loss": 0.0054, "step": 17650 }, { "epoch": 0.34, "learning_rate": 4.429154617031876e-05, "loss": 0.0215, "step": 17652 }, { "epoch": 0.34, "learning_rate": 4.4290899210061525e-05, "loss": 0.0069, "step": 17654 }, { "epoch": 0.34, "learning_rate": 4.42902522498043e-05, "loss": 0.0002, "step": 17656 }, { "epoch": 0.34, "learning_rate": 4.428960528954706e-05, "loss": 0.0133, "step": 17658 }, { "epoch": 0.34, "learning_rate": 4.428895832928983e-05, "loss": 0.0002, "step": 17660 }, { "epoch": 0.34, "learning_rate": 4.42883113690326e-05, "loss": 0.0067, "step": 17662 }, { "epoch": 0.34, "learning_rate": 4.428766440877537e-05, "loss": 0.0032, "step": 17664 }, { "epoch": 0.34, "learning_rate": 4.428701744851814e-05, "loss": 0.0092, "step": 17666 }, { "epoch": 0.34, "learning_rate": 4.428637048826091e-05, "loss": 0.0126, "step": 17668 }, { "epoch": 0.34, "learning_rate": 4.428572352800368e-05, "loss": 0.0026, "step": 17670 }, { "epoch": 0.34, "learning_rate": 4.428507656774645e-05, "loss": 0.0016, "step": 17672 }, { "epoch": 0.34, "learning_rate": 4.4284429607489216e-05, "loss": 0.0018, "step": 17674 }, { "epoch": 0.34, "learning_rate": 4.428378264723198e-05, "loss": 0.0014, "step": 17676 }, { "epoch": 0.34, "learning_rate": 4.4283135686974754e-05, "loss": 0.0003, "step": 17678 }, { "epoch": 0.34, "learning_rate": 4.4282488726717516e-05, "loss": 0.0073, "step": 17680 }, { "epoch": 0.34, "learning_rate": 4.4281841766460286e-05, "loss": 0.0012, "step": 17682 }, { "epoch": 0.34, "learning_rate": 4.4281194806203055e-05, "loss": 0.0075, "step": 17684 }, { "epoch": 0.34, "learning_rate": 4.4280547845945824e-05, "loss": 0.0022, "step": 17686 }, { "epoch": 0.34, "learning_rate": 4.42799008856886e-05, "loss": 0.0003, "step": 17688 }, { "epoch": 0.34, "learning_rate": 4.427925392543136e-05, "loss": 0.0008, "step": 17690 }, { "epoch": 0.34, "learning_rate": 4.427860696517413e-05, "loss": 0.0022, "step": 17692 }, { "epoch": 0.34, "learning_rate": 4.42779600049169e-05, "loss": 0.0005, "step": 17694 }, { "epoch": 0.34, "learning_rate": 4.427731304465967e-05, "loss": 0.002, "step": 17696 }, { "epoch": 0.34, "learning_rate": 4.427666608440244e-05, "loss": 0.0002, "step": 17698 }, { "epoch": 0.34, "learning_rate": 4.427601912414521e-05, "loss": 0.0008, "step": 17700 }, { "epoch": 0.34, "learning_rate": 4.4275372163887976e-05, "loss": 0.0129, "step": 17702 }, { "epoch": 0.34, "learning_rate": 4.4274725203630745e-05, "loss": 0.0007, "step": 17704 }, { "epoch": 0.34, "learning_rate": 4.4274078243373515e-05, "loss": 0.001, "step": 17706 }, { "epoch": 0.34, "learning_rate": 4.427343128311628e-05, "loss": 0.0001, "step": 17708 }, { "epoch": 0.34, "learning_rate": 4.427278432285905e-05, "loss": 0.0002, "step": 17710 }, { "epoch": 0.34, "learning_rate": 4.4272137362601815e-05, "loss": 0.0025, "step": 17712 }, { "epoch": 0.34, "learning_rate": 4.4271490402344584e-05, "loss": 0.0001, "step": 17714 }, { "epoch": 0.34, "learning_rate": 4.427084344208735e-05, "loss": 0.0243, "step": 17716 }, { "epoch": 0.34, "learning_rate": 4.427019648183012e-05, "loss": 0.0077, "step": 17718 }, { "epoch": 0.34, "learning_rate": 4.426954952157289e-05, "loss": 0.0088, "step": 17720 }, { "epoch": 0.34, "learning_rate": 4.426890256131566e-05, "loss": 0.0001, "step": 17722 }, { "epoch": 0.34, "learning_rate": 4.426825560105843e-05, "loss": 0.0043, "step": 17724 }, { "epoch": 0.34, "learning_rate": 4.42676086408012e-05, "loss": 0.0133, "step": 17726 }, { "epoch": 0.34, "learning_rate": 4.426696168054397e-05, "loss": 0.001, "step": 17728 }, { "epoch": 0.34, "learning_rate": 4.426631472028674e-05, "loss": 0.0001, "step": 17730 }, { "epoch": 0.34, "learning_rate": 4.4265667760029506e-05, "loss": 0.0092, "step": 17732 }, { "epoch": 0.34, "learning_rate": 4.4265020799772275e-05, "loss": 0.0177, "step": 17734 }, { "epoch": 0.34, "learning_rate": 4.426437383951504e-05, "loss": 0.0001, "step": 17736 }, { "epoch": 0.34, "learning_rate": 4.426372687925781e-05, "loss": 0.0005, "step": 17738 }, { "epoch": 0.34, "learning_rate": 4.4263079919000575e-05, "loss": 0.0011, "step": 17740 }, { "epoch": 0.34, "learning_rate": 4.4262432958743345e-05, "loss": 0.0004, "step": 17742 }, { "epoch": 0.34, "learning_rate": 4.4261785998486114e-05, "loss": 0.0001, "step": 17744 }, { "epoch": 0.34, "learning_rate": 4.426113903822888e-05, "loss": 0.0005, "step": 17746 }, { "epoch": 0.34, "learning_rate": 4.426049207797165e-05, "loss": 0.0012, "step": 17748 }, { "epoch": 0.34, "learning_rate": 4.425984511771442e-05, "loss": 0.0007, "step": 17750 }, { "epoch": 0.34, "learning_rate": 4.425919815745719e-05, "loss": 0.0184, "step": 17752 }, { "epoch": 0.34, "learning_rate": 4.425855119719996e-05, "loss": 0.0004, "step": 17754 }, { "epoch": 0.34, "learning_rate": 4.425790423694273e-05, "loss": 0.0066, "step": 17756 }, { "epoch": 0.34, "learning_rate": 4.425725727668549e-05, "loss": 0.0002, "step": 17758 }, { "epoch": 0.34, "learning_rate": 4.4256610316428266e-05, "loss": 0.01, "step": 17760 }, { "epoch": 0.34, "learning_rate": 4.425596335617103e-05, "loss": 0.0007, "step": 17762 }, { "epoch": 0.34, "learning_rate": 4.42553163959138e-05, "loss": 0.0176, "step": 17764 }, { "epoch": 0.34, "learning_rate": 4.4254669435656574e-05, "loss": 0.0016, "step": 17766 }, { "epoch": 0.34, "learning_rate": 4.4254022475399336e-05, "loss": 0.012, "step": 17768 }, { "epoch": 0.34, "learning_rate": 4.425337551514211e-05, "loss": 0.0004, "step": 17770 }, { "epoch": 0.34, "learning_rate": 4.4252728554884874e-05, "loss": 0.0004, "step": 17772 }, { "epoch": 0.34, "learning_rate": 4.425208159462764e-05, "loss": 0.0114, "step": 17774 }, { "epoch": 0.35, "learning_rate": 4.425143463437041e-05, "loss": 0.0079, "step": 17776 }, { "epoch": 0.35, "learning_rate": 4.425078767411318e-05, "loss": 0.0003, "step": 17778 }, { "epoch": 0.35, "learning_rate": 4.425014071385595e-05, "loss": 0.0005, "step": 17780 }, { "epoch": 0.35, "learning_rate": 4.424949375359872e-05, "loss": 0.0098, "step": 17782 }, { "epoch": 0.35, "learning_rate": 4.424884679334149e-05, "loss": 0.0002, "step": 17784 }, { "epoch": 0.35, "learning_rate": 4.424819983308426e-05, "loss": 0.0044, "step": 17786 }, { "epoch": 0.35, "learning_rate": 4.424755287282703e-05, "loss": 0.0081, "step": 17788 }, { "epoch": 0.35, "learning_rate": 4.424690591256979e-05, "loss": 0.0004, "step": 17790 }, { "epoch": 0.35, "learning_rate": 4.4246258952312565e-05, "loss": 0.0002, "step": 17792 }, { "epoch": 0.35, "learning_rate": 4.424561199205533e-05, "loss": 0.0055, "step": 17794 }, { "epoch": 0.35, "learning_rate": 4.4244965031798096e-05, "loss": 0.0017, "step": 17796 }, { "epoch": 0.35, "learning_rate": 4.424431807154087e-05, "loss": 0.002, "step": 17798 }, { "epoch": 0.35, "learning_rate": 4.4243671111283634e-05, "loss": 0.0024, "step": 17800 }, { "epoch": 0.35, "learning_rate": 4.4243024151026404e-05, "loss": 0.0107, "step": 17802 }, { "epoch": 0.35, "learning_rate": 4.424237719076917e-05, "loss": 0.0081, "step": 17804 }, { "epoch": 0.35, "learning_rate": 4.424173023051194e-05, "loss": 0.0004, "step": 17806 }, { "epoch": 0.35, "learning_rate": 4.424108327025471e-05, "loss": 0.0006, "step": 17808 }, { "epoch": 0.35, "learning_rate": 4.424043630999748e-05, "loss": 0.0003, "step": 17810 }, { "epoch": 0.35, "learning_rate": 4.423978934974025e-05, "loss": 0.0125, "step": 17812 }, { "epoch": 0.35, "learning_rate": 4.423914238948302e-05, "loss": 0.032, "step": 17814 }, { "epoch": 0.35, "learning_rate": 4.423849542922579e-05, "loss": 0.0004, "step": 17816 }, { "epoch": 0.35, "learning_rate": 4.423784846896855e-05, "loss": 0.0051, "step": 17818 }, { "epoch": 0.35, "learning_rate": 4.4237201508711325e-05, "loss": 0.0002, "step": 17820 }, { "epoch": 0.35, "learning_rate": 4.423655454845409e-05, "loss": 0.0007, "step": 17822 }, { "epoch": 0.35, "learning_rate": 4.423590758819686e-05, "loss": 0.0024, "step": 17824 }, { "epoch": 0.35, "learning_rate": 4.4235260627939626e-05, "loss": 0.0206, "step": 17826 }, { "epoch": 0.35, "learning_rate": 4.4234613667682395e-05, "loss": 0.0315, "step": 17828 }, { "epoch": 0.35, "learning_rate": 4.423396670742517e-05, "loss": 0.0026, "step": 17830 }, { "epoch": 0.35, "learning_rate": 4.423331974716793e-05, "loss": 0.0001, "step": 17832 }, { "epoch": 0.35, "learning_rate": 4.42326727869107e-05, "loss": 0.0016, "step": 17834 }, { "epoch": 0.35, "learning_rate": 4.423202582665347e-05, "loss": 0.0004, "step": 17836 }, { "epoch": 0.35, "learning_rate": 4.423137886639624e-05, "loss": 0.0013, "step": 17838 }, { "epoch": 0.35, "learning_rate": 4.4230731906139e-05, "loss": 0.0012, "step": 17840 }, { "epoch": 0.35, "learning_rate": 4.423008494588178e-05, "loss": 0.0131, "step": 17842 }, { "epoch": 0.35, "learning_rate": 4.422943798562455e-05, "loss": 0.0079, "step": 17844 }, { "epoch": 0.35, "learning_rate": 4.4228791025367317e-05, "loss": 0.0001, "step": 17846 }, { "epoch": 0.35, "learning_rate": 4.4228144065110086e-05, "loss": 0.0019, "step": 17848 }, { "epoch": 0.35, "learning_rate": 4.422749710485285e-05, "loss": 0.0005, "step": 17850 }, { "epoch": 0.35, "learning_rate": 4.4226850144595624e-05, "loss": 0.0013, "step": 17852 }, { "epoch": 0.35, "learning_rate": 4.4226203184338386e-05, "loss": 0.0104, "step": 17854 }, { "epoch": 0.35, "learning_rate": 4.4225556224081155e-05, "loss": 0.0003, "step": 17856 }, { "epoch": 0.35, "learning_rate": 4.4224909263823924e-05, "loss": 0.0004, "step": 17858 }, { "epoch": 0.35, "learning_rate": 4.4224262303566693e-05, "loss": 0.008, "step": 17860 }, { "epoch": 0.35, "learning_rate": 4.422361534330946e-05, "loss": 0.005, "step": 17862 }, { "epoch": 0.35, "learning_rate": 4.422296838305223e-05, "loss": 0.0003, "step": 17864 }, { "epoch": 0.35, "learning_rate": 4.4222321422795e-05, "loss": 0.0002, "step": 17866 }, { "epoch": 0.35, "learning_rate": 4.422167446253777e-05, "loss": 0.003, "step": 17868 }, { "epoch": 0.35, "learning_rate": 4.422102750228054e-05, "loss": 0.0025, "step": 17870 }, { "epoch": 0.35, "learning_rate": 4.42203805420233e-05, "loss": 0.0007, "step": 17872 }, { "epoch": 0.35, "learning_rate": 4.421973358176608e-05, "loss": 0.0038, "step": 17874 }, { "epoch": 0.35, "learning_rate": 4.4219086621508846e-05, "loss": 0.0001, "step": 17876 }, { "epoch": 0.35, "learning_rate": 4.421843966125161e-05, "loss": 0.0006, "step": 17878 }, { "epoch": 0.35, "learning_rate": 4.4217792700994384e-05, "loss": 0.0117, "step": 17880 }, { "epoch": 0.35, "learning_rate": 4.4217145740737147e-05, "loss": 0.0039, "step": 17882 }, { "epoch": 0.35, "learning_rate": 4.4216498780479916e-05, "loss": 0.0055, "step": 17884 }, { "epoch": 0.35, "learning_rate": 4.4215851820222685e-05, "loss": 0.0047, "step": 17886 }, { "epoch": 0.35, "learning_rate": 4.4215204859965454e-05, "loss": 0.0003, "step": 17888 }, { "epoch": 0.35, "learning_rate": 4.421455789970822e-05, "loss": 0.0001, "step": 17890 }, { "epoch": 0.35, "learning_rate": 4.421391093945099e-05, "loss": 0.008, "step": 17892 }, { "epoch": 0.35, "learning_rate": 4.421326397919376e-05, "loss": 0.0014, "step": 17894 }, { "epoch": 0.35, "learning_rate": 4.421261701893653e-05, "loss": 0.0383, "step": 17896 }, { "epoch": 0.35, "learning_rate": 4.42119700586793e-05, "loss": 0.0006, "step": 17898 }, { "epoch": 0.35, "learning_rate": 4.421132309842206e-05, "loss": 0.013, "step": 17900 }, { "epoch": 0.35, "learning_rate": 4.421067613816484e-05, "loss": 0.0005, "step": 17902 }, { "epoch": 0.35, "learning_rate": 4.42100291779076e-05, "loss": 0.0227, "step": 17904 }, { "epoch": 0.35, "learning_rate": 4.420938221765037e-05, "loss": 0.0004, "step": 17906 }, { "epoch": 0.35, "learning_rate": 4.420873525739314e-05, "loss": 0.0033, "step": 17908 }, { "epoch": 0.35, "learning_rate": 4.420808829713591e-05, "loss": 0.011, "step": 17910 }, { "epoch": 0.35, "learning_rate": 4.420744133687868e-05, "loss": 0.0003, "step": 17912 }, { "epoch": 0.35, "learning_rate": 4.4206794376621445e-05, "loss": 0.0003, "step": 17914 }, { "epoch": 0.35, "learning_rate": 4.4206147416364214e-05, "loss": 0.0203, "step": 17916 }, { "epoch": 0.35, "learning_rate": 4.420550045610698e-05, "loss": 0.0002, "step": 17918 }, { "epoch": 0.35, "learning_rate": 4.420485349584975e-05, "loss": 0.0045, "step": 17920 }, { "epoch": 0.35, "learning_rate": 4.420420653559252e-05, "loss": 0.0119, "step": 17922 }, { "epoch": 0.35, "learning_rate": 4.420355957533529e-05, "loss": 0.0053, "step": 17924 }, { "epoch": 0.35, "learning_rate": 4.420291261507806e-05, "loss": 0.0003, "step": 17926 }, { "epoch": 0.35, "learning_rate": 4.420226565482083e-05, "loss": 0.0027, "step": 17928 }, { "epoch": 0.35, "learning_rate": 4.42016186945636e-05, "loss": 0.0033, "step": 17930 }, { "epoch": 0.35, "learning_rate": 4.420097173430636e-05, "loss": 0.0005, "step": 17932 }, { "epoch": 0.35, "learning_rate": 4.4200324774049136e-05, "loss": 0.0003, "step": 17934 }, { "epoch": 0.35, "learning_rate": 4.41996778137919e-05, "loss": 0.018, "step": 17936 }, { "epoch": 0.35, "learning_rate": 4.419903085353467e-05, "loss": 0.0017, "step": 17938 }, { "epoch": 0.35, "learning_rate": 4.4198383893277436e-05, "loss": 0.0002, "step": 17940 }, { "epoch": 0.35, "learning_rate": 4.4197736933020206e-05, "loss": 0.0499, "step": 17942 }, { "epoch": 0.35, "learning_rate": 4.4197089972762975e-05, "loss": 0.0001, "step": 17944 }, { "epoch": 0.35, "learning_rate": 4.4196443012505744e-05, "loss": 0.0042, "step": 17946 }, { "epoch": 0.35, "learning_rate": 4.419579605224851e-05, "loss": 0.0067, "step": 17948 }, { "epoch": 0.35, "learning_rate": 4.419514909199128e-05, "loss": 0.0002, "step": 17950 }, { "epoch": 0.35, "learning_rate": 4.419450213173405e-05, "loss": 0.0015, "step": 17952 }, { "epoch": 0.35, "learning_rate": 4.419385517147682e-05, "loss": 0.0008, "step": 17954 }, { "epoch": 0.35, "learning_rate": 4.419320821121959e-05, "loss": 0.0165, "step": 17956 }, { "epoch": 0.35, "learning_rate": 4.419256125096236e-05, "loss": 0.0002, "step": 17958 }, { "epoch": 0.35, "learning_rate": 4.419191429070512e-05, "loss": 0.0001, "step": 17960 }, { "epoch": 0.35, "learning_rate": 4.4191267330447896e-05, "loss": 0.0086, "step": 17962 }, { "epoch": 0.35, "learning_rate": 4.419062037019066e-05, "loss": 0.0041, "step": 17964 }, { "epoch": 0.35, "learning_rate": 4.418997340993343e-05, "loss": 0.0009, "step": 17966 }, { "epoch": 0.35, "learning_rate": 4.41893264496762e-05, "loss": 0.0005, "step": 17968 }, { "epoch": 0.35, "learning_rate": 4.4188679489418966e-05, "loss": 0.0013, "step": 17970 }, { "epoch": 0.35, "learning_rate": 4.4188032529161735e-05, "loss": 0.0346, "step": 17972 }, { "epoch": 0.35, "learning_rate": 4.4187385568904504e-05, "loss": 0.009, "step": 17974 }, { "epoch": 0.35, "learning_rate": 4.418673860864727e-05, "loss": 0.0011, "step": 17976 }, { "epoch": 0.35, "learning_rate": 4.418609164839004e-05, "loss": 0.0013, "step": 17978 }, { "epoch": 0.35, "learning_rate": 4.418544468813281e-05, "loss": 0.0005, "step": 17980 }, { "epoch": 0.35, "learning_rate": 4.4184797727875574e-05, "loss": 0.0004, "step": 17982 }, { "epoch": 0.35, "learning_rate": 4.418415076761835e-05, "loss": 0.0109, "step": 17984 }, { "epoch": 0.35, "learning_rate": 4.418350380736111e-05, "loss": 0.0047, "step": 17986 }, { "epoch": 0.35, "learning_rate": 4.418285684710389e-05, "loss": 0.0129, "step": 17988 }, { "epoch": 0.35, "learning_rate": 4.418220988684666e-05, "loss": 0.0098, "step": 17990 }, { "epoch": 0.35, "learning_rate": 4.418156292658942e-05, "loss": 0.0005, "step": 17992 }, { "epoch": 0.35, "learning_rate": 4.4180915966332195e-05, "loss": 0.0017, "step": 17994 }, { "epoch": 0.35, "learning_rate": 4.418026900607496e-05, "loss": 0.0001, "step": 17996 }, { "epoch": 0.35, "learning_rate": 4.4179622045817726e-05, "loss": 0.003, "step": 17998 }, { "epoch": 0.35, "learning_rate": 4.4178975085560495e-05, "loss": 0.0164, "step": 18000 }, { "epoch": 0.35, "learning_rate": 4.4178328125303265e-05, "loss": 0.0066, "step": 18002 }, { "epoch": 0.35, "learning_rate": 4.4177681165046034e-05, "loss": 0.0015, "step": 18004 }, { "epoch": 0.35, "learning_rate": 4.41770342047888e-05, "loss": 0.0155, "step": 18006 }, { "epoch": 0.35, "learning_rate": 4.417638724453157e-05, "loss": 0.0047, "step": 18008 }, { "epoch": 0.35, "learning_rate": 4.417574028427434e-05, "loss": 0.022, "step": 18010 }, { "epoch": 0.35, "learning_rate": 4.417509332401711e-05, "loss": 0.0032, "step": 18012 }, { "epoch": 0.35, "learning_rate": 4.417444636375987e-05, "loss": 0.008, "step": 18014 }, { "epoch": 0.35, "learning_rate": 4.417379940350265e-05, "loss": 0.0002, "step": 18016 }, { "epoch": 0.35, "learning_rate": 4.417315244324541e-05, "loss": 0.0008, "step": 18018 }, { "epoch": 0.35, "learning_rate": 4.417250548298818e-05, "loss": 0.0025, "step": 18020 }, { "epoch": 0.35, "learning_rate": 4.4171858522730955e-05, "loss": 0.007, "step": 18022 }, { "epoch": 0.35, "learning_rate": 4.417121156247372e-05, "loss": 0.0001, "step": 18024 }, { "epoch": 0.35, "learning_rate": 4.417056460221649e-05, "loss": 0.0016, "step": 18026 }, { "epoch": 0.35, "learning_rate": 4.4169917641959256e-05, "loss": 0.0001, "step": 18028 }, { "epoch": 0.35, "learning_rate": 4.4169270681702025e-05, "loss": 0.0044, "step": 18030 }, { "epoch": 0.35, "learning_rate": 4.4168623721444794e-05, "loss": 0.0327, "step": 18032 }, { "epoch": 0.35, "learning_rate": 4.416797676118756e-05, "loss": 0.0052, "step": 18034 }, { "epoch": 0.35, "learning_rate": 4.416732980093033e-05, "loss": 0.0001, "step": 18036 }, { "epoch": 0.35, "learning_rate": 4.41666828406731e-05, "loss": 0.0262, "step": 18038 }, { "epoch": 0.35, "learning_rate": 4.416603588041587e-05, "loss": 0.0003, "step": 18040 }, { "epoch": 0.35, "learning_rate": 4.416538892015863e-05, "loss": 0.0102, "step": 18042 }, { "epoch": 0.35, "learning_rate": 4.416474195990141e-05, "loss": 0.0022, "step": 18044 }, { "epoch": 0.35, "learning_rate": 4.416409499964417e-05, "loss": 0.0005, "step": 18046 }, { "epoch": 0.35, "learning_rate": 4.416344803938694e-05, "loss": 0.001, "step": 18048 }, { "epoch": 0.35, "learning_rate": 4.416280107912971e-05, "loss": 0.0146, "step": 18050 }, { "epoch": 0.35, "learning_rate": 4.416215411887248e-05, "loss": 0.0004, "step": 18052 }, { "epoch": 0.35, "learning_rate": 4.4161507158615254e-05, "loss": 0.0011, "step": 18054 }, { "epoch": 0.35, "learning_rate": 4.4160860198358016e-05, "loss": 0.0004, "step": 18056 }, { "epoch": 0.35, "learning_rate": 4.4160213238100785e-05, "loss": 0.0005, "step": 18058 }, { "epoch": 0.35, "learning_rate": 4.4159566277843554e-05, "loss": 0.0019, "step": 18060 }, { "epoch": 0.35, "learning_rate": 4.4158919317586323e-05, "loss": 0.0009, "step": 18062 }, { "epoch": 0.35, "learning_rate": 4.4158272357329086e-05, "loss": 0.0002, "step": 18064 }, { "epoch": 0.35, "learning_rate": 4.415762539707186e-05, "loss": 0.0031, "step": 18066 }, { "epoch": 0.35, "learning_rate": 4.415697843681463e-05, "loss": 0.0025, "step": 18068 }, { "epoch": 0.35, "learning_rate": 4.41563314765574e-05, "loss": 0.0007, "step": 18070 }, { "epoch": 0.35, "learning_rate": 4.415568451630017e-05, "loss": 0.0023, "step": 18072 }, { "epoch": 0.35, "learning_rate": 4.415503755604293e-05, "loss": 0.0003, "step": 18074 }, { "epoch": 0.35, "learning_rate": 4.415439059578571e-05, "loss": 0.0058, "step": 18076 }, { "epoch": 0.35, "learning_rate": 4.415374363552847e-05, "loss": 0.0001, "step": 18078 }, { "epoch": 0.35, "learning_rate": 4.415309667527124e-05, "loss": 0.0022, "step": 18080 }, { "epoch": 0.35, "learning_rate": 4.415244971501401e-05, "loss": 0.0005, "step": 18082 }, { "epoch": 0.35, "learning_rate": 4.415180275475678e-05, "loss": 0.0001, "step": 18084 }, { "epoch": 0.35, "learning_rate": 4.4151155794499546e-05, "loss": 0.0236, "step": 18086 }, { "epoch": 0.35, "learning_rate": 4.4150508834242315e-05, "loss": 0.0002, "step": 18088 }, { "epoch": 0.35, "learning_rate": 4.4149861873985084e-05, "loss": 0.0004, "step": 18090 }, { "epoch": 0.35, "learning_rate": 4.414921491372785e-05, "loss": 0.0003, "step": 18092 }, { "epoch": 0.35, "learning_rate": 4.414856795347062e-05, "loss": 0.0081, "step": 18094 }, { "epoch": 0.35, "learning_rate": 4.4147920993213384e-05, "loss": 0.0007, "step": 18096 }, { "epoch": 0.35, "learning_rate": 4.414727403295616e-05, "loss": 0.0004, "step": 18098 }, { "epoch": 0.35, "learning_rate": 4.414662707269893e-05, "loss": 0.0103, "step": 18100 }, { "epoch": 0.35, "learning_rate": 4.414598011244169e-05, "loss": 0.0094, "step": 18102 }, { "epoch": 0.35, "learning_rate": 4.414533315218447e-05, "loss": 0.0001, "step": 18104 }, { "epoch": 0.35, "learning_rate": 4.414468619192723e-05, "loss": 0.0001, "step": 18106 }, { "epoch": 0.35, "learning_rate": 4.414403923167e-05, "loss": 0.015, "step": 18108 }, { "epoch": 0.35, "learning_rate": 4.414339227141277e-05, "loss": 0.0123, "step": 18110 }, { "epoch": 0.35, "learning_rate": 4.414274531115554e-05, "loss": 0.0001, "step": 18112 }, { "epoch": 0.35, "learning_rate": 4.4142098350898306e-05, "loss": 0.0234, "step": 18114 }, { "epoch": 0.35, "learning_rate": 4.4141451390641075e-05, "loss": 0.0001, "step": 18116 }, { "epoch": 0.35, "learning_rate": 4.4140804430383844e-05, "loss": 0.0153, "step": 18118 }, { "epoch": 0.35, "learning_rate": 4.414015747012661e-05, "loss": 0.0109, "step": 18120 }, { "epoch": 0.35, "learning_rate": 4.413951050986938e-05, "loss": 0.0013, "step": 18122 }, { "epoch": 0.35, "learning_rate": 4.4138863549612145e-05, "loss": 0.0002, "step": 18124 }, { "epoch": 0.35, "learning_rate": 4.413821658935492e-05, "loss": 0.0183, "step": 18126 }, { "epoch": 0.35, "learning_rate": 4.413756962909768e-05, "loss": 0.0002, "step": 18128 }, { "epoch": 0.35, "learning_rate": 4.413692266884046e-05, "loss": 0.0085, "step": 18130 }, { "epoch": 0.35, "learning_rate": 4.413627570858323e-05, "loss": 0.0001, "step": 18132 }, { "epoch": 0.35, "learning_rate": 4.413562874832599e-05, "loss": 0.004, "step": 18134 }, { "epoch": 0.35, "learning_rate": 4.4134981788068766e-05, "loss": 0.0133, "step": 18136 }, { "epoch": 0.35, "learning_rate": 4.413433482781153e-05, "loss": 0.0026, "step": 18138 }, { "epoch": 0.35, "learning_rate": 4.41336878675543e-05, "loss": 0.0002, "step": 18140 }, { "epoch": 0.35, "learning_rate": 4.4133040907297067e-05, "loss": 0.004, "step": 18142 }, { "epoch": 0.35, "learning_rate": 4.4132393947039836e-05, "loss": 0.0007, "step": 18144 }, { "epoch": 0.35, "learning_rate": 4.4131746986782605e-05, "loss": 0.0012, "step": 18146 }, { "epoch": 0.35, "learning_rate": 4.4131100026525374e-05, "loss": 0.0005, "step": 18148 }, { "epoch": 0.35, "learning_rate": 4.413045306626814e-05, "loss": 0.0029, "step": 18150 }, { "epoch": 0.35, "learning_rate": 4.412980610601091e-05, "loss": 0.0005, "step": 18152 }, { "epoch": 0.35, "learning_rate": 4.412915914575368e-05, "loss": 0.0001, "step": 18154 }, { "epoch": 0.35, "learning_rate": 4.412851218549644e-05, "loss": 0.0003, "step": 18156 }, { "epoch": 0.35, "learning_rate": 4.412786522523922e-05, "loss": 0.0001, "step": 18158 }, { "epoch": 0.35, "learning_rate": 4.412721826498198e-05, "loss": 0.0015, "step": 18160 }, { "epoch": 0.35, "learning_rate": 4.412657130472475e-05, "loss": 0.0001, "step": 18162 }, { "epoch": 0.35, "learning_rate": 4.412592434446752e-05, "loss": 0.0037, "step": 18164 }, { "epoch": 0.35, "learning_rate": 4.412527738421029e-05, "loss": 0.0042, "step": 18166 }, { "epoch": 0.35, "learning_rate": 4.412463042395306e-05, "loss": 0.0001, "step": 18168 }, { "epoch": 0.35, "learning_rate": 4.412398346369583e-05, "loss": 0.0002, "step": 18170 }, { "epoch": 0.35, "learning_rate": 4.4123336503438596e-05, "loss": 0.0207, "step": 18172 }, { "epoch": 0.35, "learning_rate": 4.4122689543181365e-05, "loss": 0.0011, "step": 18174 }, { "epoch": 0.35, "learning_rate": 4.4122042582924134e-05, "loss": 0.0257, "step": 18176 }, { "epoch": 0.35, "learning_rate": 4.41213956226669e-05, "loss": 0.0001, "step": 18178 }, { "epoch": 0.35, "learning_rate": 4.412074866240967e-05, "loss": 0.0093, "step": 18180 }, { "epoch": 0.35, "learning_rate": 4.412010170215244e-05, "loss": 0.0086, "step": 18182 }, { "epoch": 0.35, "learning_rate": 4.4119454741895204e-05, "loss": 0.0006, "step": 18184 }, { "epoch": 0.35, "learning_rate": 4.411880778163798e-05, "loss": 0.0004, "step": 18186 }, { "epoch": 0.35, "learning_rate": 4.411816082138074e-05, "loss": 0.0004, "step": 18188 }, { "epoch": 0.35, "learning_rate": 4.411751386112351e-05, "loss": 0.0048, "step": 18190 }, { "epoch": 0.35, "learning_rate": 4.411686690086628e-05, "loss": 0.0082, "step": 18192 }, { "epoch": 0.35, "learning_rate": 4.411621994060905e-05, "loss": 0.0086, "step": 18194 }, { "epoch": 0.35, "learning_rate": 4.411557298035182e-05, "loss": 0.0014, "step": 18196 }, { "epoch": 0.35, "learning_rate": 4.411492602009459e-05, "loss": 0.0002, "step": 18198 }, { "epoch": 0.35, "learning_rate": 4.4114279059837356e-05, "loss": 0.0008, "step": 18200 }, { "epoch": 0.35, "learning_rate": 4.4113632099580126e-05, "loss": 0.0068, "step": 18202 }, { "epoch": 0.35, "learning_rate": 4.4112985139322895e-05, "loss": 0.007, "step": 18204 }, { "epoch": 0.35, "learning_rate": 4.411233817906566e-05, "loss": 0.0001, "step": 18206 }, { "epoch": 0.35, "learning_rate": 4.411169121880843e-05, "loss": 0.0001, "step": 18208 }, { "epoch": 0.35, "learning_rate": 4.41110442585512e-05, "loss": 0.0002, "step": 18210 }, { "epoch": 0.35, "learning_rate": 4.411039729829397e-05, "loss": 0.0073, "step": 18212 }, { "epoch": 0.35, "learning_rate": 4.410975033803674e-05, "loss": 0.0011, "step": 18214 }, { "epoch": 0.35, "learning_rate": 4.41091033777795e-05, "loss": 0.0005, "step": 18216 }, { "epoch": 0.35, "learning_rate": 4.410845641752228e-05, "loss": 0.0003, "step": 18218 }, { "epoch": 0.35, "learning_rate": 4.410780945726504e-05, "loss": 0.0086, "step": 18220 }, { "epoch": 0.35, "learning_rate": 4.410716249700781e-05, "loss": 0.0015, "step": 18222 }, { "epoch": 0.35, "learning_rate": 4.410651553675058e-05, "loss": 0.0055, "step": 18224 }, { "epoch": 0.35, "learning_rate": 4.410586857649335e-05, "loss": 0.0062, "step": 18226 }, { "epoch": 0.35, "learning_rate": 4.410522161623612e-05, "loss": 0.0017, "step": 18228 }, { "epoch": 0.35, "learning_rate": 4.4104574655978886e-05, "loss": 0.0001, "step": 18230 }, { "epoch": 0.35, "learning_rate": 4.4103927695721655e-05, "loss": 0.0232, "step": 18232 }, { "epoch": 0.35, "learning_rate": 4.4103280735464424e-05, "loss": 0.0052, "step": 18234 }, { "epoch": 0.35, "learning_rate": 4.410263377520719e-05, "loss": 0.0232, "step": 18236 }, { "epoch": 0.35, "learning_rate": 4.4101986814949955e-05, "loss": 0.0009, "step": 18238 }, { "epoch": 0.35, "learning_rate": 4.410133985469273e-05, "loss": 0.0012, "step": 18240 }, { "epoch": 0.35, "learning_rate": 4.4100692894435494e-05, "loss": 0.0004, "step": 18242 }, { "epoch": 0.35, "learning_rate": 4.410004593417826e-05, "loss": 0.0018, "step": 18244 }, { "epoch": 0.35, "learning_rate": 4.409939897392104e-05, "loss": 0.0039, "step": 18246 }, { "epoch": 0.35, "learning_rate": 4.40987520136638e-05, "loss": 0.0046, "step": 18248 }, { "epoch": 0.35, "learning_rate": 4.409810505340657e-05, "loss": 0.0004, "step": 18250 }, { "epoch": 0.35, "learning_rate": 4.409745809314934e-05, "loss": 0.0028, "step": 18252 }, { "epoch": 0.35, "learning_rate": 4.409681113289211e-05, "loss": 0.0004, "step": 18254 }, { "epoch": 0.35, "learning_rate": 4.409616417263488e-05, "loss": 0.0024, "step": 18256 }, { "epoch": 0.35, "learning_rate": 4.4095517212377646e-05, "loss": 0.016, "step": 18258 }, { "epoch": 0.35, "learning_rate": 4.4094870252120415e-05, "loss": 0.001, "step": 18260 }, { "epoch": 0.35, "learning_rate": 4.4094223291863184e-05, "loss": 0.0002, "step": 18262 }, { "epoch": 0.35, "learning_rate": 4.4093576331605954e-05, "loss": 0.0021, "step": 18264 }, { "epoch": 0.35, "learning_rate": 4.4092929371348716e-05, "loss": 0.0003, "step": 18266 }, { "epoch": 0.35, "learning_rate": 4.409228241109149e-05, "loss": 0.0002, "step": 18268 }, { "epoch": 0.35, "learning_rate": 4.4091635450834254e-05, "loss": 0.0017, "step": 18270 }, { "epoch": 0.35, "learning_rate": 4.409098849057703e-05, "loss": 0.0003, "step": 18272 }, { "epoch": 0.35, "learning_rate": 4.409034153031979e-05, "loss": 0.0069, "step": 18274 }, { "epoch": 0.35, "learning_rate": 4.408969457006256e-05, "loss": 0.0034, "step": 18276 }, { "epoch": 0.35, "learning_rate": 4.408904760980534e-05, "loss": 0.0071, "step": 18278 }, { "epoch": 0.35, "learning_rate": 4.40884006495481e-05, "loss": 0.0001, "step": 18280 }, { "epoch": 0.35, "learning_rate": 4.408775368929087e-05, "loss": 0.003, "step": 18282 }, { "epoch": 0.35, "learning_rate": 4.408710672903364e-05, "loss": 0.0031, "step": 18284 }, { "epoch": 0.35, "learning_rate": 4.408645976877641e-05, "loss": 0.0067, "step": 18286 }, { "epoch": 0.35, "learning_rate": 4.408581280851917e-05, "loss": 0.0029, "step": 18288 }, { "epoch": 0.35, "learning_rate": 4.4085165848261945e-05, "loss": 0.0037, "step": 18290 }, { "epoch": 0.36, "learning_rate": 4.4084518888004714e-05, "loss": 0.0003, "step": 18292 }, { "epoch": 0.36, "learning_rate": 4.408387192774748e-05, "loss": 0.0003, "step": 18294 }, { "epoch": 0.36, "learning_rate": 4.408322496749025e-05, "loss": 0.0021, "step": 18296 }, { "epoch": 0.36, "learning_rate": 4.4082578007233014e-05, "loss": 0.0002, "step": 18298 }, { "epoch": 0.36, "learning_rate": 4.408193104697579e-05, "loss": 0.0005, "step": 18300 }, { "epoch": 0.36, "learning_rate": 4.408128408671855e-05, "loss": 0.0096, "step": 18302 }, { "epoch": 0.36, "learning_rate": 4.408063712646132e-05, "loss": 0.0065, "step": 18304 }, { "epoch": 0.36, "learning_rate": 4.407999016620409e-05, "loss": 0.0196, "step": 18306 }, { "epoch": 0.36, "learning_rate": 4.407934320594686e-05, "loss": 0.0159, "step": 18308 }, { "epoch": 0.36, "learning_rate": 4.407869624568963e-05, "loss": 0.0102, "step": 18310 }, { "epoch": 0.36, "learning_rate": 4.40780492854324e-05, "loss": 0.0008, "step": 18312 }, { "epoch": 0.36, "learning_rate": 4.407740232517517e-05, "loss": 0.0034, "step": 18314 }, { "epoch": 0.36, "learning_rate": 4.4076755364917936e-05, "loss": 0.0024, "step": 18316 }, { "epoch": 0.36, "learning_rate": 4.4076108404660705e-05, "loss": 0.0014, "step": 18318 }, { "epoch": 0.36, "learning_rate": 4.407546144440347e-05, "loss": 0.0007, "step": 18320 }, { "epoch": 0.36, "learning_rate": 4.4074814484146243e-05, "loss": 0.0121, "step": 18322 }, { "epoch": 0.36, "learning_rate": 4.407416752388901e-05, "loss": 0.0021, "step": 18324 }, { "epoch": 0.36, "learning_rate": 4.4073520563631775e-05, "loss": 0.0009, "step": 18326 }, { "epoch": 0.36, "learning_rate": 4.407287360337455e-05, "loss": 0.0028, "step": 18328 }, { "epoch": 0.36, "learning_rate": 4.407222664311731e-05, "loss": 0.0009, "step": 18330 }, { "epoch": 0.36, "learning_rate": 4.407157968286008e-05, "loss": 0.0214, "step": 18332 }, { "epoch": 0.36, "learning_rate": 4.407093272260285e-05, "loss": 0.0007, "step": 18334 }, { "epoch": 0.36, "learning_rate": 4.407028576234562e-05, "loss": 0.0059, "step": 18336 }, { "epoch": 0.36, "learning_rate": 4.406963880208839e-05, "loss": 0.0486, "step": 18338 }, { "epoch": 0.36, "learning_rate": 4.406899184183116e-05, "loss": 0.0064, "step": 18340 }, { "epoch": 0.36, "learning_rate": 4.406834488157393e-05, "loss": 0.0001, "step": 18342 }, { "epoch": 0.36, "learning_rate": 4.40676979213167e-05, "loss": 0.0123, "step": 18344 }, { "epoch": 0.36, "learning_rate": 4.4067050961059466e-05, "loss": 0.0075, "step": 18346 }, { "epoch": 0.36, "learning_rate": 4.406640400080223e-05, "loss": 0.0003, "step": 18348 }, { "epoch": 0.36, "learning_rate": 4.4065757040545004e-05, "loss": 0.0014, "step": 18350 }, { "epoch": 0.36, "learning_rate": 4.4065110080287766e-05, "loss": 0.0004, "step": 18352 }, { "epoch": 0.36, "learning_rate": 4.406446312003054e-05, "loss": 0.0144, "step": 18354 }, { "epoch": 0.36, "learning_rate": 4.406381615977331e-05, "loss": 0.0029, "step": 18356 }, { "epoch": 0.36, "learning_rate": 4.4063169199516073e-05, "loss": 0.0001, "step": 18358 }, { "epoch": 0.36, "learning_rate": 4.406252223925885e-05, "loss": 0.0089, "step": 18360 }, { "epoch": 0.36, "learning_rate": 4.406187527900161e-05, "loss": 0.0004, "step": 18362 }, { "epoch": 0.36, "learning_rate": 4.406122831874438e-05, "loss": 0.0031, "step": 18364 }, { "epoch": 0.36, "learning_rate": 4.406058135848715e-05, "loss": 0.0025, "step": 18366 }, { "epoch": 0.36, "learning_rate": 4.405993439822992e-05, "loss": 0.0002, "step": 18368 }, { "epoch": 0.36, "learning_rate": 4.405928743797269e-05, "loss": 0.0012, "step": 18370 }, { "epoch": 0.36, "learning_rate": 4.405864047771546e-05, "loss": 0.0043, "step": 18372 }, { "epoch": 0.36, "learning_rate": 4.4057993517458226e-05, "loss": 0.0078, "step": 18374 }, { "epoch": 0.36, "learning_rate": 4.4057346557200995e-05, "loss": 0.0012, "step": 18376 }, { "epoch": 0.36, "learning_rate": 4.4056699596943764e-05, "loss": 0.0006, "step": 18378 }, { "epoch": 0.36, "learning_rate": 4.4056052636686527e-05, "loss": 0.0001, "step": 18380 }, { "epoch": 0.36, "learning_rate": 4.40554056764293e-05, "loss": 0.0001, "step": 18382 }, { "epoch": 0.36, "learning_rate": 4.4054758716172065e-05, "loss": 0.0013, "step": 18384 }, { "epoch": 0.36, "learning_rate": 4.4054111755914834e-05, "loss": 0.0111, "step": 18386 }, { "epoch": 0.36, "learning_rate": 4.405346479565761e-05, "loss": 0.0212, "step": 18388 }, { "epoch": 0.36, "learning_rate": 4.405281783540037e-05, "loss": 0.0005, "step": 18390 }, { "epoch": 0.36, "learning_rate": 4.405217087514314e-05, "loss": 0.0001, "step": 18392 }, { "epoch": 0.36, "learning_rate": 4.405152391488591e-05, "loss": 0.0002, "step": 18394 }, { "epoch": 0.36, "learning_rate": 4.405087695462868e-05, "loss": 0.0005, "step": 18396 }, { "epoch": 0.36, "learning_rate": 4.405022999437145e-05, "loss": 0.0249, "step": 18398 }, { "epoch": 0.36, "learning_rate": 4.404958303411422e-05, "loss": 0.0081, "step": 18400 }, { "epoch": 0.36, "learning_rate": 4.4048936073856987e-05, "loss": 0.0001, "step": 18402 }, { "epoch": 0.36, "learning_rate": 4.4048289113599756e-05, "loss": 0.0067, "step": 18404 }, { "epoch": 0.36, "learning_rate": 4.4047642153342525e-05, "loss": 0.0045, "step": 18406 }, { "epoch": 0.36, "learning_rate": 4.404699519308529e-05, "loss": 0.0035, "step": 18408 }, { "epoch": 0.36, "learning_rate": 4.404634823282806e-05, "loss": 0.0003, "step": 18410 }, { "epoch": 0.36, "learning_rate": 4.4045701272570825e-05, "loss": 0.0002, "step": 18412 }, { "epoch": 0.36, "learning_rate": 4.4045054312313594e-05, "loss": 0.0281, "step": 18414 }, { "epoch": 0.36, "learning_rate": 4.404440735205636e-05, "loss": 0.0006, "step": 18416 }, { "epoch": 0.36, "learning_rate": 4.404376039179913e-05, "loss": 0.0203, "step": 18418 }, { "epoch": 0.36, "learning_rate": 4.40431134315419e-05, "loss": 0.0036, "step": 18420 }, { "epoch": 0.36, "learning_rate": 4.404246647128467e-05, "loss": 0.0174, "step": 18422 }, { "epoch": 0.36, "learning_rate": 4.404181951102744e-05, "loss": 0.0023, "step": 18424 }, { "epoch": 0.36, "learning_rate": 4.404117255077021e-05, "loss": 0.0001, "step": 18426 }, { "epoch": 0.36, "learning_rate": 4.404052559051298e-05, "loss": 0.0067, "step": 18428 }, { "epoch": 0.36, "learning_rate": 4.403987863025574e-05, "loss": 0.0017, "step": 18430 }, { "epoch": 0.36, "learning_rate": 4.4039231669998516e-05, "loss": 0.0003, "step": 18432 }, { "epoch": 0.36, "learning_rate": 4.4038584709741285e-05, "loss": 0.0002, "step": 18434 }, { "epoch": 0.36, "learning_rate": 4.4037937749484054e-05, "loss": 0.0111, "step": 18436 }, { "epoch": 0.36, "learning_rate": 4.403729078922682e-05, "loss": 0.0155, "step": 18438 }, { "epoch": 0.36, "learning_rate": 4.4036643828969586e-05, "loss": 0.0044, "step": 18440 }, { "epoch": 0.36, "learning_rate": 4.403599686871236e-05, "loss": 0.0001, "step": 18442 }, { "epoch": 0.36, "learning_rate": 4.4035349908455124e-05, "loss": 0.0002, "step": 18444 }, { "epoch": 0.36, "learning_rate": 4.403470294819789e-05, "loss": 0.009, "step": 18446 }, { "epoch": 0.36, "learning_rate": 4.403405598794066e-05, "loss": 0.0003, "step": 18448 }, { "epoch": 0.36, "learning_rate": 4.403340902768343e-05, "loss": 0.0045, "step": 18450 }, { "epoch": 0.36, "learning_rate": 4.40327620674262e-05, "loss": 0.0017, "step": 18452 }, { "epoch": 0.36, "learning_rate": 4.403211510716897e-05, "loss": 0.0004, "step": 18454 }, { "epoch": 0.36, "learning_rate": 4.403146814691174e-05, "loss": 0.0132, "step": 18456 }, { "epoch": 0.36, "learning_rate": 4.403082118665451e-05, "loss": 0.0283, "step": 18458 }, { "epoch": 0.36, "learning_rate": 4.4030174226397276e-05, "loss": 0.0174, "step": 18460 }, { "epoch": 0.36, "learning_rate": 4.402952726614004e-05, "loss": 0.0175, "step": 18462 }, { "epoch": 0.36, "learning_rate": 4.4028880305882815e-05, "loss": 0.008, "step": 18464 }, { "epoch": 0.36, "learning_rate": 4.402823334562558e-05, "loss": 0.0442, "step": 18466 }, { "epoch": 0.36, "learning_rate": 4.4027586385368346e-05, "loss": 0.0006, "step": 18468 }, { "epoch": 0.36, "learning_rate": 4.402693942511112e-05, "loss": 0.0087, "step": 18470 }, { "epoch": 0.36, "learning_rate": 4.4026292464853884e-05, "loss": 0.0004, "step": 18472 }, { "epoch": 0.36, "learning_rate": 4.402564550459665e-05, "loss": 0.0007, "step": 18474 }, { "epoch": 0.36, "learning_rate": 4.402499854433942e-05, "loss": 0.0001, "step": 18476 }, { "epoch": 0.36, "learning_rate": 4.402435158408219e-05, "loss": 0.0001, "step": 18478 }, { "epoch": 0.36, "learning_rate": 4.402370462382496e-05, "loss": 0.0006, "step": 18480 }, { "epoch": 0.36, "learning_rate": 4.402305766356773e-05, "loss": 0.0081, "step": 18482 }, { "epoch": 0.36, "learning_rate": 4.40224107033105e-05, "loss": 0.0028, "step": 18484 }, { "epoch": 0.36, "learning_rate": 4.402176374305327e-05, "loss": 0.0041, "step": 18486 }, { "epoch": 0.36, "learning_rate": 4.402111678279604e-05, "loss": 0.0002, "step": 18488 }, { "epoch": 0.36, "learning_rate": 4.40204698225388e-05, "loss": 0.0002, "step": 18490 }, { "epoch": 0.36, "learning_rate": 4.4019822862281575e-05, "loss": 0.0001, "step": 18492 }, { "epoch": 0.36, "learning_rate": 4.401917590202434e-05, "loss": 0.0005, "step": 18494 }, { "epoch": 0.36, "learning_rate": 4.401852894176711e-05, "loss": 0.0003, "step": 18496 }, { "epoch": 0.36, "learning_rate": 4.4017881981509875e-05, "loss": 0.0123, "step": 18498 }, { "epoch": 0.36, "learning_rate": 4.4017235021252645e-05, "loss": 0.0005, "step": 18500 }, { "epoch": 0.36, "learning_rate": 4.401658806099542e-05, "loss": 0.0005, "step": 18502 }, { "epoch": 0.36, "learning_rate": 4.401594110073818e-05, "loss": 0.0001, "step": 18504 }, { "epoch": 0.36, "learning_rate": 4.401529414048095e-05, "loss": 0.0038, "step": 18506 }, { "epoch": 0.36, "learning_rate": 4.401464718022372e-05, "loss": 0.0087, "step": 18508 }, { "epoch": 0.36, "learning_rate": 4.401400021996649e-05, "loss": 0.0013, "step": 18510 }, { "epoch": 0.36, "learning_rate": 4.401335325970926e-05, "loss": 0.001, "step": 18512 }, { "epoch": 0.36, "learning_rate": 4.401270629945203e-05, "loss": 0.0246, "step": 18514 }, { "epoch": 0.36, "learning_rate": 4.40120593391948e-05, "loss": 0.0014, "step": 18516 }, { "epoch": 0.36, "learning_rate": 4.4011412378937566e-05, "loss": 0.0211, "step": 18518 }, { "epoch": 0.36, "learning_rate": 4.4010765418680335e-05, "loss": 0.0001, "step": 18520 }, { "epoch": 0.36, "learning_rate": 4.40101184584231e-05, "loss": 0.0006, "step": 18522 }, { "epoch": 0.36, "learning_rate": 4.4009471498165874e-05, "loss": 0.0024, "step": 18524 }, { "epoch": 0.36, "learning_rate": 4.4008824537908636e-05, "loss": 0.0133, "step": 18526 }, { "epoch": 0.36, "learning_rate": 4.4008177577651405e-05, "loss": 0.0076, "step": 18528 }, { "epoch": 0.36, "learning_rate": 4.4007530617394174e-05, "loss": 0.0002, "step": 18530 }, { "epoch": 0.36, "learning_rate": 4.400688365713694e-05, "loss": 0.005, "step": 18532 }, { "epoch": 0.36, "learning_rate": 4.400623669687971e-05, "loss": 0.004, "step": 18534 }, { "epoch": 0.36, "learning_rate": 4.400558973662248e-05, "loss": 0.0211, "step": 18536 }, { "epoch": 0.36, "learning_rate": 4.400494277636525e-05, "loss": 0.0003, "step": 18538 }, { "epoch": 0.36, "learning_rate": 4.400429581610802e-05, "loss": 0.0028, "step": 18540 }, { "epoch": 0.36, "learning_rate": 4.400364885585079e-05, "loss": 0.0001, "step": 18542 }, { "epoch": 0.36, "learning_rate": 4.400300189559355e-05, "loss": 0.0001, "step": 18544 }, { "epoch": 0.36, "learning_rate": 4.400235493533633e-05, "loss": 0.0041, "step": 18546 }, { "epoch": 0.36, "learning_rate": 4.4001707975079096e-05, "loss": 0.0053, "step": 18548 }, { "epoch": 0.36, "learning_rate": 4.400106101482186e-05, "loss": 0.0001, "step": 18550 }, { "epoch": 0.36, "learning_rate": 4.4000414054564634e-05, "loss": 0.0002, "step": 18552 }, { "epoch": 0.36, "learning_rate": 4.3999767094307396e-05, "loss": 0.0005, "step": 18554 }, { "epoch": 0.36, "learning_rate": 4.3999120134050165e-05, "loss": 0.0049, "step": 18556 }, { "epoch": 0.36, "learning_rate": 4.3998473173792934e-05, "loss": 0.0002, "step": 18558 }, { "epoch": 0.36, "learning_rate": 4.3997826213535704e-05, "loss": 0.0021, "step": 18560 }, { "epoch": 0.36, "learning_rate": 4.399717925327847e-05, "loss": 0.0875, "step": 18562 }, { "epoch": 0.36, "learning_rate": 4.399653229302124e-05, "loss": 0.0012, "step": 18564 }, { "epoch": 0.36, "learning_rate": 4.399588533276401e-05, "loss": 0.003, "step": 18566 }, { "epoch": 0.36, "learning_rate": 4.399523837250678e-05, "loss": 0.0012, "step": 18568 }, { "epoch": 0.36, "learning_rate": 4.399459141224955e-05, "loss": 0.0002, "step": 18570 }, { "epoch": 0.36, "learning_rate": 4.399394445199231e-05, "loss": 0.0005, "step": 18572 }, { "epoch": 0.36, "learning_rate": 4.399329749173509e-05, "loss": 0.0028, "step": 18574 }, { "epoch": 0.36, "learning_rate": 4.399265053147785e-05, "loss": 0.0044, "step": 18576 }, { "epoch": 0.36, "learning_rate": 4.3992003571220625e-05, "loss": 0.0022, "step": 18578 }, { "epoch": 0.36, "learning_rate": 4.3991356610963394e-05, "loss": 0.0012, "step": 18580 }, { "epoch": 0.36, "learning_rate": 4.399070965070616e-05, "loss": 0.0022, "step": 18582 }, { "epoch": 0.36, "learning_rate": 4.399006269044893e-05, "loss": 0.0075, "step": 18584 }, { "epoch": 0.36, "learning_rate": 4.3989415730191695e-05, "loss": 0.0004, "step": 18586 }, { "epoch": 0.36, "learning_rate": 4.3988768769934464e-05, "loss": 0.0002, "step": 18588 }, { "epoch": 0.36, "learning_rate": 4.398812180967723e-05, "loss": 0.0004, "step": 18590 }, { "epoch": 0.36, "learning_rate": 4.398747484942e-05, "loss": 0.0001, "step": 18592 }, { "epoch": 0.36, "learning_rate": 4.398682788916277e-05, "loss": 0.0001, "step": 18594 }, { "epoch": 0.36, "learning_rate": 4.398618092890554e-05, "loss": 0.0286, "step": 18596 }, { "epoch": 0.36, "learning_rate": 4.398553396864831e-05, "loss": 0.0016, "step": 18598 }, { "epoch": 0.36, "learning_rate": 4.398488700839108e-05, "loss": 0.024, "step": 18600 }, { "epoch": 0.36, "learning_rate": 4.398424004813385e-05, "loss": 0.0126, "step": 18602 }, { "epoch": 0.36, "learning_rate": 4.398359308787661e-05, "loss": 0.0002, "step": 18604 }, { "epoch": 0.36, "learning_rate": 4.3982946127619386e-05, "loss": 0.0007, "step": 18606 }, { "epoch": 0.36, "learning_rate": 4.398229916736215e-05, "loss": 0.0006, "step": 18608 }, { "epoch": 0.36, "learning_rate": 4.398165220710492e-05, "loss": 0.0031, "step": 18610 }, { "epoch": 0.36, "learning_rate": 4.398100524684769e-05, "loss": 0.0081, "step": 18612 }, { "epoch": 0.36, "learning_rate": 4.3980358286590455e-05, "loss": 0.0016, "step": 18614 }, { "epoch": 0.36, "learning_rate": 4.3979711326333224e-05, "loss": 0.0032, "step": 18616 }, { "epoch": 0.36, "learning_rate": 4.3979064366075993e-05, "loss": 0.024, "step": 18618 }, { "epoch": 0.36, "learning_rate": 4.397841740581876e-05, "loss": 0.0001, "step": 18620 }, { "epoch": 0.36, "learning_rate": 4.397777044556153e-05, "loss": 0.0094, "step": 18622 }, { "epoch": 0.36, "learning_rate": 4.39771234853043e-05, "loss": 0.0012, "step": 18624 }, { "epoch": 0.36, "learning_rate": 4.397647652504707e-05, "loss": 0.0001, "step": 18626 }, { "epoch": 0.36, "learning_rate": 4.397582956478984e-05, "loss": 0.0041, "step": 18628 }, { "epoch": 0.36, "learning_rate": 4.397518260453261e-05, "loss": 0.0016, "step": 18630 }, { "epoch": 0.36, "learning_rate": 4.397453564427537e-05, "loss": 0.0004, "step": 18632 }, { "epoch": 0.36, "learning_rate": 4.3973888684018146e-05, "loss": 0.0003, "step": 18634 }, { "epoch": 0.36, "learning_rate": 4.397324172376091e-05, "loss": 0.0007, "step": 18636 }, { "epoch": 0.36, "learning_rate": 4.3972594763503684e-05, "loss": 0.0009, "step": 18638 }, { "epoch": 0.36, "learning_rate": 4.3971947803246447e-05, "loss": 0.0134, "step": 18640 }, { "epoch": 0.36, "learning_rate": 4.3971300842989216e-05, "loss": 0.0009, "step": 18642 }, { "epoch": 0.36, "learning_rate": 4.3970653882731985e-05, "loss": 0.0004, "step": 18644 }, { "epoch": 0.36, "learning_rate": 4.3970006922474754e-05, "loss": 0.0001, "step": 18646 }, { "epoch": 0.36, "learning_rate": 4.396935996221752e-05, "loss": 0.0024, "step": 18648 }, { "epoch": 0.36, "learning_rate": 4.396871300196029e-05, "loss": 0.0063, "step": 18650 }, { "epoch": 0.36, "learning_rate": 4.396806604170306e-05, "loss": 0.0002, "step": 18652 }, { "epoch": 0.36, "learning_rate": 4.396741908144582e-05, "loss": 0.0007, "step": 18654 }, { "epoch": 0.36, "learning_rate": 4.39667721211886e-05, "loss": 0.0071, "step": 18656 }, { "epoch": 0.36, "learning_rate": 4.396612516093137e-05, "loss": 0.0259, "step": 18658 }, { "epoch": 0.36, "learning_rate": 4.396547820067414e-05, "loss": 0.0002, "step": 18660 }, { "epoch": 0.36, "learning_rate": 4.3964831240416906e-05, "loss": 0.0001, "step": 18662 }, { "epoch": 0.36, "learning_rate": 4.396418428015967e-05, "loss": 0.0057, "step": 18664 }, { "epoch": 0.36, "learning_rate": 4.3963537319902445e-05, "loss": 0.0001, "step": 18666 }, { "epoch": 0.36, "learning_rate": 4.396289035964521e-05, "loss": 0.0002, "step": 18668 }, { "epoch": 0.36, "learning_rate": 4.3962243399387976e-05, "loss": 0.0001, "step": 18670 }, { "epoch": 0.36, "learning_rate": 4.3961596439130745e-05, "loss": 0.0002, "step": 18672 }, { "epoch": 0.36, "learning_rate": 4.3960949478873514e-05, "loss": 0.0445, "step": 18674 }, { "epoch": 0.36, "learning_rate": 4.396030251861628e-05, "loss": 0.0117, "step": 18676 }, { "epoch": 0.36, "learning_rate": 4.395965555835905e-05, "loss": 0.0007, "step": 18678 }, { "epoch": 0.36, "learning_rate": 4.395900859810182e-05, "loss": 0.0015, "step": 18680 }, { "epoch": 0.36, "learning_rate": 4.395836163784459e-05, "loss": 0.0002, "step": 18682 }, { "epoch": 0.36, "learning_rate": 4.395771467758736e-05, "loss": 0.0016, "step": 18684 }, { "epoch": 0.36, "learning_rate": 4.395706771733012e-05, "loss": 0.0029, "step": 18686 }, { "epoch": 0.36, "learning_rate": 4.39564207570729e-05, "loss": 0.0002, "step": 18688 }, { "epoch": 0.36, "learning_rate": 4.395577379681567e-05, "loss": 0.0029, "step": 18690 }, { "epoch": 0.36, "learning_rate": 4.395512683655843e-05, "loss": 0.0003, "step": 18692 }, { "epoch": 0.36, "learning_rate": 4.3954479876301205e-05, "loss": 0.0001, "step": 18694 }, { "epoch": 0.36, "learning_rate": 4.395383291604397e-05, "loss": 0.0007, "step": 18696 }, { "epoch": 0.36, "learning_rate": 4.3953185955786736e-05, "loss": 0.0006, "step": 18698 }, { "epoch": 0.36, "learning_rate": 4.3952538995529506e-05, "loss": 0.0041, "step": 18700 }, { "epoch": 0.36, "learning_rate": 4.3951892035272275e-05, "loss": 0.0022, "step": 18702 }, { "epoch": 0.36, "learning_rate": 4.3951245075015044e-05, "loss": 0.0103, "step": 18704 }, { "epoch": 0.36, "learning_rate": 4.395059811475781e-05, "loss": 0.0048, "step": 18706 }, { "epoch": 0.36, "learning_rate": 4.394995115450058e-05, "loss": 0.0059, "step": 18708 }, { "epoch": 0.36, "learning_rate": 4.394930419424335e-05, "loss": 0.0003, "step": 18710 }, { "epoch": 0.36, "learning_rate": 4.394865723398612e-05, "loss": 0.0003, "step": 18712 }, { "epoch": 0.36, "learning_rate": 4.394801027372888e-05, "loss": 0.0018, "step": 18714 }, { "epoch": 0.36, "learning_rate": 4.394736331347166e-05, "loss": 0.0064, "step": 18716 }, { "epoch": 0.36, "learning_rate": 4.394671635321442e-05, "loss": 0.0036, "step": 18718 }, { "epoch": 0.36, "learning_rate": 4.3946069392957196e-05, "loss": 0.0017, "step": 18720 }, { "epoch": 0.36, "learning_rate": 4.394542243269996e-05, "loss": 0.0001, "step": 18722 }, { "epoch": 0.36, "learning_rate": 4.394477547244273e-05, "loss": 0.0004, "step": 18724 }, { "epoch": 0.36, "learning_rate": 4.3944128512185504e-05, "loss": 0.0002, "step": 18726 }, { "epoch": 0.36, "learning_rate": 4.3943481551928266e-05, "loss": 0.0032, "step": 18728 }, { "epoch": 0.36, "learning_rate": 4.3942834591671035e-05, "loss": 0.0005, "step": 18730 }, { "epoch": 0.36, "learning_rate": 4.3942187631413804e-05, "loss": 0.0005, "step": 18732 }, { "epoch": 0.36, "learning_rate": 4.394154067115657e-05, "loss": 0.0031, "step": 18734 }, { "epoch": 0.36, "learning_rate": 4.394089371089934e-05, "loss": 0.0105, "step": 18736 }, { "epoch": 0.36, "learning_rate": 4.394024675064211e-05, "loss": 0.001, "step": 18738 }, { "epoch": 0.36, "learning_rate": 4.393959979038488e-05, "loss": 0.0002, "step": 18740 }, { "epoch": 0.36, "learning_rate": 4.393895283012765e-05, "loss": 0.0059, "step": 18742 }, { "epoch": 0.36, "learning_rate": 4.393830586987042e-05, "loss": 0.0073, "step": 18744 }, { "epoch": 0.36, "learning_rate": 4.393765890961318e-05, "loss": 0.0019, "step": 18746 }, { "epoch": 0.36, "learning_rate": 4.393701194935596e-05, "loss": 0.0061, "step": 18748 }, { "epoch": 0.36, "learning_rate": 4.393636498909872e-05, "loss": 0.0001, "step": 18750 }, { "epoch": 0.36, "learning_rate": 4.393571802884149e-05, "loss": 0.0005, "step": 18752 }, { "epoch": 0.36, "learning_rate": 4.393507106858426e-05, "loss": 0.0096, "step": 18754 }, { "epoch": 0.36, "learning_rate": 4.3934424108327026e-05, "loss": 0.0014, "step": 18756 }, { "epoch": 0.36, "learning_rate": 4.3933777148069795e-05, "loss": 0.0043, "step": 18758 }, { "epoch": 0.36, "learning_rate": 4.3933130187812565e-05, "loss": 0.0027, "step": 18760 }, { "epoch": 0.36, "learning_rate": 4.3932483227555334e-05, "loss": 0.0001, "step": 18762 }, { "epoch": 0.36, "learning_rate": 4.39318362672981e-05, "loss": 0.0014, "step": 18764 }, { "epoch": 0.36, "learning_rate": 4.393118930704087e-05, "loss": 0.0146, "step": 18766 }, { "epoch": 0.36, "learning_rate": 4.3930542346783634e-05, "loss": 0.0031, "step": 18768 }, { "epoch": 0.36, "learning_rate": 4.392989538652641e-05, "loss": 0.0005, "step": 18770 }, { "epoch": 0.36, "learning_rate": 4.392924842626918e-05, "loss": 0.0022, "step": 18772 }, { "epoch": 0.36, "learning_rate": 4.392860146601194e-05, "loss": 0.0075, "step": 18774 }, { "epoch": 0.36, "learning_rate": 4.392795450575472e-05, "loss": 0.0001, "step": 18776 }, { "epoch": 0.36, "learning_rate": 4.392730754549748e-05, "loss": 0.0003, "step": 18778 }, { "epoch": 0.36, "learning_rate": 4.3926660585240255e-05, "loss": 0.0002, "step": 18780 }, { "epoch": 0.36, "learning_rate": 4.392601362498302e-05, "loss": 0.0048, "step": 18782 }, { "epoch": 0.36, "learning_rate": 4.392536666472579e-05, "loss": 0.0005, "step": 18784 }, { "epoch": 0.36, "learning_rate": 4.3924719704468556e-05, "loss": 0.0001, "step": 18786 }, { "epoch": 0.36, "learning_rate": 4.3924072744211325e-05, "loss": 0.0013, "step": 18788 }, { "epoch": 0.36, "learning_rate": 4.3923425783954094e-05, "loss": 0.0012, "step": 18790 }, { "epoch": 0.36, "learning_rate": 4.392277882369686e-05, "loss": 0.0004, "step": 18792 }, { "epoch": 0.36, "learning_rate": 4.392213186343963e-05, "loss": 0.0081, "step": 18794 }, { "epoch": 0.36, "learning_rate": 4.3921484903182394e-05, "loss": 0.0246, "step": 18796 }, { "epoch": 0.36, "learning_rate": 4.392083794292517e-05, "loss": 0.0003, "step": 18798 }, { "epoch": 0.36, "learning_rate": 4.392019098266793e-05, "loss": 0.0093, "step": 18800 }, { "epoch": 0.36, "learning_rate": 4.391954402241071e-05, "loss": 0.0001, "step": 18802 }, { "epoch": 0.36, "learning_rate": 4.391889706215348e-05, "loss": 0.0025, "step": 18804 }, { "epoch": 0.37, "learning_rate": 4.391825010189624e-05, "loss": 0.0018, "step": 18806 }, { "epoch": 0.37, "learning_rate": 4.3917603141639016e-05, "loss": 0.0174, "step": 18808 }, { "epoch": 0.37, "learning_rate": 4.391695618138178e-05, "loss": 0.0008, "step": 18810 }, { "epoch": 0.37, "learning_rate": 4.391630922112455e-05, "loss": 0.0001, "step": 18812 }, { "epoch": 0.37, "learning_rate": 4.3915662260867316e-05, "loss": 0.0027, "step": 18814 }, { "epoch": 0.37, "learning_rate": 4.3915015300610085e-05, "loss": 0.0027, "step": 18816 }, { "epoch": 0.37, "learning_rate": 4.3914368340352854e-05, "loss": 0.0001, "step": 18818 }, { "epoch": 0.37, "learning_rate": 4.3913721380095624e-05, "loss": 0.0074, "step": 18820 }, { "epoch": 0.37, "learning_rate": 4.391307441983839e-05, "loss": 0.0011, "step": 18822 }, { "epoch": 0.37, "learning_rate": 4.391242745958116e-05, "loss": 0.0001, "step": 18824 }, { "epoch": 0.37, "learning_rate": 4.391178049932393e-05, "loss": 0.0004, "step": 18826 }, { "epoch": 0.37, "learning_rate": 4.391113353906669e-05, "loss": 0.0157, "step": 18828 }, { "epoch": 0.37, "learning_rate": 4.391048657880947e-05, "loss": 0.0034, "step": 18830 }, { "epoch": 0.37, "learning_rate": 4.390983961855223e-05, "loss": 0.0155, "step": 18832 }, { "epoch": 0.37, "learning_rate": 4.3909192658295e-05, "loss": 0.0006, "step": 18834 }, { "epoch": 0.37, "learning_rate": 4.3908545698037776e-05, "loss": 0.0018, "step": 18836 }, { "epoch": 0.37, "learning_rate": 4.390789873778054e-05, "loss": 0.0075, "step": 18838 }, { "epoch": 0.37, "learning_rate": 4.390725177752331e-05, "loss": 0.0171, "step": 18840 }, { "epoch": 0.37, "learning_rate": 4.390660481726608e-05, "loss": 0.0002, "step": 18842 }, { "epoch": 0.37, "learning_rate": 4.3905957857008846e-05, "loss": 0.0015, "step": 18844 }, { "epoch": 0.37, "learning_rate": 4.3905310896751615e-05, "loss": 0.004, "step": 18846 }, { "epoch": 0.37, "learning_rate": 4.3904663936494384e-05, "loss": 0.0008, "step": 18848 }, { "epoch": 0.37, "learning_rate": 4.390401697623715e-05, "loss": 0.0001, "step": 18850 }, { "epoch": 0.37, "learning_rate": 4.390337001597992e-05, "loss": 0.0071, "step": 18852 }, { "epoch": 0.37, "learning_rate": 4.390272305572269e-05, "loss": 0.0275, "step": 18854 }, { "epoch": 0.37, "learning_rate": 4.3902076095465453e-05, "loss": 0.0014, "step": 18856 }, { "epoch": 0.37, "learning_rate": 4.390142913520823e-05, "loss": 0.0002, "step": 18858 }, { "epoch": 0.37, "learning_rate": 4.390078217495099e-05, "loss": 0.0004, "step": 18860 }, { "epoch": 0.37, "learning_rate": 4.390013521469377e-05, "loss": 0.0042, "step": 18862 }, { "epoch": 0.37, "learning_rate": 4.389948825443653e-05, "loss": 0.0036, "step": 18864 }, { "epoch": 0.37, "learning_rate": 4.38988412941793e-05, "loss": 0.0005, "step": 18866 }, { "epoch": 0.37, "learning_rate": 4.3898194333922075e-05, "loss": 0.0103, "step": 18868 }, { "epoch": 0.37, "learning_rate": 4.389754737366484e-05, "loss": 0.0016, "step": 18870 }, { "epoch": 0.37, "learning_rate": 4.3896900413407606e-05, "loss": 0.0002, "step": 18872 }, { "epoch": 0.37, "learning_rate": 4.3896253453150375e-05, "loss": 0.0002, "step": 18874 }, { "epoch": 0.37, "learning_rate": 4.3895606492893144e-05, "loss": 0.0081, "step": 18876 }, { "epoch": 0.37, "learning_rate": 4.389495953263591e-05, "loss": 0.0084, "step": 18878 }, { "epoch": 0.37, "learning_rate": 4.389431257237868e-05, "loss": 0.0027, "step": 18880 }, { "epoch": 0.37, "learning_rate": 4.389366561212145e-05, "loss": 0.0003, "step": 18882 }, { "epoch": 0.37, "learning_rate": 4.389301865186422e-05, "loss": 0.001, "step": 18884 }, { "epoch": 0.37, "learning_rate": 4.389237169160699e-05, "loss": 0.0021, "step": 18886 }, { "epoch": 0.37, "learning_rate": 4.389172473134975e-05, "loss": 0.0083, "step": 18888 }, { "epoch": 0.37, "learning_rate": 4.389107777109253e-05, "loss": 0.0153, "step": 18890 }, { "epoch": 0.37, "learning_rate": 4.389043081083529e-05, "loss": 0.0001, "step": 18892 }, { "epoch": 0.37, "learning_rate": 4.388978385057806e-05, "loss": 0.0001, "step": 18894 }, { "epoch": 0.37, "learning_rate": 4.388913689032083e-05, "loss": 0.0001, "step": 18896 }, { "epoch": 0.37, "learning_rate": 4.38884899300636e-05, "loss": 0.0036, "step": 18898 }, { "epoch": 0.37, "learning_rate": 4.3887842969806367e-05, "loss": 0.0012, "step": 18900 }, { "epoch": 0.37, "learning_rate": 4.3887196009549136e-05, "loss": 0.0168, "step": 18902 }, { "epoch": 0.37, "learning_rate": 4.3886549049291905e-05, "loss": 0.001, "step": 18904 }, { "epoch": 0.37, "learning_rate": 4.3885902089034674e-05, "loss": 0.0583, "step": 18906 }, { "epoch": 0.37, "learning_rate": 4.388525512877744e-05, "loss": 0.0033, "step": 18908 }, { "epoch": 0.37, "learning_rate": 4.3884608168520205e-05, "loss": 0.0018, "step": 18910 }, { "epoch": 0.37, "learning_rate": 4.388396120826298e-05, "loss": 0.0007, "step": 18912 }, { "epoch": 0.37, "learning_rate": 4.388331424800575e-05, "loss": 0.006, "step": 18914 }, { "epoch": 0.37, "learning_rate": 4.388266728774851e-05, "loss": 0.0001, "step": 18916 }, { "epoch": 0.37, "learning_rate": 4.388202032749129e-05, "loss": 0.005, "step": 18918 }, { "epoch": 0.37, "learning_rate": 4.388137336723405e-05, "loss": 0.0007, "step": 18920 }, { "epoch": 0.37, "learning_rate": 4.3880726406976826e-05, "loss": 0.0004, "step": 18922 }, { "epoch": 0.37, "learning_rate": 4.388007944671959e-05, "loss": 0.0005, "step": 18924 }, { "epoch": 0.37, "learning_rate": 4.387943248646236e-05, "loss": 0.0012, "step": 18926 }, { "epoch": 0.37, "learning_rate": 4.387878552620513e-05, "loss": 0.0002, "step": 18928 }, { "epoch": 0.37, "learning_rate": 4.3878138565947896e-05, "loss": 0.003, "step": 18930 }, { "epoch": 0.37, "learning_rate": 4.3877491605690665e-05, "loss": 0.0059, "step": 18932 }, { "epoch": 0.37, "learning_rate": 4.3876844645433434e-05, "loss": 0.0003, "step": 18934 }, { "epoch": 0.37, "learning_rate": 4.38761976851762e-05, "loss": 0.0011, "step": 18936 }, { "epoch": 0.37, "learning_rate": 4.3875550724918966e-05, "loss": 0.0018, "step": 18938 }, { "epoch": 0.37, "learning_rate": 4.387490376466174e-05, "loss": 0.0009, "step": 18940 }, { "epoch": 0.37, "learning_rate": 4.3874256804404504e-05, "loss": 0.0062, "step": 18942 }, { "epoch": 0.37, "learning_rate": 4.387360984414728e-05, "loss": 0.0034, "step": 18944 }, { "epoch": 0.37, "learning_rate": 4.387296288389004e-05, "loss": 0.0006, "step": 18946 }, { "epoch": 0.37, "learning_rate": 4.387231592363281e-05, "loss": 0.0001, "step": 18948 }, { "epoch": 0.37, "learning_rate": 4.387166896337559e-05, "loss": 0.0002, "step": 18950 }, { "epoch": 0.37, "learning_rate": 4.387102200311835e-05, "loss": 0.0066, "step": 18952 }, { "epoch": 0.37, "learning_rate": 4.387037504286112e-05, "loss": 0.0009, "step": 18954 }, { "epoch": 0.37, "learning_rate": 4.386972808260389e-05, "loss": 0.0032, "step": 18956 }, { "epoch": 0.37, "learning_rate": 4.3869081122346656e-05, "loss": 0.0002, "step": 18958 }, { "epoch": 0.37, "learning_rate": 4.3868434162089426e-05, "loss": 0.0005, "step": 18960 }, { "epoch": 0.37, "learning_rate": 4.3867787201832195e-05, "loss": 0.0023, "step": 18962 }, { "epoch": 0.37, "learning_rate": 4.3867140241574964e-05, "loss": 0.0003, "step": 18964 }, { "epoch": 0.37, "learning_rate": 4.386649328131773e-05, "loss": 0.0004, "step": 18966 }, { "epoch": 0.37, "learning_rate": 4.38658463210605e-05, "loss": 0.0035, "step": 18968 }, { "epoch": 0.37, "learning_rate": 4.3865199360803264e-05, "loss": 0.005, "step": 18970 }, { "epoch": 0.37, "learning_rate": 4.386455240054604e-05, "loss": 0.0059, "step": 18972 }, { "epoch": 0.37, "learning_rate": 4.38639054402888e-05, "loss": 0.0001, "step": 18974 }, { "epoch": 0.37, "learning_rate": 4.386325848003157e-05, "loss": 0.0001, "step": 18976 }, { "epoch": 0.37, "learning_rate": 4.386261151977434e-05, "loss": 0.0077, "step": 18978 }, { "epoch": 0.37, "learning_rate": 4.386196455951711e-05, "loss": 0.0035, "step": 18980 }, { "epoch": 0.37, "learning_rate": 4.386131759925988e-05, "loss": 0.0009, "step": 18982 }, { "epoch": 0.37, "learning_rate": 4.386067063900265e-05, "loss": 0.0006, "step": 18984 }, { "epoch": 0.37, "learning_rate": 4.386002367874542e-05, "loss": 0.0001, "step": 18986 }, { "epoch": 0.37, "learning_rate": 4.3859376718488186e-05, "loss": 0.0, "step": 18988 }, { "epoch": 0.37, "learning_rate": 4.3858729758230955e-05, "loss": 0.0122, "step": 18990 }, { "epoch": 0.37, "learning_rate": 4.3858082797973724e-05, "loss": 0.0002, "step": 18992 }, { "epoch": 0.37, "learning_rate": 4.385743583771649e-05, "loss": 0.0003, "step": 18994 }, { "epoch": 0.37, "learning_rate": 4.385678887745926e-05, "loss": 0.0001, "step": 18996 }, { "epoch": 0.37, "learning_rate": 4.3856141917202025e-05, "loss": 0.0123, "step": 18998 }, { "epoch": 0.37, "learning_rate": 4.38554949569448e-05, "loss": 0.0082, "step": 19000 }, { "epoch": 0.37, "learning_rate": 4.385484799668756e-05, "loss": 0.0016, "step": 19002 }, { "epoch": 0.37, "learning_rate": 4.385420103643034e-05, "loss": 0.0073, "step": 19004 }, { "epoch": 0.37, "learning_rate": 4.38535540761731e-05, "loss": 0.005, "step": 19006 }, { "epoch": 0.37, "learning_rate": 4.385290711591587e-05, "loss": 0.0119, "step": 19008 }, { "epoch": 0.37, "learning_rate": 4.385226015565864e-05, "loss": 0.0003, "step": 19010 }, { "epoch": 0.37, "learning_rate": 4.385161319540141e-05, "loss": 0.0106, "step": 19012 }, { "epoch": 0.37, "learning_rate": 4.385096623514418e-05, "loss": 0.0032, "step": 19014 }, { "epoch": 0.37, "learning_rate": 4.3850319274886946e-05, "loss": 0.0002, "step": 19016 }, { "epoch": 0.37, "learning_rate": 4.3849672314629715e-05, "loss": 0.012, "step": 19018 }, { "epoch": 0.37, "learning_rate": 4.384902535437248e-05, "loss": 0.0005, "step": 19020 }, { "epoch": 0.37, "learning_rate": 4.3848378394115254e-05, "loss": 0.0006, "step": 19022 }, { "epoch": 0.37, "learning_rate": 4.3847731433858016e-05, "loss": 0.0002, "step": 19024 }, { "epoch": 0.37, "learning_rate": 4.384708447360079e-05, "loss": 0.0004, "step": 19026 }, { "epoch": 0.37, "learning_rate": 4.384643751334356e-05, "loss": 0.0054, "step": 19028 }, { "epoch": 0.37, "learning_rate": 4.384579055308632e-05, "loss": 0.0024, "step": 19030 }, { "epoch": 0.37, "learning_rate": 4.38451435928291e-05, "loss": 0.002, "step": 19032 }, { "epoch": 0.37, "learning_rate": 4.384449663257186e-05, "loss": 0.0025, "step": 19034 }, { "epoch": 0.37, "learning_rate": 4.384384967231463e-05, "loss": 0.0096, "step": 19036 }, { "epoch": 0.37, "learning_rate": 4.38432027120574e-05, "loss": 0.0018, "step": 19038 }, { "epoch": 0.37, "learning_rate": 4.384255575180017e-05, "loss": 0.0002, "step": 19040 }, { "epoch": 0.37, "learning_rate": 4.384190879154294e-05, "loss": 0.0001, "step": 19042 }, { "epoch": 0.37, "learning_rate": 4.384126183128571e-05, "loss": 0.0002, "step": 19044 }, { "epoch": 0.37, "learning_rate": 4.3840614871028476e-05, "loss": 0.0006, "step": 19046 }, { "epoch": 0.37, "learning_rate": 4.3839967910771245e-05, "loss": 0.0036, "step": 19048 }, { "epoch": 0.37, "learning_rate": 4.3839320950514014e-05, "loss": 0.0053, "step": 19050 }, { "epoch": 0.37, "learning_rate": 4.3838673990256776e-05, "loss": 0.0001, "step": 19052 }, { "epoch": 0.37, "learning_rate": 4.383802702999955e-05, "loss": 0.0004, "step": 19054 }, { "epoch": 0.37, "learning_rate": 4.3837380069742314e-05, "loss": 0.0017, "step": 19056 }, { "epoch": 0.37, "learning_rate": 4.3836733109485084e-05, "loss": 0.0016, "step": 19058 }, { "epoch": 0.37, "learning_rate": 4.383608614922786e-05, "loss": 0.0003, "step": 19060 }, { "epoch": 0.37, "learning_rate": 4.383543918897062e-05, "loss": 0.0002, "step": 19062 }, { "epoch": 0.37, "learning_rate": 4.38347922287134e-05, "loss": 0.0136, "step": 19064 }, { "epoch": 0.37, "learning_rate": 4.383414526845616e-05, "loss": 0.0003, "step": 19066 }, { "epoch": 0.37, "learning_rate": 4.383349830819893e-05, "loss": 0.0188, "step": 19068 }, { "epoch": 0.37, "learning_rate": 4.38328513479417e-05, "loss": 0.0219, "step": 19070 }, { "epoch": 0.37, "learning_rate": 4.383220438768447e-05, "loss": 0.0002, "step": 19072 }, { "epoch": 0.37, "learning_rate": 4.3831557427427236e-05, "loss": 0.0029, "step": 19074 }, { "epoch": 0.37, "learning_rate": 4.3830910467170005e-05, "loss": 0.0076, "step": 19076 }, { "epoch": 0.37, "learning_rate": 4.3830263506912774e-05, "loss": 0.0001, "step": 19078 }, { "epoch": 0.37, "learning_rate": 4.382961654665554e-05, "loss": 0.0004, "step": 19080 }, { "epoch": 0.37, "learning_rate": 4.382896958639831e-05, "loss": 0.006, "step": 19082 }, { "epoch": 0.37, "learning_rate": 4.3828322626141075e-05, "loss": 0.0089, "step": 19084 }, { "epoch": 0.37, "learning_rate": 4.382767566588385e-05, "loss": 0.0048, "step": 19086 }, { "epoch": 0.37, "learning_rate": 4.382702870562661e-05, "loss": 0.0015, "step": 19088 }, { "epoch": 0.37, "learning_rate": 4.382638174536938e-05, "loss": 0.0002, "step": 19090 }, { "epoch": 0.37, "learning_rate": 4.382573478511216e-05, "loss": 0.0003, "step": 19092 }, { "epoch": 0.37, "learning_rate": 4.382508782485492e-05, "loss": 0.0002, "step": 19094 }, { "epoch": 0.37, "learning_rate": 4.382444086459769e-05, "loss": 0.0001, "step": 19096 }, { "epoch": 0.37, "learning_rate": 4.382379390434046e-05, "loss": 0.0003, "step": 19098 }, { "epoch": 0.37, "learning_rate": 4.382314694408323e-05, "loss": 0.0031, "step": 19100 }, { "epoch": 0.37, "learning_rate": 4.382249998382599e-05, "loss": 0.0232, "step": 19102 }, { "epoch": 0.37, "learning_rate": 4.3821853023568766e-05, "loss": 0.0275, "step": 19104 }, { "epoch": 0.37, "learning_rate": 4.3821206063311535e-05, "loss": 0.0009, "step": 19106 }, { "epoch": 0.37, "learning_rate": 4.3820559103054304e-05, "loss": 0.0019, "step": 19108 }, { "epoch": 0.37, "learning_rate": 4.381991214279707e-05, "loss": 0.0001, "step": 19110 }, { "epoch": 0.37, "learning_rate": 4.3819265182539835e-05, "loss": 0.0001, "step": 19112 }, { "epoch": 0.37, "learning_rate": 4.381861822228261e-05, "loss": 0.0001, "step": 19114 }, { "epoch": 0.37, "learning_rate": 4.3817971262025373e-05, "loss": 0.0001, "step": 19116 }, { "epoch": 0.37, "learning_rate": 4.381732430176814e-05, "loss": 0.0054, "step": 19118 }, { "epoch": 0.37, "learning_rate": 4.381667734151091e-05, "loss": 0.0002, "step": 19120 }, { "epoch": 0.37, "learning_rate": 4.381603038125368e-05, "loss": 0.0002, "step": 19122 }, { "epoch": 0.37, "learning_rate": 4.381538342099645e-05, "loss": 0.005, "step": 19124 }, { "epoch": 0.37, "learning_rate": 4.381473646073922e-05, "loss": 0.0001, "step": 19126 }, { "epoch": 0.37, "learning_rate": 4.381408950048199e-05, "loss": 0.0002, "step": 19128 }, { "epoch": 0.37, "learning_rate": 4.381344254022476e-05, "loss": 0.0124, "step": 19130 }, { "epoch": 0.37, "learning_rate": 4.3812795579967526e-05, "loss": 0.0056, "step": 19132 }, { "epoch": 0.37, "learning_rate": 4.381214861971029e-05, "loss": 0.019, "step": 19134 }, { "epoch": 0.37, "learning_rate": 4.3811501659453064e-05, "loss": 0.0005, "step": 19136 }, { "epoch": 0.37, "learning_rate": 4.381085469919583e-05, "loss": 0.0161, "step": 19138 }, { "epoch": 0.37, "learning_rate": 4.3810207738938596e-05, "loss": 0.0287, "step": 19140 }, { "epoch": 0.37, "learning_rate": 4.380956077868137e-05, "loss": 0.0002, "step": 19142 }, { "epoch": 0.37, "learning_rate": 4.3808913818424134e-05, "loss": 0.001, "step": 19144 }, { "epoch": 0.37, "learning_rate": 4.380826685816691e-05, "loss": 0.0148, "step": 19146 }, { "epoch": 0.37, "learning_rate": 4.380761989790967e-05, "loss": 0.0032, "step": 19148 }, { "epoch": 0.37, "learning_rate": 4.380697293765244e-05, "loss": 0.0025, "step": 19150 }, { "epoch": 0.37, "learning_rate": 4.380632597739521e-05, "loss": 0.0007, "step": 19152 }, { "epoch": 0.37, "learning_rate": 4.380567901713798e-05, "loss": 0.0083, "step": 19154 }, { "epoch": 0.37, "learning_rate": 4.380503205688075e-05, "loss": 0.0146, "step": 19156 }, { "epoch": 0.37, "learning_rate": 4.380438509662352e-05, "loss": 0.0026, "step": 19158 }, { "epoch": 0.37, "learning_rate": 4.3803738136366287e-05, "loss": 0.0004, "step": 19160 }, { "epoch": 0.37, "learning_rate": 4.380309117610905e-05, "loss": 0.0001, "step": 19162 }, { "epoch": 0.37, "learning_rate": 4.3802444215851825e-05, "loss": 0.0091, "step": 19164 }, { "epoch": 0.37, "learning_rate": 4.380179725559459e-05, "loss": 0.001, "step": 19166 }, { "epoch": 0.37, "learning_rate": 4.380115029533736e-05, "loss": 0.0019, "step": 19168 }, { "epoch": 0.37, "learning_rate": 4.380050333508013e-05, "loss": 0.0007, "step": 19170 }, { "epoch": 0.37, "learning_rate": 4.3799856374822894e-05, "loss": 0.0103, "step": 19172 }, { "epoch": 0.37, "learning_rate": 4.379920941456567e-05, "loss": 0.0063, "step": 19174 }, { "epoch": 0.37, "learning_rate": 4.379856245430843e-05, "loss": 0.008, "step": 19176 }, { "epoch": 0.37, "learning_rate": 4.37979154940512e-05, "loss": 0.0086, "step": 19178 }, { "epoch": 0.37, "learning_rate": 4.379726853379397e-05, "loss": 0.003, "step": 19180 }, { "epoch": 0.37, "learning_rate": 4.379662157353674e-05, "loss": 0.0042, "step": 19182 }, { "epoch": 0.37, "learning_rate": 4.379597461327951e-05, "loss": 0.0063, "step": 19184 }, { "epoch": 0.37, "learning_rate": 4.379532765302228e-05, "loss": 0.0008, "step": 19186 }, { "epoch": 0.37, "learning_rate": 4.379468069276505e-05, "loss": 0.0001, "step": 19188 }, { "epoch": 0.37, "learning_rate": 4.3794033732507816e-05, "loss": 0.0099, "step": 19190 }, { "epoch": 0.37, "learning_rate": 4.3793386772250585e-05, "loss": 0.0076, "step": 19192 }, { "epoch": 0.37, "learning_rate": 4.379273981199335e-05, "loss": 0.0007, "step": 19194 }, { "epoch": 0.37, "learning_rate": 4.379209285173612e-05, "loss": 0.0081, "step": 19196 }, { "epoch": 0.37, "learning_rate": 4.3791445891478886e-05, "loss": 0.0049, "step": 19198 }, { "epoch": 0.37, "learning_rate": 4.3790798931221655e-05, "loss": 0.004, "step": 19200 }, { "epoch": 0.37, "learning_rate": 4.3790151970964424e-05, "loss": 0.0026, "step": 19202 }, { "epoch": 0.37, "learning_rate": 4.378950501070719e-05, "loss": 0.001, "step": 19204 }, { "epoch": 0.37, "learning_rate": 4.378885805044997e-05, "loss": 0.0098, "step": 19206 }, { "epoch": 0.37, "learning_rate": 4.378821109019273e-05, "loss": 0.0054, "step": 19208 }, { "epoch": 0.37, "learning_rate": 4.37875641299355e-05, "loss": 0.0001, "step": 19210 }, { "epoch": 0.37, "learning_rate": 4.378691716967827e-05, "loss": 0.0169, "step": 19212 }, { "epoch": 0.37, "learning_rate": 4.378627020942104e-05, "loss": 0.0097, "step": 19214 }, { "epoch": 0.37, "learning_rate": 4.378562324916381e-05, "loss": 0.001, "step": 19216 }, { "epoch": 0.37, "learning_rate": 4.3784976288906576e-05, "loss": 0.0164, "step": 19218 }, { "epoch": 0.37, "learning_rate": 4.3784329328649346e-05, "loss": 0.0002, "step": 19220 }, { "epoch": 0.37, "learning_rate": 4.378368236839211e-05, "loss": 0.0191, "step": 19222 }, { "epoch": 0.37, "learning_rate": 4.3783035408134884e-05, "loss": 0.001, "step": 19224 }, { "epoch": 0.37, "learning_rate": 4.3782388447877646e-05, "loss": 0.0048, "step": 19226 }, { "epoch": 0.37, "learning_rate": 4.378174148762042e-05, "loss": 0.0035, "step": 19228 }, { "epoch": 0.37, "learning_rate": 4.3781094527363184e-05, "loss": 0.0021, "step": 19230 }, { "epoch": 0.37, "learning_rate": 4.378044756710595e-05, "loss": 0.0052, "step": 19232 }, { "epoch": 0.37, "learning_rate": 4.377980060684872e-05, "loss": 0.0002, "step": 19234 }, { "epoch": 0.37, "learning_rate": 4.377915364659149e-05, "loss": 0.0069, "step": 19236 }, { "epoch": 0.37, "learning_rate": 4.377850668633426e-05, "loss": 0.0001, "step": 19238 }, { "epoch": 0.37, "learning_rate": 4.377785972607703e-05, "loss": 0.0018, "step": 19240 }, { "epoch": 0.37, "learning_rate": 4.37772127658198e-05, "loss": 0.0058, "step": 19242 }, { "epoch": 0.37, "learning_rate": 4.377656580556256e-05, "loss": 0.0002, "step": 19244 }, { "epoch": 0.37, "learning_rate": 4.377591884530534e-05, "loss": 0.0003, "step": 19246 }, { "epoch": 0.37, "learning_rate": 4.37752718850481e-05, "loss": 0.0046, "step": 19248 }, { "epoch": 0.37, "learning_rate": 4.3774624924790875e-05, "loss": 0.0019, "step": 19250 }, { "epoch": 0.37, "learning_rate": 4.3773977964533644e-05, "loss": 0.0115, "step": 19252 }, { "epoch": 0.37, "learning_rate": 4.3773331004276406e-05, "loss": 0.0179, "step": 19254 }, { "epoch": 0.37, "learning_rate": 4.377268404401918e-05, "loss": 0.0131, "step": 19256 }, { "epoch": 0.37, "learning_rate": 4.3772037083761945e-05, "loss": 0.0022, "step": 19258 }, { "epoch": 0.37, "learning_rate": 4.3771390123504714e-05, "loss": 0.0201, "step": 19260 }, { "epoch": 0.37, "learning_rate": 4.377074316324748e-05, "loss": 0.0028, "step": 19262 }, { "epoch": 0.37, "learning_rate": 4.377009620299025e-05, "loss": 0.0001, "step": 19264 }, { "epoch": 0.37, "learning_rate": 4.376944924273302e-05, "loss": 0.0003, "step": 19266 }, { "epoch": 0.37, "learning_rate": 4.376880228247579e-05, "loss": 0.0374, "step": 19268 }, { "epoch": 0.37, "learning_rate": 4.376815532221856e-05, "loss": 0.0001, "step": 19270 }, { "epoch": 0.37, "learning_rate": 4.376750836196133e-05, "loss": 0.0001, "step": 19272 }, { "epoch": 0.37, "learning_rate": 4.37668614017041e-05, "loss": 0.0016, "step": 19274 }, { "epoch": 0.37, "learning_rate": 4.376621444144686e-05, "loss": 0.0014, "step": 19276 }, { "epoch": 0.37, "learning_rate": 4.3765567481189635e-05, "loss": 0.0001, "step": 19278 }, { "epoch": 0.37, "learning_rate": 4.37649205209324e-05, "loss": 0.0003, "step": 19280 }, { "epoch": 0.37, "learning_rate": 4.376427356067517e-05, "loss": 0.0011, "step": 19282 }, { "epoch": 0.37, "learning_rate": 4.376362660041794e-05, "loss": 0.0031, "step": 19284 }, { "epoch": 0.37, "learning_rate": 4.3762979640160705e-05, "loss": 0.0001, "step": 19286 }, { "epoch": 0.37, "learning_rate": 4.376233267990348e-05, "loss": 0.0006, "step": 19288 }, { "epoch": 0.37, "learning_rate": 4.376168571964624e-05, "loss": 0.0044, "step": 19290 }, { "epoch": 0.37, "learning_rate": 4.376103875938901e-05, "loss": 0.0004, "step": 19292 }, { "epoch": 0.37, "learning_rate": 4.376039179913178e-05, "loss": 0.0, "step": 19294 }, { "epoch": 0.37, "learning_rate": 4.375974483887455e-05, "loss": 0.0016, "step": 19296 }, { "epoch": 0.37, "learning_rate": 4.375909787861732e-05, "loss": 0.001, "step": 19298 }, { "epoch": 0.37, "learning_rate": 4.375845091836009e-05, "loss": 0.0034, "step": 19300 }, { "epoch": 0.37, "learning_rate": 4.375780395810286e-05, "loss": 0.001, "step": 19302 }, { "epoch": 0.37, "learning_rate": 4.375715699784562e-05, "loss": 0.005, "step": 19304 }, { "epoch": 0.37, "learning_rate": 4.3756510037588396e-05, "loss": 0.0013, "step": 19306 }, { "epoch": 0.37, "learning_rate": 4.375586307733116e-05, "loss": 0.0169, "step": 19308 }, { "epoch": 0.37, "learning_rate": 4.3755216117073934e-05, "loss": 0.0097, "step": 19310 }, { "epoch": 0.37, "learning_rate": 4.3754569156816696e-05, "loss": 0.0001, "step": 19312 }, { "epoch": 0.37, "learning_rate": 4.3753922196559465e-05, "loss": 0.0042, "step": 19314 }, { "epoch": 0.37, "learning_rate": 4.375327523630224e-05, "loss": 0.0109, "step": 19316 }, { "epoch": 0.37, "learning_rate": 4.3752628276045004e-05, "loss": 0.0036, "step": 19318 }, { "epoch": 0.37, "learning_rate": 4.375198131578777e-05, "loss": 0.0307, "step": 19320 }, { "epoch": 0.38, "learning_rate": 4.375133435553054e-05, "loss": 0.0186, "step": 19322 }, { "epoch": 0.38, "learning_rate": 4.375068739527331e-05, "loss": 0.0115, "step": 19324 }, { "epoch": 0.38, "learning_rate": 4.375004043501607e-05, "loss": 0.0044, "step": 19326 }, { "epoch": 0.38, "learning_rate": 4.374939347475885e-05, "loss": 0.0002, "step": 19328 }, { "epoch": 0.38, "learning_rate": 4.374874651450162e-05, "loss": 0.0067, "step": 19330 }, { "epoch": 0.38, "learning_rate": 4.374809955424439e-05, "loss": 0.0001, "step": 19332 }, { "epoch": 0.38, "learning_rate": 4.3747452593987156e-05, "loss": 0.0138, "step": 19334 }, { "epoch": 0.38, "learning_rate": 4.374680563372992e-05, "loss": 0.0188, "step": 19336 }, { "epoch": 0.38, "learning_rate": 4.3746158673472694e-05, "loss": 0.0023, "step": 19338 }, { "epoch": 0.38, "learning_rate": 4.374551171321546e-05, "loss": 0.0001, "step": 19340 }, { "epoch": 0.38, "learning_rate": 4.3744864752958226e-05, "loss": 0.0022, "step": 19342 }, { "epoch": 0.38, "learning_rate": 4.3744217792700995e-05, "loss": 0.0017, "step": 19344 }, { "epoch": 0.38, "learning_rate": 4.3743570832443764e-05, "loss": 0.0015, "step": 19346 }, { "epoch": 0.38, "learning_rate": 4.374292387218654e-05, "loss": 0.0002, "step": 19348 }, { "epoch": 0.38, "learning_rate": 4.37422769119293e-05, "loss": 0.0004, "step": 19350 }, { "epoch": 0.38, "learning_rate": 4.374162995167207e-05, "loss": 0.0017, "step": 19352 }, { "epoch": 0.38, "learning_rate": 4.374098299141484e-05, "loss": 0.0002, "step": 19354 }, { "epoch": 0.38, "learning_rate": 4.374033603115761e-05, "loss": 0.0001, "step": 19356 }, { "epoch": 0.38, "learning_rate": 4.373968907090037e-05, "loss": 0.036, "step": 19358 }, { "epoch": 0.38, "learning_rate": 4.373904211064315e-05, "loss": 0.0002, "step": 19360 }, { "epoch": 0.38, "learning_rate": 4.373839515038592e-05, "loss": 0.0061, "step": 19362 }, { "epoch": 0.38, "learning_rate": 4.373774819012868e-05, "loss": 0.0004, "step": 19364 }, { "epoch": 0.38, "learning_rate": 4.3737101229871455e-05, "loss": 0.0026, "step": 19366 }, { "epoch": 0.38, "learning_rate": 4.373645426961422e-05, "loss": 0.0024, "step": 19368 }, { "epoch": 0.38, "learning_rate": 4.373580730935699e-05, "loss": 0.014, "step": 19370 }, { "epoch": 0.38, "learning_rate": 4.3735160349099755e-05, "loss": 0.0002, "step": 19372 }, { "epoch": 0.38, "learning_rate": 4.3734513388842524e-05, "loss": 0.0057, "step": 19374 }, { "epoch": 0.38, "learning_rate": 4.3733866428585293e-05, "loss": 0.0, "step": 19376 }, { "epoch": 0.38, "learning_rate": 4.373321946832806e-05, "loss": 0.0003, "step": 19378 }, { "epoch": 0.38, "learning_rate": 4.373257250807083e-05, "loss": 0.0028, "step": 19380 }, { "epoch": 0.38, "learning_rate": 4.37319255478136e-05, "loss": 0.0056, "step": 19382 }, { "epoch": 0.38, "learning_rate": 4.373127858755637e-05, "loss": 0.0039, "step": 19384 }, { "epoch": 0.38, "learning_rate": 4.373063162729913e-05, "loss": 0.0037, "step": 19386 }, { "epoch": 0.38, "learning_rate": 4.372998466704191e-05, "loss": 0.0023, "step": 19388 }, { "epoch": 0.38, "learning_rate": 4.372933770678467e-05, "loss": 0.0073, "step": 19390 }, { "epoch": 0.38, "learning_rate": 4.3728690746527446e-05, "loss": 0.0086, "step": 19392 }, { "epoch": 0.38, "learning_rate": 4.3728043786270215e-05, "loss": 0.0024, "step": 19394 }, { "epoch": 0.38, "learning_rate": 4.372739682601298e-05, "loss": 0.0003, "step": 19396 }, { "epoch": 0.38, "learning_rate": 4.372674986575575e-05, "loss": 0.0001, "step": 19398 }, { "epoch": 0.38, "learning_rate": 4.3726102905498516e-05, "loss": 0.0112, "step": 19400 }, { "epoch": 0.38, "learning_rate": 4.3725455945241285e-05, "loss": 0.0001, "step": 19402 }, { "epoch": 0.38, "learning_rate": 4.3724808984984054e-05, "loss": 0.027, "step": 19404 }, { "epoch": 0.38, "learning_rate": 4.372416202472682e-05, "loss": 0.0003, "step": 19406 }, { "epoch": 0.38, "learning_rate": 4.372351506446959e-05, "loss": 0.0001, "step": 19408 }, { "epoch": 0.38, "learning_rate": 4.372286810421236e-05, "loss": 0.0002, "step": 19410 }, { "epoch": 0.38, "learning_rate": 4.372222114395513e-05, "loss": 0.0041, "step": 19412 }, { "epoch": 0.38, "learning_rate": 4.37215741836979e-05, "loss": 0.0005, "step": 19414 }, { "epoch": 0.38, "learning_rate": 4.372092722344067e-05, "loss": 0.006, "step": 19416 }, { "epoch": 0.38, "learning_rate": 4.372028026318343e-05, "loss": 0.0045, "step": 19418 }, { "epoch": 0.38, "learning_rate": 4.3719633302926207e-05, "loss": 0.0002, "step": 19420 }, { "epoch": 0.38, "learning_rate": 4.371898634266897e-05, "loss": 0.004, "step": 19422 }, { "epoch": 0.38, "learning_rate": 4.371833938241174e-05, "loss": 0.0085, "step": 19424 }, { "epoch": 0.38, "learning_rate": 4.371769242215451e-05, "loss": 0.0006, "step": 19426 }, { "epoch": 0.38, "learning_rate": 4.3717045461897276e-05, "loss": 0.0001, "step": 19428 }, { "epoch": 0.38, "learning_rate": 4.371639850164005e-05, "loss": 0.0002, "step": 19430 }, { "epoch": 0.38, "learning_rate": 4.3715751541382814e-05, "loss": 0.0008, "step": 19432 }, { "epoch": 0.38, "learning_rate": 4.371510458112558e-05, "loss": 0.0064, "step": 19434 }, { "epoch": 0.38, "learning_rate": 4.371445762086835e-05, "loss": 0.0018, "step": 19436 }, { "epoch": 0.38, "learning_rate": 4.371381066061112e-05, "loss": 0.0145, "step": 19438 }, { "epoch": 0.38, "learning_rate": 4.371316370035389e-05, "loss": 0.0004, "step": 19440 }, { "epoch": 0.38, "learning_rate": 4.371251674009666e-05, "loss": 0.0045, "step": 19442 }, { "epoch": 0.38, "learning_rate": 4.371186977983943e-05, "loss": 0.0017, "step": 19444 }, { "epoch": 0.38, "learning_rate": 4.371122281958219e-05, "loss": 0.0007, "step": 19446 }, { "epoch": 0.38, "learning_rate": 4.371057585932497e-05, "loss": 0.0061, "step": 19448 }, { "epoch": 0.38, "learning_rate": 4.370992889906773e-05, "loss": 0.0002, "step": 19450 }, { "epoch": 0.38, "learning_rate": 4.3709281938810505e-05, "loss": 0.0169, "step": 19452 }, { "epoch": 0.38, "learning_rate": 4.370863497855327e-05, "loss": 0.0015, "step": 19454 }, { "epoch": 0.38, "learning_rate": 4.3707988018296036e-05, "loss": 0.0018, "step": 19456 }, { "epoch": 0.38, "learning_rate": 4.3707341058038806e-05, "loss": 0.0001, "step": 19458 }, { "epoch": 0.38, "learning_rate": 4.3706694097781575e-05, "loss": 0.0001, "step": 19460 }, { "epoch": 0.38, "learning_rate": 4.3706047137524344e-05, "loss": 0.0012, "step": 19462 }, { "epoch": 0.38, "learning_rate": 4.370540017726711e-05, "loss": 0.0097, "step": 19464 }, { "epoch": 0.38, "learning_rate": 4.370475321700988e-05, "loss": 0.0096, "step": 19466 }, { "epoch": 0.38, "learning_rate": 4.3704106256752644e-05, "loss": 0.001, "step": 19468 }, { "epoch": 0.38, "learning_rate": 4.370345929649542e-05, "loss": 0.0002, "step": 19470 }, { "epoch": 0.38, "learning_rate": 4.370281233623819e-05, "loss": 0.0047, "step": 19472 }, { "epoch": 0.38, "learning_rate": 4.370216537598096e-05, "loss": 0.0003, "step": 19474 }, { "epoch": 0.38, "learning_rate": 4.370151841572373e-05, "loss": 0.0036, "step": 19476 }, { "epoch": 0.38, "learning_rate": 4.370087145546649e-05, "loss": 0.0003, "step": 19478 }, { "epoch": 0.38, "learning_rate": 4.3700224495209265e-05, "loss": 0.0811, "step": 19480 }, { "epoch": 0.38, "learning_rate": 4.369957753495203e-05, "loss": 0.0099, "step": 19482 }, { "epoch": 0.38, "learning_rate": 4.36989305746948e-05, "loss": 0.0001, "step": 19484 }, { "epoch": 0.38, "learning_rate": 4.3698283614437566e-05, "loss": 0.0138, "step": 19486 }, { "epoch": 0.38, "learning_rate": 4.3697636654180335e-05, "loss": 0.0001, "step": 19488 }, { "epoch": 0.38, "learning_rate": 4.3696989693923104e-05, "loss": 0.0932, "step": 19490 }, { "epoch": 0.38, "learning_rate": 4.369634273366587e-05, "loss": 0.0091, "step": 19492 }, { "epoch": 0.38, "learning_rate": 4.369569577340864e-05, "loss": 0.0106, "step": 19494 }, { "epoch": 0.38, "learning_rate": 4.369504881315141e-05, "loss": 0.0053, "step": 19496 }, { "epoch": 0.38, "learning_rate": 4.369440185289418e-05, "loss": 0.005, "step": 19498 }, { "epoch": 0.38, "learning_rate": 4.369375489263694e-05, "loss": 0.0177, "step": 19500 }, { "epoch": 0.38, "learning_rate": 4.369310793237972e-05, "loss": 0.0027, "step": 19502 }, { "epoch": 0.38, "learning_rate": 4.369246097212248e-05, "loss": 0.0018, "step": 19504 }, { "epoch": 0.38, "learning_rate": 4.369181401186525e-05, "loss": 0.0004, "step": 19506 }, { "epoch": 0.38, "learning_rate": 4.3691167051608026e-05, "loss": 0.0005, "step": 19508 }, { "epoch": 0.38, "learning_rate": 4.369052009135079e-05, "loss": 0.0066, "step": 19510 }, { "epoch": 0.38, "learning_rate": 4.3689873131093564e-05, "loss": 0.0002, "step": 19512 }, { "epoch": 0.38, "learning_rate": 4.3689226170836326e-05, "loss": 0.0014, "step": 19514 }, { "epoch": 0.38, "learning_rate": 4.3688579210579095e-05, "loss": 0.0003, "step": 19516 }, { "epoch": 0.38, "learning_rate": 4.3687932250321865e-05, "loss": 0.0065, "step": 19518 }, { "epoch": 0.38, "learning_rate": 4.3687285290064634e-05, "loss": 0.0003, "step": 19520 }, { "epoch": 0.38, "learning_rate": 4.36866383298074e-05, "loss": 0.0085, "step": 19522 }, { "epoch": 0.38, "learning_rate": 4.368599136955017e-05, "loss": 0.0009, "step": 19524 }, { "epoch": 0.38, "learning_rate": 4.368534440929294e-05, "loss": 0.0004, "step": 19526 }, { "epoch": 0.38, "learning_rate": 4.36846974490357e-05, "loss": 0.0003, "step": 19528 }, { "epoch": 0.38, "learning_rate": 4.368405048877848e-05, "loss": 0.0003, "step": 19530 }, { "epoch": 0.38, "learning_rate": 4.368340352852124e-05, "loss": 0.0061, "step": 19532 }, { "epoch": 0.38, "learning_rate": 4.368275656826402e-05, "loss": 0.0004, "step": 19534 }, { "epoch": 0.38, "learning_rate": 4.368210960800678e-05, "loss": 0.0003, "step": 19536 }, { "epoch": 0.38, "learning_rate": 4.368146264774955e-05, "loss": 0.0002, "step": 19538 }, { "epoch": 0.38, "learning_rate": 4.3680815687492324e-05, "loss": 0.0027, "step": 19540 }, { "epoch": 0.38, "learning_rate": 4.368016872723509e-05, "loss": 0.0004, "step": 19542 }, { "epoch": 0.38, "learning_rate": 4.3679521766977856e-05, "loss": 0.0046, "step": 19544 }, { "epoch": 0.38, "learning_rate": 4.3678874806720625e-05, "loss": 0.0024, "step": 19546 }, { "epoch": 0.38, "learning_rate": 4.3678227846463394e-05, "loss": 0.0166, "step": 19548 }, { "epoch": 0.38, "learning_rate": 4.367758088620616e-05, "loss": 0.0062, "step": 19550 }, { "epoch": 0.38, "learning_rate": 4.367693392594893e-05, "loss": 0.0003, "step": 19552 }, { "epoch": 0.38, "learning_rate": 4.36762869656917e-05, "loss": 0.0009, "step": 19554 }, { "epoch": 0.38, "learning_rate": 4.367564000543447e-05, "loss": 0.0044, "step": 19556 }, { "epoch": 0.38, "learning_rate": 4.367499304517724e-05, "loss": 0.0003, "step": 19558 }, { "epoch": 0.38, "learning_rate": 4.367434608492e-05, "loss": 0.0166, "step": 19560 }, { "epoch": 0.38, "learning_rate": 4.367369912466278e-05, "loss": 0.003, "step": 19562 }, { "epoch": 0.38, "learning_rate": 4.367305216440554e-05, "loss": 0.0102, "step": 19564 }, { "epoch": 0.38, "learning_rate": 4.367240520414831e-05, "loss": 0.0175, "step": 19566 }, { "epoch": 0.38, "learning_rate": 4.367175824389108e-05, "loss": 0.0002, "step": 19568 }, { "epoch": 0.38, "learning_rate": 4.367111128363385e-05, "loss": 0.0008, "step": 19570 }, { "epoch": 0.38, "learning_rate": 4.367046432337662e-05, "loss": 0.0002, "step": 19572 }, { "epoch": 0.38, "learning_rate": 4.3669817363119385e-05, "loss": 0.0005, "step": 19574 }, { "epoch": 0.38, "learning_rate": 4.3669170402862154e-05, "loss": 0.0002, "step": 19576 }, { "epoch": 0.38, "learning_rate": 4.3668523442604924e-05, "loss": 0.0002, "step": 19578 }, { "epoch": 0.38, "learning_rate": 4.366787648234769e-05, "loss": 0.0002, "step": 19580 }, { "epoch": 0.38, "learning_rate": 4.3667229522090455e-05, "loss": 0.001, "step": 19582 }, { "epoch": 0.38, "learning_rate": 4.366658256183323e-05, "loss": 0.0037, "step": 19584 }, { "epoch": 0.38, "learning_rate": 4.3665935601576e-05, "loss": 0.002, "step": 19586 }, { "epoch": 0.38, "learning_rate": 4.366528864131876e-05, "loss": 0.0026, "step": 19588 }, { "epoch": 0.38, "learning_rate": 4.366464168106154e-05, "loss": 0.0008, "step": 19590 }, { "epoch": 0.38, "learning_rate": 4.36639947208043e-05, "loss": 0.0001, "step": 19592 }, { "epoch": 0.38, "learning_rate": 4.3663347760547076e-05, "loss": 0.006, "step": 19594 }, { "epoch": 0.38, "learning_rate": 4.366270080028984e-05, "loss": 0.0002, "step": 19596 }, { "epoch": 0.38, "learning_rate": 4.366205384003261e-05, "loss": 0.0009, "step": 19598 }, { "epoch": 0.38, "learning_rate": 4.366140687977538e-05, "loss": 0.0001, "step": 19600 }, { "epoch": 0.38, "learning_rate": 4.3660759919518146e-05, "loss": 0.0015, "step": 19602 }, { "epoch": 0.38, "learning_rate": 4.3660112959260915e-05, "loss": 0.0001, "step": 19604 }, { "epoch": 0.38, "learning_rate": 4.3659465999003684e-05, "loss": 0.0136, "step": 19606 }, { "epoch": 0.38, "learning_rate": 4.365881903874645e-05, "loss": 0.0082, "step": 19608 }, { "epoch": 0.38, "learning_rate": 4.3658172078489215e-05, "loss": 0.0004, "step": 19610 }, { "epoch": 0.38, "learning_rate": 4.365752511823199e-05, "loss": 0.0002, "step": 19612 }, { "epoch": 0.38, "learning_rate": 4.3656878157974753e-05, "loss": 0.003, "step": 19614 }, { "epoch": 0.38, "learning_rate": 4.365623119771753e-05, "loss": 0.0026, "step": 19616 }, { "epoch": 0.38, "learning_rate": 4.36555842374603e-05, "loss": 0.0004, "step": 19618 }, { "epoch": 0.38, "learning_rate": 4.365493727720306e-05, "loss": 0.001, "step": 19620 }, { "epoch": 0.38, "learning_rate": 4.3654290316945837e-05, "loss": 0.0035, "step": 19622 }, { "epoch": 0.38, "learning_rate": 4.36536433566886e-05, "loss": 0.0077, "step": 19624 }, { "epoch": 0.38, "learning_rate": 4.365299639643137e-05, "loss": 0.0087, "step": 19626 }, { "epoch": 0.38, "learning_rate": 4.365234943617414e-05, "loss": 0.0037, "step": 19628 }, { "epoch": 0.38, "learning_rate": 4.3651702475916906e-05, "loss": 0.0175, "step": 19630 }, { "epoch": 0.38, "learning_rate": 4.3651055515659675e-05, "loss": 0.0134, "step": 19632 }, { "epoch": 0.38, "learning_rate": 4.3650408555402444e-05, "loss": 0.0001, "step": 19634 }, { "epoch": 0.38, "learning_rate": 4.3649761595145213e-05, "loss": 0.002, "step": 19636 }, { "epoch": 0.38, "learning_rate": 4.364911463488798e-05, "loss": 0.0007, "step": 19638 }, { "epoch": 0.38, "learning_rate": 4.364846767463075e-05, "loss": 0.0049, "step": 19640 }, { "epoch": 0.38, "learning_rate": 4.3647820714373514e-05, "loss": 0.0002, "step": 19642 }, { "epoch": 0.38, "learning_rate": 4.364717375411629e-05, "loss": 0.0025, "step": 19644 }, { "epoch": 0.38, "learning_rate": 4.364652679385905e-05, "loss": 0.0107, "step": 19646 }, { "epoch": 0.38, "learning_rate": 4.364587983360182e-05, "loss": 0.0028, "step": 19648 }, { "epoch": 0.38, "learning_rate": 4.36452328733446e-05, "loss": 0.0001, "step": 19650 }, { "epoch": 0.38, "learning_rate": 4.364458591308736e-05, "loss": 0.0029, "step": 19652 }, { "epoch": 0.38, "learning_rate": 4.3643938952830135e-05, "loss": 0.0158, "step": 19654 }, { "epoch": 0.38, "learning_rate": 4.36432919925729e-05, "loss": 0.0, "step": 19656 }, { "epoch": 0.38, "learning_rate": 4.3642645032315667e-05, "loss": 0.0001, "step": 19658 }, { "epoch": 0.38, "learning_rate": 4.3641998072058436e-05, "loss": 0.0012, "step": 19660 }, { "epoch": 0.38, "learning_rate": 4.3641351111801205e-05, "loss": 0.0082, "step": 19662 }, { "epoch": 0.38, "learning_rate": 4.3640704151543974e-05, "loss": 0.013, "step": 19664 }, { "epoch": 0.38, "learning_rate": 4.364005719128674e-05, "loss": 0.0019, "step": 19666 }, { "epoch": 0.38, "learning_rate": 4.363941023102951e-05, "loss": 0.0325, "step": 19668 }, { "epoch": 0.38, "learning_rate": 4.3638763270772274e-05, "loss": 0.0084, "step": 19670 }, { "epoch": 0.38, "learning_rate": 4.363811631051505e-05, "loss": 0.0003, "step": 19672 }, { "epoch": 0.38, "learning_rate": 4.363746935025781e-05, "loss": 0.001, "step": 19674 }, { "epoch": 0.38, "learning_rate": 4.363682239000059e-05, "loss": 0.0032, "step": 19676 }, { "epoch": 0.38, "learning_rate": 4.363617542974335e-05, "loss": 0.0033, "step": 19678 }, { "epoch": 0.38, "learning_rate": 4.363552846948612e-05, "loss": 0.0079, "step": 19680 }, { "epoch": 0.38, "learning_rate": 4.363488150922889e-05, "loss": 0.0115, "step": 19682 }, { "epoch": 0.38, "learning_rate": 4.363423454897166e-05, "loss": 0.0031, "step": 19684 }, { "epoch": 0.38, "learning_rate": 4.363358758871443e-05, "loss": 0.0014, "step": 19686 }, { "epoch": 0.38, "learning_rate": 4.3632940628457196e-05, "loss": 0.0045, "step": 19688 }, { "epoch": 0.38, "learning_rate": 4.3632293668199965e-05, "loss": 0.0136, "step": 19690 }, { "epoch": 0.38, "learning_rate": 4.363164670794273e-05, "loss": 0.0002, "step": 19692 }, { "epoch": 0.38, "learning_rate": 4.36309997476855e-05, "loss": 0.0001, "step": 19694 }, { "epoch": 0.38, "learning_rate": 4.363035278742827e-05, "loss": 0.0039, "step": 19696 }, { "epoch": 0.38, "learning_rate": 4.362970582717104e-05, "loss": 0.0073, "step": 19698 }, { "epoch": 0.38, "learning_rate": 4.362905886691381e-05, "loss": 0.0013, "step": 19700 }, { "epoch": 0.38, "learning_rate": 4.362841190665657e-05, "loss": 0.0036, "step": 19702 }, { "epoch": 0.38, "learning_rate": 4.362776494639935e-05, "loss": 0.0115, "step": 19704 }, { "epoch": 0.38, "learning_rate": 4.362711798614211e-05, "loss": 0.0004, "step": 19706 }, { "epoch": 0.38, "learning_rate": 4.362647102588488e-05, "loss": 0.0098, "step": 19708 }, { "epoch": 0.38, "learning_rate": 4.362582406562765e-05, "loss": 0.0092, "step": 19710 }, { "epoch": 0.38, "learning_rate": 4.362517710537042e-05, "loss": 0.0132, "step": 19712 }, { "epoch": 0.38, "learning_rate": 4.362453014511319e-05, "loss": 0.0016, "step": 19714 }, { "epoch": 0.38, "learning_rate": 4.3623883184855956e-05, "loss": 0.0039, "step": 19716 }, { "epoch": 0.38, "learning_rate": 4.3623236224598726e-05, "loss": 0.0001, "step": 19718 }, { "epoch": 0.38, "learning_rate": 4.3622589264341495e-05, "loss": 0.009, "step": 19720 }, { "epoch": 0.38, "learning_rate": 4.3621942304084264e-05, "loss": 0.0073, "step": 19722 }, { "epoch": 0.38, "learning_rate": 4.3621295343827026e-05, "loss": 0.0004, "step": 19724 }, { "epoch": 0.38, "learning_rate": 4.36206483835698e-05, "loss": 0.0062, "step": 19726 }, { "epoch": 0.38, "learning_rate": 4.3620001423312564e-05, "loss": 0.0005, "step": 19728 }, { "epoch": 0.38, "learning_rate": 4.361935446305533e-05, "loss": 0.0019, "step": 19730 }, { "epoch": 0.38, "learning_rate": 4.361870750279811e-05, "loss": 0.0017, "step": 19732 }, { "epoch": 0.38, "learning_rate": 4.361806054254087e-05, "loss": 0.0009, "step": 19734 }, { "epoch": 0.38, "learning_rate": 4.361741358228365e-05, "loss": 0.0002, "step": 19736 }, { "epoch": 0.38, "learning_rate": 4.361676662202641e-05, "loss": 0.0001, "step": 19738 }, { "epoch": 0.38, "learning_rate": 4.361611966176918e-05, "loss": 0.0105, "step": 19740 }, { "epoch": 0.38, "learning_rate": 4.361547270151195e-05, "loss": 0.0019, "step": 19742 }, { "epoch": 0.38, "learning_rate": 4.361482574125472e-05, "loss": 0.009, "step": 19744 }, { "epoch": 0.38, "learning_rate": 4.3614178780997486e-05, "loss": 0.0016, "step": 19746 }, { "epoch": 0.38, "learning_rate": 4.3613531820740255e-05, "loss": 0.0032, "step": 19748 }, { "epoch": 0.38, "learning_rate": 4.3612884860483024e-05, "loss": 0.005, "step": 19750 }, { "epoch": 0.38, "learning_rate": 4.3612237900225786e-05, "loss": 0.0046, "step": 19752 }, { "epoch": 0.38, "learning_rate": 4.361159093996856e-05, "loss": 0.0001, "step": 19754 }, { "epoch": 0.38, "learning_rate": 4.3610943979711325e-05, "loss": 0.0004, "step": 19756 }, { "epoch": 0.38, "learning_rate": 4.36102970194541e-05, "loss": 0.0323, "step": 19758 }, { "epoch": 0.38, "learning_rate": 4.360965005919686e-05, "loss": 0.0006, "step": 19760 }, { "epoch": 0.38, "learning_rate": 4.360900309893963e-05, "loss": 0.0038, "step": 19762 }, { "epoch": 0.38, "learning_rate": 4.360835613868241e-05, "loss": 0.0109, "step": 19764 }, { "epoch": 0.38, "learning_rate": 4.360770917842517e-05, "loss": 0.0003, "step": 19766 }, { "epoch": 0.38, "learning_rate": 4.360706221816794e-05, "loss": 0.0003, "step": 19768 }, { "epoch": 0.38, "learning_rate": 4.360641525791071e-05, "loss": 0.0024, "step": 19770 }, { "epoch": 0.38, "learning_rate": 4.360576829765348e-05, "loss": 0.0003, "step": 19772 }, { "epoch": 0.38, "learning_rate": 4.3605121337396246e-05, "loss": 0.0007, "step": 19774 }, { "epoch": 0.38, "learning_rate": 4.3604474377139015e-05, "loss": 0.0092, "step": 19776 }, { "epoch": 0.38, "learning_rate": 4.3603827416881785e-05, "loss": 0.0026, "step": 19778 }, { "epoch": 0.38, "learning_rate": 4.3603180456624554e-05, "loss": 0.0155, "step": 19780 }, { "epoch": 0.38, "learning_rate": 4.360253349636732e-05, "loss": 0.0001, "step": 19782 }, { "epoch": 0.38, "learning_rate": 4.3601886536110085e-05, "loss": 0.0092, "step": 19784 }, { "epoch": 0.38, "learning_rate": 4.360123957585286e-05, "loss": 0.0006, "step": 19786 }, { "epoch": 0.38, "learning_rate": 4.360059261559562e-05, "loss": 0.0002, "step": 19788 }, { "epoch": 0.38, "learning_rate": 4.359994565533839e-05, "loss": 0.0135, "step": 19790 }, { "epoch": 0.38, "learning_rate": 4.359929869508116e-05, "loss": 0.0001, "step": 19792 }, { "epoch": 0.38, "learning_rate": 4.359865173482393e-05, "loss": 0.0003, "step": 19794 }, { "epoch": 0.38, "learning_rate": 4.3598004774566706e-05, "loss": 0.0031, "step": 19796 }, { "epoch": 0.38, "learning_rate": 4.359735781430947e-05, "loss": 0.0006, "step": 19798 }, { "epoch": 0.38, "learning_rate": 4.359671085405224e-05, "loss": 0.0002, "step": 19800 }, { "epoch": 0.38, "learning_rate": 4.359606389379501e-05, "loss": 0.0091, "step": 19802 }, { "epoch": 0.38, "learning_rate": 4.3595416933537776e-05, "loss": 0.0034, "step": 19804 }, { "epoch": 0.38, "learning_rate": 4.359476997328054e-05, "loss": 0.0001, "step": 19806 }, { "epoch": 0.38, "learning_rate": 4.3594123013023314e-05, "loss": 0.0002, "step": 19808 }, { "epoch": 0.38, "learning_rate": 4.359347605276608e-05, "loss": 0.0002, "step": 19810 }, { "epoch": 0.38, "learning_rate": 4.3592829092508845e-05, "loss": 0.0217, "step": 19812 }, { "epoch": 0.38, "learning_rate": 4.359218213225162e-05, "loss": 0.0012, "step": 19814 }, { "epoch": 0.38, "learning_rate": 4.3591535171994384e-05, "loss": 0.0069, "step": 19816 }, { "epoch": 0.38, "learning_rate": 4.359088821173716e-05, "loss": 0.0002, "step": 19818 }, { "epoch": 0.38, "learning_rate": 4.359024125147992e-05, "loss": 0.0002, "step": 19820 }, { "epoch": 0.38, "learning_rate": 4.358959429122269e-05, "loss": 0.0001, "step": 19822 }, { "epoch": 0.38, "learning_rate": 4.358894733096546e-05, "loss": 0.0004, "step": 19824 }, { "epoch": 0.38, "learning_rate": 4.358830037070823e-05, "loss": 0.0001, "step": 19826 }, { "epoch": 0.38, "learning_rate": 4.3587653410451e-05, "loss": 0.0139, "step": 19828 }, { "epoch": 0.38, "learning_rate": 4.358700645019377e-05, "loss": 0.0189, "step": 19830 }, { "epoch": 0.38, "learning_rate": 4.3586359489936536e-05, "loss": 0.0032, "step": 19832 }, { "epoch": 0.38, "learning_rate": 4.35857125296793e-05, "loss": 0.0071, "step": 19834 }, { "epoch": 0.38, "learning_rate": 4.3585065569422074e-05, "loss": 0.0017, "step": 19836 }, { "epoch": 0.39, "learning_rate": 4.358441860916484e-05, "loss": 0.0001, "step": 19838 }, { "epoch": 0.39, "learning_rate": 4.358377164890761e-05, "loss": 0.0042, "step": 19840 }, { "epoch": 0.39, "learning_rate": 4.358312468865038e-05, "loss": 0.0035, "step": 19842 }, { "epoch": 0.39, "learning_rate": 4.3582477728393144e-05, "loss": 0.0001, "step": 19844 }, { "epoch": 0.39, "learning_rate": 4.358183076813592e-05, "loss": 0.0007, "step": 19846 }, { "epoch": 0.39, "learning_rate": 4.358118380787868e-05, "loss": 0.0004, "step": 19848 }, { "epoch": 0.39, "learning_rate": 4.358053684762145e-05, "loss": 0.1025, "step": 19850 }, { "epoch": 0.39, "learning_rate": 4.357988988736422e-05, "loss": 0.0016, "step": 19852 }, { "epoch": 0.39, "learning_rate": 4.357924292710699e-05, "loss": 0.0003, "step": 19854 }, { "epoch": 0.39, "learning_rate": 4.357859596684976e-05, "loss": 0.0097, "step": 19856 }, { "epoch": 0.39, "learning_rate": 4.357794900659253e-05, "loss": 0.0344, "step": 19858 }, { "epoch": 0.39, "learning_rate": 4.35773020463353e-05, "loss": 0.018, "step": 19860 }, { "epoch": 0.39, "learning_rate": 4.3576655086078066e-05, "loss": 0.0003, "step": 19862 }, { "epoch": 0.39, "learning_rate": 4.3576008125820835e-05, "loss": 0.0239, "step": 19864 }, { "epoch": 0.39, "learning_rate": 4.35753611655636e-05, "loss": 0.0116, "step": 19866 }, { "epoch": 0.39, "learning_rate": 4.357471420530637e-05, "loss": 0.0077, "step": 19868 }, { "epoch": 0.39, "learning_rate": 4.3574067245049135e-05, "loss": 0.0007, "step": 19870 }, { "epoch": 0.39, "learning_rate": 4.3573420284791904e-05, "loss": 0.0006, "step": 19872 }, { "epoch": 0.39, "learning_rate": 4.357277332453468e-05, "loss": 0.0041, "step": 19874 }, { "epoch": 0.39, "learning_rate": 4.357212636427744e-05, "loss": 0.0115, "step": 19876 }, { "epoch": 0.39, "learning_rate": 4.357147940402022e-05, "loss": 0.0002, "step": 19878 }, { "epoch": 0.39, "learning_rate": 4.357083244376298e-05, "loss": 0.0228, "step": 19880 }, { "epoch": 0.39, "learning_rate": 4.357018548350575e-05, "loss": 0.0199, "step": 19882 }, { "epoch": 0.39, "learning_rate": 4.356953852324852e-05, "loss": 0.0001, "step": 19884 }, { "epoch": 0.39, "learning_rate": 4.356889156299129e-05, "loss": 0.0001, "step": 19886 }, { "epoch": 0.39, "learning_rate": 4.356824460273406e-05, "loss": 0.0011, "step": 19888 }, { "epoch": 0.39, "learning_rate": 4.3567597642476826e-05, "loss": 0.0103, "step": 19890 }, { "epoch": 0.39, "learning_rate": 4.3566950682219595e-05, "loss": 0.0071, "step": 19892 }, { "epoch": 0.39, "learning_rate": 4.356630372196236e-05, "loss": 0.023, "step": 19894 }, { "epoch": 0.39, "learning_rate": 4.3565656761705133e-05, "loss": 0.0063, "step": 19896 }, { "epoch": 0.39, "learning_rate": 4.3565009801447896e-05, "loss": 0.0087, "step": 19898 }, { "epoch": 0.39, "learning_rate": 4.356436284119067e-05, "loss": 0.0003, "step": 19900 }, { "epoch": 0.39, "learning_rate": 4.3563715880933434e-05, "loss": 0.0002, "step": 19902 }, { "epoch": 0.39, "learning_rate": 4.35630689206762e-05, "loss": 0.0187, "step": 19904 }, { "epoch": 0.39, "learning_rate": 4.356242196041897e-05, "loss": 0.0001, "step": 19906 }, { "epoch": 0.39, "learning_rate": 4.356177500016174e-05, "loss": 0.0003, "step": 19908 }, { "epoch": 0.39, "learning_rate": 4.356112803990451e-05, "loss": 0.0077, "step": 19910 }, { "epoch": 0.39, "learning_rate": 4.356048107964728e-05, "loss": 0.0093, "step": 19912 }, { "epoch": 0.39, "learning_rate": 4.355983411939005e-05, "loss": 0.0009, "step": 19914 }, { "epoch": 0.39, "learning_rate": 4.355918715913282e-05, "loss": 0.0107, "step": 19916 }, { "epoch": 0.39, "learning_rate": 4.3558540198875587e-05, "loss": 0.0002, "step": 19918 }, { "epoch": 0.39, "learning_rate": 4.3557893238618356e-05, "loss": 0.0011, "step": 19920 }, { "epoch": 0.39, "learning_rate": 4.3557246278361125e-05, "loss": 0.0031, "step": 19922 }, { "epoch": 0.39, "learning_rate": 4.3556599318103894e-05, "loss": 0.007, "step": 19924 }, { "epoch": 0.39, "learning_rate": 4.3555952357846656e-05, "loss": 0.0223, "step": 19926 }, { "epoch": 0.39, "learning_rate": 4.355530539758943e-05, "loss": 0.0001, "step": 19928 }, { "epoch": 0.39, "learning_rate": 4.3554658437332194e-05, "loss": 0.0001, "step": 19930 }, { "epoch": 0.39, "learning_rate": 4.355401147707496e-05, "loss": 0.0004, "step": 19932 }, { "epoch": 0.39, "learning_rate": 4.355336451681773e-05, "loss": 0.0001, "step": 19934 }, { "epoch": 0.39, "learning_rate": 4.35527175565605e-05, "loss": 0.0002, "step": 19936 }, { "epoch": 0.39, "learning_rate": 4.355207059630327e-05, "loss": 0.0008, "step": 19938 }, { "epoch": 0.39, "learning_rate": 4.355142363604604e-05, "loss": 0.0053, "step": 19940 }, { "epoch": 0.39, "learning_rate": 4.355077667578881e-05, "loss": 0.0021, "step": 19942 }, { "epoch": 0.39, "learning_rate": 4.355012971553158e-05, "loss": 0.0001, "step": 19944 }, { "epoch": 0.39, "learning_rate": 4.354948275527435e-05, "loss": 0.0044, "step": 19946 }, { "epoch": 0.39, "learning_rate": 4.354883579501711e-05, "loss": 0.0014, "step": 19948 }, { "epoch": 0.39, "learning_rate": 4.3548188834759885e-05, "loss": 0.0001, "step": 19950 }, { "epoch": 0.39, "learning_rate": 4.3547541874502654e-05, "loss": 0.0002, "step": 19952 }, { "epoch": 0.39, "learning_rate": 4.3546894914245417e-05, "loss": 0.0026, "step": 19954 }, { "epoch": 0.39, "learning_rate": 4.354624795398819e-05, "loss": 0.0016, "step": 19956 }, { "epoch": 0.39, "learning_rate": 4.3545600993730955e-05, "loss": 0.0021, "step": 19958 }, { "epoch": 0.39, "learning_rate": 4.354495403347373e-05, "loss": 0.0001, "step": 19960 }, { "epoch": 0.39, "learning_rate": 4.354430707321649e-05, "loss": 0.0361, "step": 19962 }, { "epoch": 0.39, "learning_rate": 4.354366011295926e-05, "loss": 0.0013, "step": 19964 }, { "epoch": 0.39, "learning_rate": 4.354301315270203e-05, "loss": 0.0003, "step": 19966 }, { "epoch": 0.39, "learning_rate": 4.35423661924448e-05, "loss": 0.0073, "step": 19968 }, { "epoch": 0.39, "learning_rate": 4.354171923218757e-05, "loss": 0.0014, "step": 19970 }, { "epoch": 0.39, "learning_rate": 4.354107227193034e-05, "loss": 0.0248, "step": 19972 }, { "epoch": 0.39, "learning_rate": 4.354042531167311e-05, "loss": 0.0384, "step": 19974 }, { "epoch": 0.39, "learning_rate": 4.353977835141587e-05, "loss": 0.007, "step": 19976 }, { "epoch": 0.39, "learning_rate": 4.3539131391158646e-05, "loss": 0.0002, "step": 19978 }, { "epoch": 0.39, "learning_rate": 4.353848443090141e-05, "loss": 0.0003, "step": 19980 }, { "epoch": 0.39, "learning_rate": 4.3537837470644184e-05, "loss": 0.0082, "step": 19982 }, { "epoch": 0.39, "learning_rate": 4.3537190510386946e-05, "loss": 0.0015, "step": 19984 }, { "epoch": 0.39, "learning_rate": 4.3536543550129715e-05, "loss": 0.0001, "step": 19986 }, { "epoch": 0.39, "learning_rate": 4.353589658987249e-05, "loss": 0.0167, "step": 19988 }, { "epoch": 0.39, "learning_rate": 4.353524962961525e-05, "loss": 0.0001, "step": 19990 }, { "epoch": 0.39, "learning_rate": 4.353460266935802e-05, "loss": 0.0149, "step": 19992 }, { "epoch": 0.39, "learning_rate": 4.353395570910079e-05, "loss": 0.002, "step": 19994 }, { "epoch": 0.39, "learning_rate": 4.353330874884356e-05, "loss": 0.0005, "step": 19996 }, { "epoch": 0.39, "learning_rate": 4.353266178858633e-05, "loss": 0.0051, "step": 19998 }, { "epoch": 0.39, "learning_rate": 4.35320148283291e-05, "loss": 0.0003, "step": 20000 }, { "epoch": 0.39, "learning_rate": 4.353136786807187e-05, "loss": 0.0029, "step": 20002 }, { "epoch": 0.39, "learning_rate": 4.353072090781464e-05, "loss": 0.0044, "step": 20004 }, { "epoch": 0.39, "learning_rate": 4.3530073947557406e-05, "loss": 0.0083, "step": 20006 }, { "epoch": 0.39, "learning_rate": 4.352942698730017e-05, "loss": 0.0026, "step": 20008 }, { "epoch": 0.39, "learning_rate": 4.3528780027042944e-05, "loss": 0.0068, "step": 20010 }, { "epoch": 0.39, "learning_rate": 4.3528133066785706e-05, "loss": 0.0033, "step": 20012 }, { "epoch": 0.39, "learning_rate": 4.3527486106528475e-05, "loss": 0.022, "step": 20014 }, { "epoch": 0.39, "learning_rate": 4.3526839146271245e-05, "loss": 0.0001, "step": 20016 }, { "epoch": 0.39, "learning_rate": 4.3526192186014014e-05, "loss": 0.0003, "step": 20018 }, { "epoch": 0.39, "learning_rate": 4.352554522575679e-05, "loss": 0.0002, "step": 20020 }, { "epoch": 0.39, "learning_rate": 4.352489826549955e-05, "loss": 0.0071, "step": 20022 }, { "epoch": 0.39, "learning_rate": 4.352425130524232e-05, "loss": 0.0002, "step": 20024 }, { "epoch": 0.39, "learning_rate": 4.352360434498509e-05, "loss": 0.0018, "step": 20026 }, { "epoch": 0.39, "learning_rate": 4.352295738472786e-05, "loss": 0.0001, "step": 20028 }, { "epoch": 0.39, "learning_rate": 4.352231042447063e-05, "loss": 0.0003, "step": 20030 }, { "epoch": 0.39, "learning_rate": 4.35216634642134e-05, "loss": 0.0002, "step": 20032 }, { "epoch": 0.39, "learning_rate": 4.3521016503956166e-05, "loss": 0.0027, "step": 20034 }, { "epoch": 0.39, "learning_rate": 4.352036954369893e-05, "loss": 0.0006, "step": 20036 }, { "epoch": 0.39, "learning_rate": 4.3519722583441705e-05, "loss": 0.0001, "step": 20038 }, { "epoch": 0.39, "learning_rate": 4.351907562318447e-05, "loss": 0.0038, "step": 20040 }, { "epoch": 0.39, "learning_rate": 4.351842866292724e-05, "loss": 0.0002, "step": 20042 }, { "epoch": 0.39, "learning_rate": 4.3517781702670005e-05, "loss": 0.0063, "step": 20044 }, { "epoch": 0.39, "learning_rate": 4.3517134742412774e-05, "loss": 0.0664, "step": 20046 }, { "epoch": 0.39, "learning_rate": 4.351648778215554e-05, "loss": 0.0083, "step": 20048 }, { "epoch": 0.39, "learning_rate": 4.351584082189831e-05, "loss": 0.0007, "step": 20050 }, { "epoch": 0.39, "learning_rate": 4.351519386164108e-05, "loss": 0.0022, "step": 20052 }, { "epoch": 0.39, "learning_rate": 4.351454690138385e-05, "loss": 0.0001, "step": 20054 }, { "epoch": 0.39, "learning_rate": 4.351389994112662e-05, "loss": 0.0111, "step": 20056 }, { "epoch": 0.39, "learning_rate": 4.351325298086938e-05, "loss": 0.0047, "step": 20058 }, { "epoch": 0.39, "learning_rate": 4.351260602061216e-05, "loss": 0.001, "step": 20060 }, { "epoch": 0.39, "learning_rate": 4.351195906035492e-05, "loss": 0.0033, "step": 20062 }, { "epoch": 0.39, "learning_rate": 4.3511312100097696e-05, "loss": 0.0001, "step": 20064 }, { "epoch": 0.39, "learning_rate": 4.3510665139840465e-05, "loss": 0.0074, "step": 20066 }, { "epoch": 0.39, "learning_rate": 4.351001817958323e-05, "loss": 0.0133, "step": 20068 }, { "epoch": 0.39, "learning_rate": 4.3509371219326e-05, "loss": 0.0011, "step": 20070 }, { "epoch": 0.39, "learning_rate": 4.3508724259068765e-05, "loss": 0.0002, "step": 20072 }, { "epoch": 0.39, "learning_rate": 4.3508077298811534e-05, "loss": 0.0001, "step": 20074 }, { "epoch": 0.39, "learning_rate": 4.3507430338554304e-05, "loss": 0.0001, "step": 20076 }, { "epoch": 0.39, "learning_rate": 4.350678337829707e-05, "loss": 0.0001, "step": 20078 }, { "epoch": 0.39, "learning_rate": 4.350613641803984e-05, "loss": 0.0002, "step": 20080 }, { "epoch": 0.39, "learning_rate": 4.350548945778261e-05, "loss": 0.0019, "step": 20082 }, { "epoch": 0.39, "learning_rate": 4.350484249752538e-05, "loss": 0.0003, "step": 20084 }, { "epoch": 0.39, "learning_rate": 4.350419553726815e-05, "loss": 0.0004, "step": 20086 }, { "epoch": 0.39, "learning_rate": 4.350354857701092e-05, "loss": 0.0027, "step": 20088 }, { "epoch": 0.39, "learning_rate": 4.350290161675368e-05, "loss": 0.0004, "step": 20090 }, { "epoch": 0.39, "learning_rate": 4.3502254656496456e-05, "loss": 0.0052, "step": 20092 }, { "epoch": 0.39, "learning_rate": 4.350160769623922e-05, "loss": 0.0032, "step": 20094 }, { "epoch": 0.39, "learning_rate": 4.350096073598199e-05, "loss": 0.0001, "step": 20096 }, { "epoch": 0.39, "learning_rate": 4.3500313775724763e-05, "loss": 0.0003, "step": 20098 }, { "epoch": 0.39, "learning_rate": 4.3499666815467526e-05, "loss": 0.001, "step": 20100 }, { "epoch": 0.39, "learning_rate": 4.34990198552103e-05, "loss": 0.0217, "step": 20102 }, { "epoch": 0.39, "learning_rate": 4.3498372894953064e-05, "loss": 0.0001, "step": 20104 }, { "epoch": 0.39, "learning_rate": 4.349772593469583e-05, "loss": 0.0007, "step": 20106 }, { "epoch": 0.39, "learning_rate": 4.34970789744386e-05, "loss": 0.0008, "step": 20108 }, { "epoch": 0.39, "learning_rate": 4.349643201418137e-05, "loss": 0.0038, "step": 20110 }, { "epoch": 0.39, "learning_rate": 4.349578505392414e-05, "loss": 0.0005, "step": 20112 }, { "epoch": 0.39, "learning_rate": 4.349513809366691e-05, "loss": 0.0002, "step": 20114 }, { "epoch": 0.39, "learning_rate": 4.349449113340968e-05, "loss": 0.0032, "step": 20116 }, { "epoch": 0.39, "learning_rate": 4.349384417315244e-05, "loss": 0.0004, "step": 20118 }, { "epoch": 0.39, "learning_rate": 4.349319721289522e-05, "loss": 0.0003, "step": 20120 }, { "epoch": 0.39, "learning_rate": 4.349255025263798e-05, "loss": 0.0001, "step": 20122 }, { "epoch": 0.39, "learning_rate": 4.3491903292380755e-05, "loss": 0.0035, "step": 20124 }, { "epoch": 0.39, "learning_rate": 4.349125633212352e-05, "loss": 0.0001, "step": 20126 }, { "epoch": 0.39, "learning_rate": 4.3490609371866286e-05, "loss": 0.0067, "step": 20128 }, { "epoch": 0.39, "learning_rate": 4.348996241160906e-05, "loss": 0.0085, "step": 20130 }, { "epoch": 0.39, "learning_rate": 4.3489315451351824e-05, "loss": 0.0001, "step": 20132 }, { "epoch": 0.39, "learning_rate": 4.3488668491094593e-05, "loss": 0.0005, "step": 20134 }, { "epoch": 0.39, "learning_rate": 4.348802153083736e-05, "loss": 0.0012, "step": 20136 }, { "epoch": 0.39, "learning_rate": 4.348737457058013e-05, "loss": 0.0124, "step": 20138 }, { "epoch": 0.39, "learning_rate": 4.34867276103229e-05, "loss": 0.0002, "step": 20140 }, { "epoch": 0.39, "learning_rate": 4.348608065006567e-05, "loss": 0.0002, "step": 20142 }, { "epoch": 0.39, "learning_rate": 4.348543368980844e-05, "loss": 0.0002, "step": 20144 }, { "epoch": 0.39, "learning_rate": 4.348478672955121e-05, "loss": 0.0057, "step": 20146 }, { "epoch": 0.39, "learning_rate": 4.348413976929398e-05, "loss": 0.0001, "step": 20148 }, { "epoch": 0.39, "learning_rate": 4.348349280903674e-05, "loss": 0.0, "step": 20150 }, { "epoch": 0.39, "learning_rate": 4.3482845848779515e-05, "loss": 0.0001, "step": 20152 }, { "epoch": 0.39, "learning_rate": 4.348219888852228e-05, "loss": 0.0001, "step": 20154 }, { "epoch": 0.39, "learning_rate": 4.3481551928265047e-05, "loss": 0.0004, "step": 20156 }, { "epoch": 0.39, "learning_rate": 4.3480904968007816e-05, "loss": 0.0001, "step": 20158 }, { "epoch": 0.39, "learning_rate": 4.3480258007750585e-05, "loss": 0.0083, "step": 20160 }, { "epoch": 0.39, "learning_rate": 4.3479611047493354e-05, "loss": 0.0023, "step": 20162 }, { "epoch": 0.39, "learning_rate": 4.347896408723612e-05, "loss": 0.0077, "step": 20164 }, { "epoch": 0.39, "learning_rate": 4.347831712697889e-05, "loss": 0.0007, "step": 20166 }, { "epoch": 0.39, "learning_rate": 4.347767016672166e-05, "loss": 0.0004, "step": 20168 }, { "epoch": 0.39, "learning_rate": 4.347702320646443e-05, "loss": 0.0052, "step": 20170 }, { "epoch": 0.39, "learning_rate": 4.347637624620719e-05, "loss": 0.0001, "step": 20172 }, { "epoch": 0.39, "learning_rate": 4.347572928594997e-05, "loss": 0.028, "step": 20174 }, { "epoch": 0.39, "learning_rate": 4.347508232569274e-05, "loss": 0.0002, "step": 20176 }, { "epoch": 0.39, "learning_rate": 4.34744353654355e-05, "loss": 0.0, "step": 20178 }, { "epoch": 0.39, "learning_rate": 4.3473788405178276e-05, "loss": 0.0001, "step": 20180 }, { "epoch": 0.39, "learning_rate": 4.347314144492104e-05, "loss": 0.0026, "step": 20182 }, { "epoch": 0.39, "learning_rate": 4.3472494484663814e-05, "loss": 0.0007, "step": 20184 }, { "epoch": 0.39, "learning_rate": 4.3471847524406576e-05, "loss": 0.0001, "step": 20186 }, { "epoch": 0.39, "learning_rate": 4.3471200564149345e-05, "loss": 0.0005, "step": 20188 }, { "epoch": 0.39, "learning_rate": 4.3470553603892114e-05, "loss": 0.0002, "step": 20190 }, { "epoch": 0.39, "learning_rate": 4.346990664363488e-05, "loss": 0.0046, "step": 20192 }, { "epoch": 0.39, "learning_rate": 4.346925968337765e-05, "loss": 0.0002, "step": 20194 }, { "epoch": 0.39, "learning_rate": 4.346861272312042e-05, "loss": 0.0041, "step": 20196 }, { "epoch": 0.39, "learning_rate": 4.346796576286319e-05, "loss": 0.0005, "step": 20198 }, { "epoch": 0.39, "learning_rate": 4.346731880260595e-05, "loss": 0.0068, "step": 20200 }, { "epoch": 0.39, "learning_rate": 4.346667184234873e-05, "loss": 0.0021, "step": 20202 }, { "epoch": 0.39, "learning_rate": 4.346602488209149e-05, "loss": 0.002, "step": 20204 }, { "epoch": 0.39, "learning_rate": 4.346537792183427e-05, "loss": 0.0025, "step": 20206 }, { "epoch": 0.39, "learning_rate": 4.346473096157703e-05, "loss": 0.0003, "step": 20208 }, { "epoch": 0.39, "learning_rate": 4.34640840013198e-05, "loss": 0.0031, "step": 20210 }, { "epoch": 0.39, "learning_rate": 4.3463437041062574e-05, "loss": 0.0007, "step": 20212 }, { "epoch": 0.39, "learning_rate": 4.3462790080805336e-05, "loss": 0.0048, "step": 20214 }, { "epoch": 0.39, "learning_rate": 4.3462143120548106e-05, "loss": 0.0072, "step": 20216 }, { "epoch": 0.39, "learning_rate": 4.3461496160290875e-05, "loss": 0.0004, "step": 20218 }, { "epoch": 0.39, "learning_rate": 4.3460849200033644e-05, "loss": 0.0009, "step": 20220 }, { "epoch": 0.39, "learning_rate": 4.346020223977641e-05, "loss": 0.0003, "step": 20222 }, { "epoch": 0.39, "learning_rate": 4.345955527951918e-05, "loss": 0.0091, "step": 20224 }, { "epoch": 0.39, "learning_rate": 4.345890831926195e-05, "loss": 0.0001, "step": 20226 }, { "epoch": 0.39, "learning_rate": 4.345826135900472e-05, "loss": 0.0005, "step": 20228 }, { "epoch": 0.39, "learning_rate": 4.345761439874749e-05, "loss": 0.0098, "step": 20230 }, { "epoch": 0.39, "learning_rate": 4.345696743849025e-05, "loss": 0.0009, "step": 20232 }, { "epoch": 0.39, "learning_rate": 4.345632047823303e-05, "loss": 0.0038, "step": 20234 }, { "epoch": 0.39, "learning_rate": 4.345567351797579e-05, "loss": 0.0001, "step": 20236 }, { "epoch": 0.39, "learning_rate": 4.345502655771856e-05, "loss": 0.0002, "step": 20238 }, { "epoch": 0.39, "learning_rate": 4.345437959746133e-05, "loss": 0.0002, "step": 20240 }, { "epoch": 0.39, "learning_rate": 4.34537326372041e-05, "loss": 0.0003, "step": 20242 }, { "epoch": 0.39, "learning_rate": 4.345308567694687e-05, "loss": 0.0004, "step": 20244 }, { "epoch": 0.39, "learning_rate": 4.3452438716689635e-05, "loss": 0.0056, "step": 20246 }, { "epoch": 0.39, "learning_rate": 4.3451791756432404e-05, "loss": 0.0018, "step": 20248 }, { "epoch": 0.39, "learning_rate": 4.345114479617517e-05, "loss": 0.0031, "step": 20250 }, { "epoch": 0.39, "learning_rate": 4.345049783591794e-05, "loss": 0.0004, "step": 20252 }, { "epoch": 0.39, "learning_rate": 4.344985087566071e-05, "loss": 0.0035, "step": 20254 }, { "epoch": 0.39, "learning_rate": 4.344920391540348e-05, "loss": 0.013, "step": 20256 }, { "epoch": 0.39, "learning_rate": 4.344855695514625e-05, "loss": 0.0001, "step": 20258 }, { "epoch": 0.39, "learning_rate": 4.344790999488901e-05, "loss": 0.0104, "step": 20260 }, { "epoch": 0.39, "learning_rate": 4.344726303463179e-05, "loss": 0.0001, "step": 20262 }, { "epoch": 0.39, "learning_rate": 4.344661607437455e-05, "loss": 0.0009, "step": 20264 }, { "epoch": 0.39, "learning_rate": 4.3445969114117326e-05, "loss": 0.0035, "step": 20266 }, { "epoch": 0.39, "learning_rate": 4.344532215386009e-05, "loss": 0.0034, "step": 20268 }, { "epoch": 0.39, "learning_rate": 4.344467519360286e-05, "loss": 0.0001, "step": 20270 }, { "epoch": 0.39, "learning_rate": 4.3444028233345626e-05, "loss": 0.0001, "step": 20272 }, { "epoch": 0.39, "learning_rate": 4.3443381273088395e-05, "loss": 0.0007, "step": 20274 }, { "epoch": 0.39, "learning_rate": 4.3442734312831165e-05, "loss": 0.0008, "step": 20276 }, { "epoch": 0.39, "learning_rate": 4.3442087352573934e-05, "loss": 0.0, "step": 20278 }, { "epoch": 0.39, "learning_rate": 4.34414403923167e-05, "loss": 0.0002, "step": 20280 }, { "epoch": 0.39, "learning_rate": 4.344079343205947e-05, "loss": 0.0026, "step": 20282 }, { "epoch": 0.39, "learning_rate": 4.344014647180224e-05, "loss": 0.0072, "step": 20284 }, { "epoch": 0.39, "learning_rate": 4.3439499511545e-05, "loss": 0.0001, "step": 20286 }, { "epoch": 0.39, "learning_rate": 4.343885255128778e-05, "loss": 0.0001, "step": 20288 }, { "epoch": 0.39, "learning_rate": 4.343820559103055e-05, "loss": 0.0001, "step": 20290 }, { "epoch": 0.39, "learning_rate": 4.343755863077331e-05, "loss": 0.0048, "step": 20292 }, { "epoch": 0.39, "learning_rate": 4.3436911670516086e-05, "loss": 0.0091, "step": 20294 }, { "epoch": 0.39, "learning_rate": 4.343626471025885e-05, "loss": 0.003, "step": 20296 }, { "epoch": 0.39, "learning_rate": 4.343561775000162e-05, "loss": 0.0045, "step": 20298 }, { "epoch": 0.39, "learning_rate": 4.343497078974439e-05, "loss": 0.0024, "step": 20300 }, { "epoch": 0.39, "learning_rate": 4.3434323829487156e-05, "loss": 0.0002, "step": 20302 }, { "epoch": 0.39, "learning_rate": 4.3433676869229925e-05, "loss": 0.0079, "step": 20304 }, { "epoch": 0.39, "learning_rate": 4.3433029908972694e-05, "loss": 0.0118, "step": 20306 }, { "epoch": 0.39, "learning_rate": 4.343238294871546e-05, "loss": 0.0022, "step": 20308 }, { "epoch": 0.39, "learning_rate": 4.343173598845823e-05, "loss": 0.0066, "step": 20310 }, { "epoch": 0.39, "learning_rate": 4.3431089028201e-05, "loss": 0.0001, "step": 20312 }, { "epoch": 0.39, "learning_rate": 4.3430442067943764e-05, "loss": 0.0039, "step": 20314 }, { "epoch": 0.39, "learning_rate": 4.342979510768654e-05, "loss": 0.0096, "step": 20316 }, { "epoch": 0.39, "learning_rate": 4.34291481474293e-05, "loss": 0.0002, "step": 20318 }, { "epoch": 0.39, "learning_rate": 4.342850118717207e-05, "loss": 0.0001, "step": 20320 }, { "epoch": 0.39, "learning_rate": 4.342785422691485e-05, "loss": 0.0027, "step": 20322 }, { "epoch": 0.39, "learning_rate": 4.342720726665761e-05, "loss": 0.0429, "step": 20324 }, { "epoch": 0.39, "learning_rate": 4.3426560306400385e-05, "loss": 0.0007, "step": 20326 }, { "epoch": 0.39, "learning_rate": 4.342591334614315e-05, "loss": 0.0001, "step": 20328 }, { "epoch": 0.39, "learning_rate": 4.3425266385885916e-05, "loss": 0.0018, "step": 20330 }, { "epoch": 0.39, "learning_rate": 4.3424619425628685e-05, "loss": 0.0037, "step": 20332 }, { "epoch": 0.39, "learning_rate": 4.3423972465371454e-05, "loss": 0.0001, "step": 20334 }, { "epoch": 0.39, "learning_rate": 4.3423325505114224e-05, "loss": 0.0017, "step": 20336 }, { "epoch": 0.39, "learning_rate": 4.342267854485699e-05, "loss": 0.0079, "step": 20338 }, { "epoch": 0.39, "learning_rate": 4.342203158459976e-05, "loss": 0.0001, "step": 20340 }, { "epoch": 0.39, "learning_rate": 4.3421384624342524e-05, "loss": 0.0029, "step": 20342 }, { "epoch": 0.39, "learning_rate": 4.34207376640853e-05, "loss": 0.0001, "step": 20344 }, { "epoch": 0.39, "learning_rate": 4.342009070382806e-05, "loss": 0.0014, "step": 20346 }, { "epoch": 0.39, "learning_rate": 4.341944374357084e-05, "loss": 0.0043, "step": 20348 }, { "epoch": 0.39, "learning_rate": 4.34187967833136e-05, "loss": 0.0036, "step": 20350 }, { "epoch": 0.4, "learning_rate": 4.341814982305637e-05, "loss": 0.0008, "step": 20352 }, { "epoch": 0.4, "learning_rate": 4.3417502862799145e-05, "loss": 0.0001, "step": 20354 }, { "epoch": 0.4, "learning_rate": 4.341685590254191e-05, "loss": 0.0002, "step": 20356 }, { "epoch": 0.4, "learning_rate": 4.341620894228468e-05, "loss": 0.0067, "step": 20358 }, { "epoch": 0.4, "learning_rate": 4.3415561982027446e-05, "loss": 0.0067, "step": 20360 }, { "epoch": 0.4, "learning_rate": 4.3414915021770215e-05, "loss": 0.0031, "step": 20362 }, { "epoch": 0.4, "learning_rate": 4.3414268061512984e-05, "loss": 0.0015, "step": 20364 }, { "epoch": 0.4, "learning_rate": 4.341362110125575e-05, "loss": 0.0385, "step": 20366 }, { "epoch": 0.4, "learning_rate": 4.341297414099852e-05, "loss": 0.0007, "step": 20368 }, { "epoch": 0.4, "learning_rate": 4.341232718074129e-05, "loss": 0.0013, "step": 20370 }, { "epoch": 0.4, "learning_rate": 4.341168022048406e-05, "loss": 0.0001, "step": 20372 }, { "epoch": 0.4, "learning_rate": 4.341103326022682e-05, "loss": 0.002, "step": 20374 }, { "epoch": 0.4, "learning_rate": 4.34103862999696e-05, "loss": 0.001, "step": 20376 }, { "epoch": 0.4, "learning_rate": 4.340973933971236e-05, "loss": 0.0002, "step": 20378 }, { "epoch": 0.4, "learning_rate": 4.340909237945513e-05, "loss": 0.0069, "step": 20380 }, { "epoch": 0.4, "learning_rate": 4.34084454191979e-05, "loss": 0.0002, "step": 20382 }, { "epoch": 0.4, "learning_rate": 4.340779845894067e-05, "loss": 0.0085, "step": 20384 }, { "epoch": 0.4, "learning_rate": 4.340715149868344e-05, "loss": 0.0002, "step": 20386 }, { "epoch": 0.4, "learning_rate": 4.3406504538426206e-05, "loss": 0.0001, "step": 20388 }, { "epoch": 0.4, "learning_rate": 4.3405857578168975e-05, "loss": 0.0004, "step": 20390 }, { "epoch": 0.4, "learning_rate": 4.3405210617911744e-05, "loss": 0.0002, "step": 20392 }, { "epoch": 0.4, "learning_rate": 4.3404563657654513e-05, "loss": 0.0024, "step": 20394 }, { "epoch": 0.4, "learning_rate": 4.3403916697397276e-05, "loss": 0.0035, "step": 20396 }, { "epoch": 0.4, "learning_rate": 4.340326973714005e-05, "loss": 0.0021, "step": 20398 }, { "epoch": 0.4, "learning_rate": 4.340262277688282e-05, "loss": 0.0001, "step": 20400 }, { "epoch": 0.4, "learning_rate": 4.340197581662558e-05, "loss": 0.0003, "step": 20402 }, { "epoch": 0.4, "learning_rate": 4.340132885636836e-05, "loss": 0.0001, "step": 20404 }, { "epoch": 0.4, "learning_rate": 4.340068189611112e-05, "loss": 0.0001, "step": 20406 }, { "epoch": 0.4, "learning_rate": 4.34000349358539e-05, "loss": 0.0132, "step": 20408 }, { "epoch": 0.4, "learning_rate": 4.339938797559666e-05, "loss": 0.0035, "step": 20410 }, { "epoch": 0.4, "learning_rate": 4.339874101533943e-05, "loss": 0.0004, "step": 20412 }, { "epoch": 0.4, "learning_rate": 4.33980940550822e-05, "loss": 0.0006, "step": 20414 }, { "epoch": 0.4, "learning_rate": 4.3397447094824967e-05, "loss": 0.0026, "step": 20416 }, { "epoch": 0.4, "learning_rate": 4.3396800134567736e-05, "loss": 0.0019, "step": 20418 }, { "epoch": 0.4, "learning_rate": 4.3396153174310505e-05, "loss": 0.0002, "step": 20420 }, { "epoch": 0.4, "learning_rate": 4.3395506214053274e-05, "loss": 0.0002, "step": 20422 }, { "epoch": 0.4, "learning_rate": 4.339485925379604e-05, "loss": 0.0137, "step": 20424 }, { "epoch": 0.4, "learning_rate": 4.339421229353881e-05, "loss": 0.0001, "step": 20426 }, { "epoch": 0.4, "learning_rate": 4.3393565333281574e-05, "loss": 0.0001, "step": 20428 }, { "epoch": 0.4, "learning_rate": 4.339291837302435e-05, "loss": 0.0033, "step": 20430 }, { "epoch": 0.4, "learning_rate": 4.339227141276712e-05, "loss": 0.0122, "step": 20432 }, { "epoch": 0.4, "learning_rate": 4.339162445250988e-05, "loss": 0.0054, "step": 20434 }, { "epoch": 0.4, "learning_rate": 4.339097749225266e-05, "loss": 0.0021, "step": 20436 }, { "epoch": 0.4, "learning_rate": 4.339033053199542e-05, "loss": 0.0026, "step": 20438 }, { "epoch": 0.4, "learning_rate": 4.338968357173819e-05, "loss": 0.0025, "step": 20440 }, { "epoch": 0.4, "learning_rate": 4.338903661148096e-05, "loss": 0.0024, "step": 20442 }, { "epoch": 0.4, "learning_rate": 4.338838965122373e-05, "loss": 0.0014, "step": 20444 }, { "epoch": 0.4, "learning_rate": 4.3387742690966496e-05, "loss": 0.0001, "step": 20446 }, { "epoch": 0.4, "learning_rate": 4.3387095730709265e-05, "loss": 0.0095, "step": 20448 }, { "epoch": 0.4, "learning_rate": 4.3386448770452034e-05, "loss": 0.0044, "step": 20450 }, { "epoch": 0.4, "learning_rate": 4.33858018101948e-05, "loss": 0.0001, "step": 20452 }, { "epoch": 0.4, "learning_rate": 4.338515484993757e-05, "loss": 0.0005, "step": 20454 }, { "epoch": 0.4, "learning_rate": 4.3384507889680335e-05, "loss": 0.0002, "step": 20456 }, { "epoch": 0.4, "learning_rate": 4.338386092942311e-05, "loss": 0.0026, "step": 20458 }, { "epoch": 0.4, "learning_rate": 4.338321396916587e-05, "loss": 0.0033, "step": 20460 }, { "epoch": 0.4, "learning_rate": 4.338256700890864e-05, "loss": 0.0096, "step": 20462 }, { "epoch": 0.4, "learning_rate": 4.338192004865141e-05, "loss": 0.0003, "step": 20464 }, { "epoch": 0.4, "learning_rate": 4.338127308839418e-05, "loss": 0.0094, "step": 20466 }, { "epoch": 0.4, "learning_rate": 4.3380626128136956e-05, "loss": 0.0005, "step": 20468 }, { "epoch": 0.4, "learning_rate": 4.337997916787972e-05, "loss": 0.0026, "step": 20470 }, { "epoch": 0.4, "learning_rate": 4.337933220762249e-05, "loss": 0.0012, "step": 20472 }, { "epoch": 0.4, "learning_rate": 4.3378685247365256e-05, "loss": 0.0195, "step": 20474 }, { "epoch": 0.4, "learning_rate": 4.3378038287108026e-05, "loss": 0.0131, "step": 20476 }, { "epoch": 0.4, "learning_rate": 4.3377391326850795e-05, "loss": 0.0071, "step": 20478 }, { "epoch": 0.4, "learning_rate": 4.3376744366593564e-05, "loss": 0.0021, "step": 20480 }, { "epoch": 0.4, "learning_rate": 4.337609740633633e-05, "loss": 0.0369, "step": 20482 }, { "epoch": 0.4, "learning_rate": 4.3375450446079095e-05, "loss": 0.0036, "step": 20484 }, { "epoch": 0.4, "learning_rate": 4.337480348582187e-05, "loss": 0.0003, "step": 20486 }, { "epoch": 0.4, "learning_rate": 4.337415652556463e-05, "loss": 0.0001, "step": 20488 }, { "epoch": 0.4, "learning_rate": 4.337350956530741e-05, "loss": 0.0001, "step": 20490 }, { "epoch": 0.4, "learning_rate": 4.337286260505017e-05, "loss": 0.0011, "step": 20492 }, { "epoch": 0.4, "learning_rate": 4.337221564479294e-05, "loss": 0.0012, "step": 20494 }, { "epoch": 0.4, "learning_rate": 4.337156868453571e-05, "loss": 0.0015, "step": 20496 }, { "epoch": 0.4, "learning_rate": 4.337092172427848e-05, "loss": 0.0002, "step": 20498 }, { "epoch": 0.4, "learning_rate": 4.337027476402125e-05, "loss": 0.0007, "step": 20500 }, { "epoch": 0.4, "learning_rate": 4.336962780376402e-05, "loss": 0.0001, "step": 20502 }, { "epoch": 0.4, "learning_rate": 4.3368980843506786e-05, "loss": 0.0001, "step": 20504 }, { "epoch": 0.4, "learning_rate": 4.3368333883249555e-05, "loss": 0.0003, "step": 20506 }, { "epoch": 0.4, "learning_rate": 4.3367686922992324e-05, "loss": 0.0027, "step": 20508 }, { "epoch": 0.4, "learning_rate": 4.336703996273509e-05, "loss": 0.0003, "step": 20510 }, { "epoch": 0.4, "learning_rate": 4.336639300247786e-05, "loss": 0.0002, "step": 20512 }, { "epoch": 0.4, "learning_rate": 4.336574604222063e-05, "loss": 0.0003, "step": 20514 }, { "epoch": 0.4, "learning_rate": 4.3365099081963394e-05, "loss": 0.0131, "step": 20516 }, { "epoch": 0.4, "learning_rate": 4.336445212170617e-05, "loss": 0.0006, "step": 20518 }, { "epoch": 0.4, "learning_rate": 4.336380516144893e-05, "loss": 0.0034, "step": 20520 }, { "epoch": 0.4, "learning_rate": 4.33631582011917e-05, "loss": 0.0004, "step": 20522 }, { "epoch": 0.4, "learning_rate": 4.336251124093447e-05, "loss": 0.0001, "step": 20524 }, { "epoch": 0.4, "learning_rate": 4.336186428067724e-05, "loss": 0.0005, "step": 20526 }, { "epoch": 0.4, "learning_rate": 4.336121732042001e-05, "loss": 0.0021, "step": 20528 }, { "epoch": 0.4, "learning_rate": 4.336057036016278e-05, "loss": 0.0001, "step": 20530 }, { "epoch": 0.4, "learning_rate": 4.3359923399905546e-05, "loss": 0.002, "step": 20532 }, { "epoch": 0.4, "learning_rate": 4.3359276439648315e-05, "loss": 0.0077, "step": 20534 }, { "epoch": 0.4, "learning_rate": 4.3358629479391085e-05, "loss": 0.0004, "step": 20536 }, { "epoch": 0.4, "learning_rate": 4.335798251913385e-05, "loss": 0.0062, "step": 20538 }, { "epoch": 0.4, "learning_rate": 4.335733555887662e-05, "loss": 0.0003, "step": 20540 }, { "epoch": 0.4, "learning_rate": 4.3356688598619385e-05, "loss": 0.0061, "step": 20542 }, { "epoch": 0.4, "learning_rate": 4.3356041638362154e-05, "loss": 0.007, "step": 20544 }, { "epoch": 0.4, "learning_rate": 4.335539467810493e-05, "loss": 0.0224, "step": 20546 }, { "epoch": 0.4, "learning_rate": 4.335474771784769e-05, "loss": 0.0007, "step": 20548 }, { "epoch": 0.4, "learning_rate": 4.335410075759047e-05, "loss": 0.0341, "step": 20550 }, { "epoch": 0.4, "learning_rate": 4.335345379733323e-05, "loss": 0.0019, "step": 20552 }, { "epoch": 0.4, "learning_rate": 4.3352806837076e-05, "loss": 0.011, "step": 20554 }, { "epoch": 0.4, "learning_rate": 4.335215987681877e-05, "loss": 0.0039, "step": 20556 }, { "epoch": 0.4, "learning_rate": 4.335151291656154e-05, "loss": 0.0001, "step": 20558 }, { "epoch": 0.4, "learning_rate": 4.335086595630431e-05, "loss": 0.0073, "step": 20560 }, { "epoch": 0.4, "learning_rate": 4.3350218996047076e-05, "loss": 0.0003, "step": 20562 }, { "epoch": 0.4, "learning_rate": 4.3349572035789845e-05, "loss": 0.0011, "step": 20564 }, { "epoch": 0.4, "learning_rate": 4.3348925075532614e-05, "loss": 0.0002, "step": 20566 }, { "epoch": 0.4, "learning_rate": 4.334827811527538e-05, "loss": 0.0004, "step": 20568 }, { "epoch": 0.4, "learning_rate": 4.3347631155018145e-05, "loss": 0.0001, "step": 20570 }, { "epoch": 0.4, "learning_rate": 4.334698419476092e-05, "loss": 0.0102, "step": 20572 }, { "epoch": 0.4, "learning_rate": 4.3346337234503684e-05, "loss": 0.0, "step": 20574 }, { "epoch": 0.4, "learning_rate": 4.334569027424645e-05, "loss": 0.0189, "step": 20576 }, { "epoch": 0.4, "learning_rate": 4.334504331398923e-05, "loss": 0.0033, "step": 20578 }, { "epoch": 0.4, "learning_rate": 4.334439635373199e-05, "loss": 0.0006, "step": 20580 }, { "epoch": 0.4, "learning_rate": 4.334374939347476e-05, "loss": 0.0066, "step": 20582 }, { "epoch": 0.4, "learning_rate": 4.334310243321753e-05, "loss": 0.0001, "step": 20584 }, { "epoch": 0.4, "learning_rate": 4.33424554729603e-05, "loss": 0.0002, "step": 20586 }, { "epoch": 0.4, "learning_rate": 4.334180851270307e-05, "loss": 0.028, "step": 20588 }, { "epoch": 0.4, "learning_rate": 4.3341161552445836e-05, "loss": 0.0, "step": 20590 }, { "epoch": 0.4, "learning_rate": 4.3340514592188605e-05, "loss": 0.0005, "step": 20592 }, { "epoch": 0.4, "learning_rate": 4.3339867631931374e-05, "loss": 0.0038, "step": 20594 }, { "epoch": 0.4, "learning_rate": 4.3339220671674144e-05, "loss": 0.0007, "step": 20596 }, { "epoch": 0.4, "learning_rate": 4.3338573711416906e-05, "loss": 0.0001, "step": 20598 }, { "epoch": 0.4, "learning_rate": 4.333792675115968e-05, "loss": 0.0012, "step": 20600 }, { "epoch": 0.4, "learning_rate": 4.3337279790902444e-05, "loss": 0.0, "step": 20602 }, { "epoch": 0.4, "learning_rate": 4.333663283064521e-05, "loss": 0.0008, "step": 20604 }, { "epoch": 0.4, "learning_rate": 4.333598587038798e-05, "loss": 0.0029, "step": 20606 }, { "epoch": 0.4, "learning_rate": 4.333533891013075e-05, "loss": 0.0001, "step": 20608 }, { "epoch": 0.4, "learning_rate": 4.333469194987353e-05, "loss": 0.0053, "step": 20610 }, { "epoch": 0.4, "learning_rate": 4.333404498961629e-05, "loss": 0.0008, "step": 20612 }, { "epoch": 0.4, "learning_rate": 4.333339802935906e-05, "loss": 0.0059, "step": 20614 }, { "epoch": 0.4, "learning_rate": 4.333275106910183e-05, "loss": 0.0002, "step": 20616 }, { "epoch": 0.4, "learning_rate": 4.33321041088446e-05, "loss": 0.0088, "step": 20618 }, { "epoch": 0.4, "learning_rate": 4.333145714858736e-05, "loss": 0.0029, "step": 20620 }, { "epoch": 0.4, "learning_rate": 4.3330810188330135e-05, "loss": 0.0099, "step": 20622 }, { "epoch": 0.4, "learning_rate": 4.3330163228072904e-05, "loss": 0.0077, "step": 20624 }, { "epoch": 0.4, "learning_rate": 4.3329516267815666e-05, "loss": 0.0127, "step": 20626 }, { "epoch": 0.4, "learning_rate": 4.332886930755844e-05, "loss": 0.0001, "step": 20628 }, { "epoch": 0.4, "learning_rate": 4.3328222347301204e-05, "loss": 0.0021, "step": 20630 }, { "epoch": 0.4, "learning_rate": 4.332757538704398e-05, "loss": 0.0002, "step": 20632 }, { "epoch": 0.4, "learning_rate": 4.332692842678674e-05, "loss": 0.0007, "step": 20634 }, { "epoch": 0.4, "learning_rate": 4.332628146652951e-05, "loss": 0.0097, "step": 20636 }, { "epoch": 0.4, "learning_rate": 4.332563450627228e-05, "loss": 0.0016, "step": 20638 }, { "epoch": 0.4, "learning_rate": 4.332498754601505e-05, "loss": 0.0186, "step": 20640 }, { "epoch": 0.4, "learning_rate": 4.332434058575782e-05, "loss": 0.0002, "step": 20642 }, { "epoch": 0.4, "learning_rate": 4.332369362550059e-05, "loss": 0.0061, "step": 20644 }, { "epoch": 0.4, "learning_rate": 4.332304666524336e-05, "loss": 0.0005, "step": 20646 }, { "epoch": 0.4, "learning_rate": 4.3322399704986126e-05, "loss": 0.0066, "step": 20648 }, { "epoch": 0.4, "learning_rate": 4.3321752744728895e-05, "loss": 0.0018, "step": 20650 }, { "epoch": 0.4, "learning_rate": 4.332110578447166e-05, "loss": 0.0051, "step": 20652 }, { "epoch": 0.4, "learning_rate": 4.3320458824214433e-05, "loss": 0.0065, "step": 20654 }, { "epoch": 0.4, "learning_rate": 4.33198118639572e-05, "loss": 0.0021, "step": 20656 }, { "epoch": 0.4, "learning_rate": 4.3319164903699965e-05, "loss": 0.0134, "step": 20658 }, { "epoch": 0.4, "learning_rate": 4.331851794344274e-05, "loss": 0.0022, "step": 20660 }, { "epoch": 0.4, "learning_rate": 4.33178709831855e-05, "loss": 0.0006, "step": 20662 }, { "epoch": 0.4, "learning_rate": 4.331722402292827e-05, "loss": 0.0005, "step": 20664 }, { "epoch": 0.4, "learning_rate": 4.331657706267104e-05, "loss": 0.0053, "step": 20666 }, { "epoch": 0.4, "learning_rate": 4.331593010241381e-05, "loss": 0.0004, "step": 20668 }, { "epoch": 0.4, "learning_rate": 4.331528314215658e-05, "loss": 0.0002, "step": 20670 }, { "epoch": 0.4, "learning_rate": 4.331463618189935e-05, "loss": 0.0002, "step": 20672 }, { "epoch": 0.4, "learning_rate": 4.331398922164212e-05, "loss": 0.0043, "step": 20674 }, { "epoch": 0.4, "learning_rate": 4.3313342261384887e-05, "loss": 0.0002, "step": 20676 }, { "epoch": 0.4, "learning_rate": 4.3312695301127656e-05, "loss": 0.0006, "step": 20678 }, { "epoch": 0.4, "learning_rate": 4.331204834087042e-05, "loss": 0.0001, "step": 20680 }, { "epoch": 0.4, "learning_rate": 4.3311401380613194e-05, "loss": 0.0085, "step": 20682 }, { "epoch": 0.4, "learning_rate": 4.3310754420355956e-05, "loss": 0.0011, "step": 20684 }, { "epoch": 0.4, "learning_rate": 4.3310107460098725e-05, "loss": 0.0056, "step": 20686 }, { "epoch": 0.4, "learning_rate": 4.3309460499841494e-05, "loss": 0.0015, "step": 20688 }, { "epoch": 0.4, "learning_rate": 4.330881353958426e-05, "loss": 0.009, "step": 20690 }, { "epoch": 0.4, "learning_rate": 4.330816657932704e-05, "loss": 0.002, "step": 20692 }, { "epoch": 0.4, "learning_rate": 4.33075196190698e-05, "loss": 0.0003, "step": 20694 }, { "epoch": 0.4, "learning_rate": 4.330687265881257e-05, "loss": 0.0004, "step": 20696 }, { "epoch": 0.4, "learning_rate": 4.330622569855534e-05, "loss": 0.0185, "step": 20698 }, { "epoch": 0.4, "learning_rate": 4.330557873829811e-05, "loss": 0.0001, "step": 20700 }, { "epoch": 0.4, "learning_rate": 4.330493177804088e-05, "loss": 0.0002, "step": 20702 }, { "epoch": 0.4, "learning_rate": 4.330428481778365e-05, "loss": 0.0003, "step": 20704 }, { "epoch": 0.4, "learning_rate": 4.3303637857526416e-05, "loss": 0.0098, "step": 20706 }, { "epoch": 0.4, "learning_rate": 4.3302990897269185e-05, "loss": 0.0009, "step": 20708 }, { "epoch": 0.4, "learning_rate": 4.3302343937011954e-05, "loss": 0.0007, "step": 20710 }, { "epoch": 0.4, "learning_rate": 4.3301696976754717e-05, "loss": 0.0077, "step": 20712 }, { "epoch": 0.4, "learning_rate": 4.330105001649749e-05, "loss": 0.0178, "step": 20714 }, { "epoch": 0.4, "learning_rate": 4.3300403056240255e-05, "loss": 0.0002, "step": 20716 }, { "epoch": 0.4, "learning_rate": 4.3299756095983024e-05, "loss": 0.0044, "step": 20718 }, { "epoch": 0.4, "learning_rate": 4.329910913572579e-05, "loss": 0.0001, "step": 20720 }, { "epoch": 0.4, "learning_rate": 4.329846217546856e-05, "loss": 0.0003, "step": 20722 }, { "epoch": 0.4, "learning_rate": 4.329781521521133e-05, "loss": 0.001, "step": 20724 }, { "epoch": 0.4, "learning_rate": 4.32971682549541e-05, "loss": 0.0031, "step": 20726 }, { "epoch": 0.4, "learning_rate": 4.329652129469687e-05, "loss": 0.0001, "step": 20728 }, { "epoch": 0.4, "learning_rate": 4.329587433443964e-05, "loss": 0.0068, "step": 20730 }, { "epoch": 0.4, "learning_rate": 4.329522737418241e-05, "loss": 0.0001, "step": 20732 }, { "epoch": 0.4, "learning_rate": 4.3294580413925176e-05, "loss": 0.0122, "step": 20734 }, { "epoch": 0.4, "learning_rate": 4.3293933453667946e-05, "loss": 0.0163, "step": 20736 }, { "epoch": 0.4, "learning_rate": 4.3293286493410715e-05, "loss": 0.0347, "step": 20738 }, { "epoch": 0.4, "learning_rate": 4.329263953315348e-05, "loss": 0.0, "step": 20740 }, { "epoch": 0.4, "learning_rate": 4.329199257289625e-05, "loss": 0.0116, "step": 20742 }, { "epoch": 0.4, "learning_rate": 4.3291345612639015e-05, "loss": 0.0001, "step": 20744 }, { "epoch": 0.4, "learning_rate": 4.3290698652381784e-05, "loss": 0.0004, "step": 20746 }, { "epoch": 0.4, "learning_rate": 4.329005169212455e-05, "loss": 0.002, "step": 20748 }, { "epoch": 0.4, "learning_rate": 4.328940473186732e-05, "loss": 0.0002, "step": 20750 }, { "epoch": 0.4, "learning_rate": 4.328875777161009e-05, "loss": 0.0088, "step": 20752 }, { "epoch": 0.4, "learning_rate": 4.328811081135286e-05, "loss": 0.0001, "step": 20754 }, { "epoch": 0.4, "learning_rate": 4.328746385109563e-05, "loss": 0.0002, "step": 20756 }, { "epoch": 0.4, "learning_rate": 4.32868168908384e-05, "loss": 0.0011, "step": 20758 }, { "epoch": 0.4, "learning_rate": 4.328616993058117e-05, "loss": 0.0005, "step": 20760 }, { "epoch": 0.4, "learning_rate": 4.328552297032393e-05, "loss": 0.0127, "step": 20762 }, { "epoch": 0.4, "learning_rate": 4.3284876010066706e-05, "loss": 0.0064, "step": 20764 }, { "epoch": 0.4, "learning_rate": 4.328422904980947e-05, "loss": 0.0088, "step": 20766 }, { "epoch": 0.4, "learning_rate": 4.328358208955224e-05, "loss": 0.0023, "step": 20768 }, { "epoch": 0.4, "learning_rate": 4.328293512929501e-05, "loss": 0.0009, "step": 20770 }, { "epoch": 0.4, "learning_rate": 4.3282288169037776e-05, "loss": 0.0, "step": 20772 }, { "epoch": 0.4, "learning_rate": 4.328164120878055e-05, "loss": 0.0007, "step": 20774 }, { "epoch": 0.4, "learning_rate": 4.3280994248523314e-05, "loss": 0.0008, "step": 20776 }, { "epoch": 0.4, "learning_rate": 4.328034728826608e-05, "loss": 0.0065, "step": 20778 }, { "epoch": 0.4, "learning_rate": 4.327970032800885e-05, "loss": 0.0001, "step": 20780 }, { "epoch": 0.4, "learning_rate": 4.327905336775162e-05, "loss": 0.0001, "step": 20782 }, { "epoch": 0.4, "learning_rate": 4.327840640749439e-05, "loss": 0.0006, "step": 20784 }, { "epoch": 0.4, "learning_rate": 4.327775944723716e-05, "loss": 0.0003, "step": 20786 }, { "epoch": 0.4, "learning_rate": 4.327711248697993e-05, "loss": 0.0076, "step": 20788 }, { "epoch": 0.4, "learning_rate": 4.32764655267227e-05, "loss": 0.0042, "step": 20790 }, { "epoch": 0.4, "learning_rate": 4.3275818566465466e-05, "loss": 0.0086, "step": 20792 }, { "epoch": 0.4, "learning_rate": 4.327517160620823e-05, "loss": 0.0006, "step": 20794 }, { "epoch": 0.4, "learning_rate": 4.3274524645951005e-05, "loss": 0.0086, "step": 20796 }, { "epoch": 0.4, "learning_rate": 4.327387768569377e-05, "loss": 0.0057, "step": 20798 }, { "epoch": 0.4, "learning_rate": 4.3273230725436536e-05, "loss": 0.0145, "step": 20800 }, { "epoch": 0.4, "learning_rate": 4.327258376517931e-05, "loss": 0.0017, "step": 20802 }, { "epoch": 0.4, "learning_rate": 4.3271936804922074e-05, "loss": 0.0031, "step": 20804 }, { "epoch": 0.4, "learning_rate": 4.327128984466484e-05, "loss": 0.0109, "step": 20806 }, { "epoch": 0.4, "learning_rate": 4.327064288440761e-05, "loss": 0.0059, "step": 20808 }, { "epoch": 0.4, "learning_rate": 4.326999592415038e-05, "loss": 0.0062, "step": 20810 }, { "epoch": 0.4, "learning_rate": 4.326934896389315e-05, "loss": 0.0013, "step": 20812 }, { "epoch": 0.4, "learning_rate": 4.326870200363592e-05, "loss": 0.0001, "step": 20814 }, { "epoch": 0.4, "learning_rate": 4.326805504337869e-05, "loss": 0.0021, "step": 20816 }, { "epoch": 0.4, "learning_rate": 4.326740808312146e-05, "loss": 0.0001, "step": 20818 }, { "epoch": 0.4, "learning_rate": 4.326676112286423e-05, "loss": 0.0094, "step": 20820 }, { "epoch": 0.4, "learning_rate": 4.326611416260699e-05, "loss": 0.0002, "step": 20822 }, { "epoch": 0.4, "learning_rate": 4.3265467202349765e-05, "loss": 0.0004, "step": 20824 }, { "epoch": 0.4, "learning_rate": 4.326482024209253e-05, "loss": 0.0192, "step": 20826 }, { "epoch": 0.4, "learning_rate": 4.3264173281835296e-05, "loss": 0.0001, "step": 20828 }, { "epoch": 0.4, "learning_rate": 4.3263526321578065e-05, "loss": 0.0182, "step": 20830 }, { "epoch": 0.4, "learning_rate": 4.3262879361320834e-05, "loss": 0.0001, "step": 20832 }, { "epoch": 0.4, "learning_rate": 4.326223240106361e-05, "loss": 0.0001, "step": 20834 }, { "epoch": 0.4, "learning_rate": 4.326158544080637e-05, "loss": 0.0118, "step": 20836 }, { "epoch": 0.4, "learning_rate": 4.326093848054914e-05, "loss": 0.0076, "step": 20838 }, { "epoch": 0.4, "learning_rate": 4.326029152029191e-05, "loss": 0.0, "step": 20840 }, { "epoch": 0.4, "learning_rate": 4.325964456003468e-05, "loss": 0.0009, "step": 20842 }, { "epoch": 0.4, "learning_rate": 4.325899759977744e-05, "loss": 0.002, "step": 20844 }, { "epoch": 0.4, "learning_rate": 4.325835063952022e-05, "loss": 0.0002, "step": 20846 }, { "epoch": 0.4, "learning_rate": 4.325770367926299e-05, "loss": 0.0003, "step": 20848 }, { "epoch": 0.4, "learning_rate": 4.3257056719005756e-05, "loss": 0.004, "step": 20850 }, { "epoch": 0.4, "learning_rate": 4.3256409758748525e-05, "loss": 0.0001, "step": 20852 }, { "epoch": 0.4, "learning_rate": 4.325576279849129e-05, "loss": 0.0002, "step": 20854 }, { "epoch": 0.4, "learning_rate": 4.3255115838234064e-05, "loss": 0.0091, "step": 20856 }, { "epoch": 0.4, "learning_rate": 4.3254468877976826e-05, "loss": 0.0222, "step": 20858 }, { "epoch": 0.4, "learning_rate": 4.3253821917719595e-05, "loss": 0.0068, "step": 20860 }, { "epoch": 0.4, "learning_rate": 4.3253174957462364e-05, "loss": 0.0088, "step": 20862 }, { "epoch": 0.4, "learning_rate": 4.325252799720513e-05, "loss": 0.0049, "step": 20864 }, { "epoch": 0.4, "learning_rate": 4.32518810369479e-05, "loss": 0.0003, "step": 20866 }, { "epoch": 0.41, "learning_rate": 4.325123407669067e-05, "loss": 0.0043, "step": 20868 }, { "epoch": 0.41, "learning_rate": 4.325058711643344e-05, "loss": 0.0206, "step": 20870 }, { "epoch": 0.41, "learning_rate": 4.324994015617621e-05, "loss": 0.0027, "step": 20872 }, { "epoch": 0.41, "learning_rate": 4.324929319591898e-05, "loss": 0.0003, "step": 20874 }, { "epoch": 0.41, "learning_rate": 4.324864623566174e-05, "loss": 0.0083, "step": 20876 }, { "epoch": 0.41, "learning_rate": 4.324799927540452e-05, "loss": 0.0004, "step": 20878 }, { "epoch": 0.41, "learning_rate": 4.3247352315147286e-05, "loss": 0.0002, "step": 20880 }, { "epoch": 0.41, "learning_rate": 4.324670535489005e-05, "loss": 0.0023, "step": 20882 }, { "epoch": 0.41, "learning_rate": 4.3246058394632824e-05, "loss": 0.0001, "step": 20884 }, { "epoch": 0.41, "learning_rate": 4.3245411434375586e-05, "loss": 0.0005, "step": 20886 }, { "epoch": 0.41, "learning_rate": 4.3244764474118355e-05, "loss": 0.0068, "step": 20888 }, { "epoch": 0.41, "learning_rate": 4.3244117513861124e-05, "loss": 0.0019, "step": 20890 }, { "epoch": 0.41, "learning_rate": 4.3243470553603893e-05, "loss": 0.0001, "step": 20892 }, { "epoch": 0.41, "learning_rate": 4.324282359334666e-05, "loss": 0.0002, "step": 20894 }, { "epoch": 0.41, "learning_rate": 4.324217663308943e-05, "loss": 0.0036, "step": 20896 }, { "epoch": 0.41, "learning_rate": 4.32415296728322e-05, "loss": 0.0001, "step": 20898 }, { "epoch": 0.41, "learning_rate": 4.324120619270358e-05, "loss": 0.0283, "step": 20900 }, { "epoch": 0.41, "learning_rate": 4.3240559232446354e-05, "loss": 0.0007, "step": 20902 }, { "epoch": 0.41, "learning_rate": 4.3239912272189123e-05, "loss": 0.0015, "step": 20904 }, { "epoch": 0.41, "learning_rate": 4.323926531193189e-05, "loss": 0.0011, "step": 20906 }, { "epoch": 0.41, "learning_rate": 4.323861835167466e-05, "loss": 0.0071, "step": 20908 }, { "epoch": 0.41, "learning_rate": 4.3237971391417424e-05, "loss": 0.0002, "step": 20910 }, { "epoch": 0.41, "learning_rate": 4.32373244311602e-05, "loss": 0.0005, "step": 20912 }, { "epoch": 0.41, "learning_rate": 4.323667747090296e-05, "loss": 0.0015, "step": 20914 }, { "epoch": 0.41, "learning_rate": 4.323603051064573e-05, "loss": 0.0021, "step": 20916 }, { "epoch": 0.41, "learning_rate": 4.32353835503885e-05, "loss": 0.0001, "step": 20918 }, { "epoch": 0.41, "learning_rate": 4.323473659013127e-05, "loss": 0.0002, "step": 20920 }, { "epoch": 0.41, "learning_rate": 4.323408962987404e-05, "loss": 0.0, "step": 20922 }, { "epoch": 0.41, "learning_rate": 4.323344266961681e-05, "loss": 0.0021, "step": 20924 }, { "epoch": 0.41, "learning_rate": 4.3232795709359577e-05, "loss": 0.0086, "step": 20926 }, { "epoch": 0.41, "learning_rate": 4.3232148749102346e-05, "loss": 0.0067, "step": 20928 }, { "epoch": 0.41, "learning_rate": 4.3231501788845115e-05, "loss": 0.013, "step": 20930 }, { "epoch": 0.41, "learning_rate": 4.323085482858788e-05, "loss": 0.0164, "step": 20932 }, { "epoch": 0.41, "learning_rate": 4.323020786833065e-05, "loss": 0.0148, "step": 20934 }, { "epoch": 0.41, "learning_rate": 4.322956090807342e-05, "loss": 0.001, "step": 20936 }, { "epoch": 0.41, "learning_rate": 4.3228913947816184e-05, "loss": 0.0018, "step": 20938 }, { "epoch": 0.41, "learning_rate": 4.322826698755896e-05, "loss": 0.0002, "step": 20940 }, { "epoch": 0.41, "learning_rate": 4.322762002730172e-05, "loss": 0.01, "step": 20942 }, { "epoch": 0.41, "learning_rate": 4.32269730670445e-05, "loss": 0.0029, "step": 20944 }, { "epoch": 0.41, "learning_rate": 4.322632610678726e-05, "loss": 0.0035, "step": 20946 }, { "epoch": 0.41, "learning_rate": 4.322567914653003e-05, "loss": 0.0017, "step": 20948 }, { "epoch": 0.41, "learning_rate": 4.32250321862728e-05, "loss": 0.0002, "step": 20950 }, { "epoch": 0.41, "learning_rate": 4.322438522601557e-05, "loss": 0.0123, "step": 20952 }, { "epoch": 0.41, "learning_rate": 4.322373826575834e-05, "loss": 0.001, "step": 20954 }, { "epoch": 0.41, "learning_rate": 4.3223091305501106e-05, "loss": 0.0001, "step": 20956 }, { "epoch": 0.41, "learning_rate": 4.3222444345243875e-05, "loss": 0.0001, "step": 20958 }, { "epoch": 0.41, "learning_rate": 4.322179738498664e-05, "loss": 0.0011, "step": 20960 }, { "epoch": 0.41, "learning_rate": 4.322115042472941e-05, "loss": 0.0001, "step": 20962 }, { "epoch": 0.41, "learning_rate": 4.3220503464472176e-05, "loss": 0.0085, "step": 20964 }, { "epoch": 0.41, "learning_rate": 4.321985650421495e-05, "loss": 0.0001, "step": 20966 }, { "epoch": 0.41, "learning_rate": 4.3219209543957714e-05, "loss": 0.0001, "step": 20968 }, { "epoch": 0.41, "learning_rate": 4.321856258370048e-05, "loss": 0.0197, "step": 20970 }, { "epoch": 0.41, "learning_rate": 4.321791562344326e-05, "loss": 0.008, "step": 20972 }, { "epoch": 0.41, "learning_rate": 4.321726866318602e-05, "loss": 0.0003, "step": 20974 }, { "epoch": 0.41, "learning_rate": 4.321662170292879e-05, "loss": 0.0001, "step": 20976 }, { "epoch": 0.41, "learning_rate": 4.321597474267156e-05, "loss": 0.0064, "step": 20978 }, { "epoch": 0.41, "learning_rate": 4.321532778241433e-05, "loss": 0.0002, "step": 20980 }, { "epoch": 0.41, "learning_rate": 4.32146808221571e-05, "loss": 0.0004, "step": 20982 }, { "epoch": 0.41, "learning_rate": 4.3214033861899866e-05, "loss": 0.0001, "step": 20984 }, { "epoch": 0.41, "learning_rate": 4.3213386901642636e-05, "loss": 0.0085, "step": 20986 }, { "epoch": 0.41, "learning_rate": 4.3212739941385405e-05, "loss": 0.0052, "step": 20988 }, { "epoch": 0.41, "learning_rate": 4.3212092981128174e-05, "loss": 0.0011, "step": 20990 }, { "epoch": 0.41, "learning_rate": 4.3211446020870936e-05, "loss": 0.0086, "step": 20992 }, { "epoch": 0.41, "learning_rate": 4.321079906061371e-05, "loss": 0.0002, "step": 20994 }, { "epoch": 0.41, "learning_rate": 4.3210152100356474e-05, "loss": 0.0014, "step": 20996 }, { "epoch": 0.41, "learning_rate": 4.320950514009924e-05, "loss": 0.0001, "step": 20998 }, { "epoch": 0.41, "learning_rate": 4.320885817984201e-05, "loss": 0.0009, "step": 21000 }, { "epoch": 0.41, "learning_rate": 4.320821121958478e-05, "loss": 0.0001, "step": 21002 }, { "epoch": 0.41, "learning_rate": 4.320756425932756e-05, "loss": 0.0002, "step": 21004 }, { "epoch": 0.41, "learning_rate": 4.320691729907032e-05, "loss": 0.0002, "step": 21006 }, { "epoch": 0.41, "learning_rate": 4.320627033881309e-05, "loss": 0.0052, "step": 21008 }, { "epoch": 0.41, "learning_rate": 4.320562337855586e-05, "loss": 0.0003, "step": 21010 }, { "epoch": 0.41, "learning_rate": 4.320497641829863e-05, "loss": 0.031, "step": 21012 }, { "epoch": 0.41, "learning_rate": 4.3204329458041396e-05, "loss": 0.0185, "step": 21014 }, { "epoch": 0.41, "learning_rate": 4.3203682497784165e-05, "loss": 0.0004, "step": 21016 }, { "epoch": 0.41, "learning_rate": 4.3203035537526934e-05, "loss": 0.0001, "step": 21018 }, { "epoch": 0.41, "learning_rate": 4.3202388577269696e-05, "loss": 0.0001, "step": 21020 }, { "epoch": 0.41, "learning_rate": 4.320174161701247e-05, "loss": 0.0028, "step": 21022 }, { "epoch": 0.41, "learning_rate": 4.3201094656755235e-05, "loss": 0.0011, "step": 21024 }, { "epoch": 0.41, "learning_rate": 4.320044769649801e-05, "loss": 0.0014, "step": 21026 }, { "epoch": 0.41, "learning_rate": 4.319980073624077e-05, "loss": 0.0013, "step": 21028 }, { "epoch": 0.41, "learning_rate": 4.319915377598354e-05, "loss": 0.0052, "step": 21030 }, { "epoch": 0.41, "learning_rate": 4.319850681572631e-05, "loss": 0.0001, "step": 21032 }, { "epoch": 0.41, "learning_rate": 4.319785985546908e-05, "loss": 0.0001, "step": 21034 }, { "epoch": 0.41, "learning_rate": 4.319721289521185e-05, "loss": 0.0001, "step": 21036 }, { "epoch": 0.41, "learning_rate": 4.319656593495462e-05, "loss": 0.0026, "step": 21038 }, { "epoch": 0.41, "learning_rate": 4.319591897469739e-05, "loss": 0.0004, "step": 21040 }, { "epoch": 0.41, "learning_rate": 4.319527201444015e-05, "loss": 0.001, "step": 21042 }, { "epoch": 0.41, "learning_rate": 4.3194625054182925e-05, "loss": 0.0001, "step": 21044 }, { "epoch": 0.41, "learning_rate": 4.319397809392569e-05, "loss": 0.0014, "step": 21046 }, { "epoch": 0.41, "learning_rate": 4.3193331133668464e-05, "loss": 0.0003, "step": 21048 }, { "epoch": 0.41, "learning_rate": 4.319268417341123e-05, "loss": 0.0001, "step": 21050 }, { "epoch": 0.41, "learning_rate": 4.3192037213153995e-05, "loss": 0.0035, "step": 21052 }, { "epoch": 0.41, "learning_rate": 4.319139025289677e-05, "loss": 0.0045, "step": 21054 }, { "epoch": 0.41, "learning_rate": 4.319074329263953e-05, "loss": 0.0002, "step": 21056 }, { "epoch": 0.41, "learning_rate": 4.31900963323823e-05, "loss": 0.0004, "step": 21058 }, { "epoch": 0.41, "learning_rate": 4.318944937212507e-05, "loss": 0.0002, "step": 21060 }, { "epoch": 0.41, "learning_rate": 4.318880241186784e-05, "loss": 0.0001, "step": 21062 }, { "epoch": 0.41, "learning_rate": 4.318815545161061e-05, "loss": 0.0001, "step": 21064 }, { "epoch": 0.41, "learning_rate": 4.318750849135338e-05, "loss": 0.0006, "step": 21066 }, { "epoch": 0.41, "learning_rate": 4.318686153109615e-05, "loss": 0.0005, "step": 21068 }, { "epoch": 0.41, "learning_rate": 4.318621457083892e-05, "loss": 0.0086, "step": 21070 }, { "epoch": 0.41, "learning_rate": 4.3185567610581686e-05, "loss": 0.0025, "step": 21072 }, { "epoch": 0.41, "learning_rate": 4.318492065032445e-05, "loss": 0.0016, "step": 21074 }, { "epoch": 0.41, "learning_rate": 4.3184273690067224e-05, "loss": 0.0005, "step": 21076 }, { "epoch": 0.41, "learning_rate": 4.3183626729809986e-05, "loss": 0.0014, "step": 21078 }, { "epoch": 0.41, "learning_rate": 4.3182979769552755e-05, "loss": 0.0024, "step": 21080 }, { "epoch": 0.41, "learning_rate": 4.318233280929553e-05, "loss": 0.0043, "step": 21082 }, { "epoch": 0.41, "learning_rate": 4.3181685849038294e-05, "loss": 0.0007, "step": 21084 }, { "epoch": 0.41, "learning_rate": 4.318103888878107e-05, "loss": 0.0003, "step": 21086 }, { "epoch": 0.41, "learning_rate": 4.318039192852383e-05, "loss": 0.0021, "step": 21088 }, { "epoch": 0.41, "learning_rate": 4.31797449682666e-05, "loss": 0.0037, "step": 21090 }, { "epoch": 0.41, "learning_rate": 4.317909800800937e-05, "loss": 0.0012, "step": 21092 }, { "epoch": 0.41, "learning_rate": 4.317845104775214e-05, "loss": 0.0008, "step": 21094 }, { "epoch": 0.41, "learning_rate": 4.317780408749491e-05, "loss": 0.0013, "step": 21096 }, { "epoch": 0.41, "learning_rate": 4.317715712723768e-05, "loss": 0.0001, "step": 21098 }, { "epoch": 0.41, "learning_rate": 4.3176510166980446e-05, "loss": 0.0072, "step": 21100 }, { "epoch": 0.41, "learning_rate": 4.317586320672321e-05, "loss": 0.0059, "step": 21102 }, { "epoch": 0.41, "learning_rate": 4.3175216246465984e-05, "loss": 0.0003, "step": 21104 }, { "epoch": 0.41, "learning_rate": 4.317456928620875e-05, "loss": 0.0076, "step": 21106 }, { "epoch": 0.41, "learning_rate": 4.317392232595152e-05, "loss": 0.0006, "step": 21108 }, { "epoch": 0.41, "learning_rate": 4.3173275365694285e-05, "loss": 0.0084, "step": 21110 }, { "epoch": 0.41, "learning_rate": 4.3172628405437054e-05, "loss": 0.0022, "step": 21112 }, { "epoch": 0.41, "learning_rate": 4.317198144517983e-05, "loss": 0.0062, "step": 21114 }, { "epoch": 0.41, "learning_rate": 4.317133448492259e-05, "loss": 0.008, "step": 21116 }, { "epoch": 0.41, "learning_rate": 4.317068752466536e-05, "loss": 0.0035, "step": 21118 }, { "epoch": 0.41, "learning_rate": 4.317004056440813e-05, "loss": 0.0039, "step": 21120 }, { "epoch": 0.41, "learning_rate": 4.31693936041509e-05, "loss": 0.0132, "step": 21122 }, { "epoch": 0.41, "learning_rate": 4.316874664389366e-05, "loss": 0.039, "step": 21124 }, { "epoch": 0.41, "learning_rate": 4.316809968363644e-05, "loss": 0.0001, "step": 21126 }, { "epoch": 0.41, "learning_rate": 4.316745272337921e-05, "loss": 0.0008, "step": 21128 }, { "epoch": 0.41, "learning_rate": 4.3166805763121976e-05, "loss": 0.0105, "step": 21130 }, { "epoch": 0.41, "learning_rate": 4.3166158802864745e-05, "loss": 0.0001, "step": 21132 }, { "epoch": 0.41, "learning_rate": 4.316551184260751e-05, "loss": 0.0024, "step": 21134 }, { "epoch": 0.41, "learning_rate": 4.316486488235028e-05, "loss": 0.0133, "step": 21136 }, { "epoch": 0.41, "learning_rate": 4.3164217922093045e-05, "loss": 0.0003, "step": 21138 }, { "epoch": 0.41, "learning_rate": 4.3163570961835814e-05, "loss": 0.0025, "step": 21140 }, { "epoch": 0.41, "learning_rate": 4.3162924001578583e-05, "loss": 0.0088, "step": 21142 }, { "epoch": 0.41, "learning_rate": 4.316227704132135e-05, "loss": 0.0236, "step": 21144 }, { "epoch": 0.41, "learning_rate": 4.316163008106412e-05, "loss": 0.0047, "step": 21146 }, { "epoch": 0.41, "learning_rate": 4.316098312080689e-05, "loss": 0.0043, "step": 21148 }, { "epoch": 0.41, "learning_rate": 4.316033616054966e-05, "loss": 0.0001, "step": 21150 }, { "epoch": 0.41, "learning_rate": 4.315968920029243e-05, "loss": 0.0048, "step": 21152 }, { "epoch": 0.41, "learning_rate": 4.31590422400352e-05, "loss": 0.0002, "step": 21154 }, { "epoch": 0.41, "learning_rate": 4.315839527977796e-05, "loss": 0.0027, "step": 21156 }, { "epoch": 0.41, "learning_rate": 4.3157748319520736e-05, "loss": 0.0001, "step": 21158 }, { "epoch": 0.41, "learning_rate": 4.3157101359263505e-05, "loss": 0.0002, "step": 21160 }, { "epoch": 0.41, "learning_rate": 4.315645439900627e-05, "loss": 0.0031, "step": 21162 }, { "epoch": 0.41, "learning_rate": 4.3155807438749043e-05, "loss": 0.0089, "step": 21164 }, { "epoch": 0.41, "learning_rate": 4.3155160478491806e-05, "loss": 0.0022, "step": 21166 }, { "epoch": 0.41, "learning_rate": 4.315451351823458e-05, "loss": 0.0, "step": 21168 }, { "epoch": 0.41, "learning_rate": 4.3153866557977344e-05, "loss": 0.0002, "step": 21170 }, { "epoch": 0.41, "learning_rate": 4.315321959772011e-05, "loss": 0.0009, "step": 21172 }, { "epoch": 0.41, "learning_rate": 4.315257263746288e-05, "loss": 0.0001, "step": 21174 }, { "epoch": 0.41, "learning_rate": 4.315192567720565e-05, "loss": 0.0126, "step": 21176 }, { "epoch": 0.41, "learning_rate": 4.315127871694842e-05, "loss": 0.0003, "step": 21178 }, { "epoch": 0.41, "learning_rate": 4.315063175669119e-05, "loss": 0.0063, "step": 21180 }, { "epoch": 0.41, "learning_rate": 4.314998479643396e-05, "loss": 0.0002, "step": 21182 }, { "epoch": 0.41, "learning_rate": 4.314933783617672e-05, "loss": 0.004, "step": 21184 }, { "epoch": 0.41, "learning_rate": 4.3148690875919497e-05, "loss": 0.003, "step": 21186 }, { "epoch": 0.41, "learning_rate": 4.314804391566226e-05, "loss": 0.0022, "step": 21188 }, { "epoch": 0.41, "learning_rate": 4.3147396955405035e-05, "loss": 0.0146, "step": 21190 }, { "epoch": 0.41, "learning_rate": 4.3146749995147804e-05, "loss": 0.0126, "step": 21192 }, { "epoch": 0.41, "learning_rate": 4.3146103034890566e-05, "loss": 0.0002, "step": 21194 }, { "epoch": 0.41, "learning_rate": 4.314545607463334e-05, "loss": 0.0133, "step": 21196 }, { "epoch": 0.41, "learning_rate": 4.3144809114376104e-05, "loss": 0.0002, "step": 21198 }, { "epoch": 0.41, "learning_rate": 4.314416215411887e-05, "loss": 0.0001, "step": 21200 }, { "epoch": 0.41, "learning_rate": 4.314351519386164e-05, "loss": 0.0145, "step": 21202 }, { "epoch": 0.41, "learning_rate": 4.314286823360441e-05, "loss": 0.0084, "step": 21204 }, { "epoch": 0.41, "learning_rate": 4.314222127334718e-05, "loss": 0.013, "step": 21206 }, { "epoch": 0.41, "learning_rate": 4.314157431308995e-05, "loss": 0.004, "step": 21208 }, { "epoch": 0.41, "learning_rate": 4.314092735283272e-05, "loss": 0.0011, "step": 21210 }, { "epoch": 0.41, "learning_rate": 4.314028039257549e-05, "loss": 0.0025, "step": 21212 }, { "epoch": 0.41, "learning_rate": 4.313963343231826e-05, "loss": 0.0348, "step": 21214 }, { "epoch": 0.41, "learning_rate": 4.313898647206102e-05, "loss": 0.0006, "step": 21216 }, { "epoch": 0.41, "learning_rate": 4.3138339511803795e-05, "loss": 0.0001, "step": 21218 }, { "epoch": 0.41, "learning_rate": 4.313769255154656e-05, "loss": 0.0027, "step": 21220 }, { "epoch": 0.41, "learning_rate": 4.3137045591289327e-05, "loss": 0.002, "step": 21222 }, { "epoch": 0.41, "learning_rate": 4.3136398631032096e-05, "loss": 0.0004, "step": 21224 }, { "epoch": 0.41, "learning_rate": 4.3135751670774865e-05, "loss": 0.001, "step": 21226 }, { "epoch": 0.41, "learning_rate": 4.313510471051764e-05, "loss": 0.0001, "step": 21228 }, { "epoch": 0.41, "learning_rate": 4.31344577502604e-05, "loss": 0.0007, "step": 21230 }, { "epoch": 0.41, "learning_rate": 4.313381079000317e-05, "loss": 0.0009, "step": 21232 }, { "epoch": 0.41, "learning_rate": 4.313316382974594e-05, "loss": 0.026, "step": 21234 }, { "epoch": 0.41, "learning_rate": 4.313251686948871e-05, "loss": 0.0026, "step": 21236 }, { "epoch": 0.41, "learning_rate": 4.313186990923148e-05, "loss": 0.0, "step": 21238 }, { "epoch": 0.41, "learning_rate": 4.313122294897425e-05, "loss": 0.0, "step": 21240 }, { "epoch": 0.41, "learning_rate": 4.313057598871702e-05, "loss": 0.0003, "step": 21242 }, { "epoch": 0.41, "learning_rate": 4.312992902845978e-05, "loss": 0.0, "step": 21244 }, { "epoch": 0.41, "learning_rate": 4.3129282068202556e-05, "loss": 0.004, "step": 21246 }, { "epoch": 0.41, "learning_rate": 4.312863510794532e-05, "loss": 0.0024, "step": 21248 }, { "epoch": 0.41, "learning_rate": 4.3127988147688094e-05, "loss": 0.0049, "step": 21250 }, { "epoch": 0.41, "learning_rate": 4.3127341187430856e-05, "loss": 0.0001, "step": 21252 }, { "epoch": 0.41, "learning_rate": 4.3126694227173625e-05, "loss": 0.0001, "step": 21254 }, { "epoch": 0.41, "learning_rate": 4.3126047266916394e-05, "loss": 0.0004, "step": 21256 }, { "epoch": 0.41, "learning_rate": 4.312540030665916e-05, "loss": 0.0029, "step": 21258 }, { "epoch": 0.41, "learning_rate": 4.312475334640193e-05, "loss": 0.0015, "step": 21260 }, { "epoch": 0.41, "learning_rate": 4.31241063861447e-05, "loss": 0.0001, "step": 21262 }, { "epoch": 0.41, "learning_rate": 4.312345942588747e-05, "loss": 0.001, "step": 21264 }, { "epoch": 0.41, "learning_rate": 4.312281246563023e-05, "loss": 0.0007, "step": 21266 }, { "epoch": 0.41, "learning_rate": 4.312216550537301e-05, "loss": 0.002, "step": 21268 }, { "epoch": 0.41, "learning_rate": 4.312151854511577e-05, "loss": 0.0009, "step": 21270 }, { "epoch": 0.41, "learning_rate": 4.312087158485855e-05, "loss": 0.0022, "step": 21272 }, { "epoch": 0.41, "learning_rate": 4.3120224624601316e-05, "loss": 0.0001, "step": 21274 }, { "epoch": 0.41, "learning_rate": 4.311957766434408e-05, "loss": 0.0002, "step": 21276 }, { "epoch": 0.41, "learning_rate": 4.3118930704086854e-05, "loss": 0.0001, "step": 21278 }, { "epoch": 0.41, "learning_rate": 4.3118283743829616e-05, "loss": 0.0017, "step": 21280 }, { "epoch": 0.41, "learning_rate": 4.3117636783572386e-05, "loss": 0.007, "step": 21282 }, { "epoch": 0.41, "learning_rate": 4.3116989823315155e-05, "loss": 0.0043, "step": 21284 }, { "epoch": 0.41, "learning_rate": 4.3116342863057924e-05, "loss": 0.0001, "step": 21286 }, { "epoch": 0.41, "learning_rate": 4.311569590280069e-05, "loss": 0.0007, "step": 21288 }, { "epoch": 0.41, "learning_rate": 4.311504894254346e-05, "loss": 0.0128, "step": 21290 }, { "epoch": 0.41, "learning_rate": 4.311440198228623e-05, "loss": 0.0026, "step": 21292 }, { "epoch": 0.41, "learning_rate": 4.3113755022029e-05, "loss": 0.0087, "step": 21294 }, { "epoch": 0.41, "learning_rate": 4.311310806177177e-05, "loss": 0.0221, "step": 21296 }, { "epoch": 0.41, "learning_rate": 4.311246110151453e-05, "loss": 0.0009, "step": 21298 }, { "epoch": 0.41, "learning_rate": 4.311181414125731e-05, "loss": 0.0001, "step": 21300 }, { "epoch": 0.41, "learning_rate": 4.311116718100007e-05, "loss": 0.0038, "step": 21302 }, { "epoch": 0.41, "learning_rate": 4.311052022074284e-05, "loss": 0.0002, "step": 21304 }, { "epoch": 0.41, "learning_rate": 4.3109873260485615e-05, "loss": 0.0005, "step": 21306 }, { "epoch": 0.41, "learning_rate": 4.310922630022838e-05, "loss": 0.0003, "step": 21308 }, { "epoch": 0.41, "learning_rate": 4.310857933997115e-05, "loss": 0.0244, "step": 21310 }, { "epoch": 0.41, "learning_rate": 4.3107932379713915e-05, "loss": 0.0004, "step": 21312 }, { "epoch": 0.41, "learning_rate": 4.3107285419456684e-05, "loss": 0.0003, "step": 21314 }, { "epoch": 0.41, "learning_rate": 4.310663845919945e-05, "loss": 0.0056, "step": 21316 }, { "epoch": 0.41, "learning_rate": 4.310599149894222e-05, "loss": 0.0001, "step": 21318 }, { "epoch": 0.41, "learning_rate": 4.310534453868499e-05, "loss": 0.0006, "step": 21320 }, { "epoch": 0.41, "learning_rate": 4.310469757842776e-05, "loss": 0.0008, "step": 21322 }, { "epoch": 0.41, "learning_rate": 4.310405061817053e-05, "loss": 0.0157, "step": 21324 }, { "epoch": 0.41, "learning_rate": 4.310340365791329e-05, "loss": 0.001, "step": 21326 }, { "epoch": 0.41, "learning_rate": 4.310275669765607e-05, "loss": 0.0012, "step": 21328 }, { "epoch": 0.41, "learning_rate": 4.310210973739883e-05, "loss": 0.0129, "step": 21330 }, { "epoch": 0.41, "learning_rate": 4.3101462777141606e-05, "loss": 0.0439, "step": 21332 }, { "epoch": 0.41, "learning_rate": 4.310081581688437e-05, "loss": 0.0001, "step": 21334 }, { "epoch": 0.41, "learning_rate": 4.310016885662714e-05, "loss": 0.0, "step": 21336 }, { "epoch": 0.41, "learning_rate": 4.309952189636991e-05, "loss": 0.0002, "step": 21338 }, { "epoch": 0.41, "learning_rate": 4.3098874936112675e-05, "loss": 0.007, "step": 21340 }, { "epoch": 0.41, "learning_rate": 4.3098227975855444e-05, "loss": 0.0099, "step": 21342 }, { "epoch": 0.41, "learning_rate": 4.3097581015598214e-05, "loss": 0.0048, "step": 21344 }, { "epoch": 0.41, "learning_rate": 4.309693405534098e-05, "loss": 0.0129, "step": 21346 }, { "epoch": 0.41, "learning_rate": 4.309628709508375e-05, "loss": 0.0193, "step": 21348 }, { "epoch": 0.41, "learning_rate": 4.309564013482652e-05, "loss": 0.0027, "step": 21350 }, { "epoch": 0.41, "learning_rate": 4.309499317456929e-05, "loss": 0.0027, "step": 21352 }, { "epoch": 0.41, "learning_rate": 4.309434621431206e-05, "loss": 0.0016, "step": 21354 }, { "epoch": 0.41, "learning_rate": 4.309369925405483e-05, "loss": 0.0014, "step": 21356 }, { "epoch": 0.41, "learning_rate": 4.309305229379759e-05, "loss": 0.0002, "step": 21358 }, { "epoch": 0.41, "learning_rate": 4.3092405333540366e-05, "loss": 0.0026, "step": 21360 }, { "epoch": 0.41, "learning_rate": 4.309175837328313e-05, "loss": 0.0001, "step": 21362 }, { "epoch": 0.41, "learning_rate": 4.30911114130259e-05, "loss": 0.0002, "step": 21364 }, { "epoch": 0.41, "learning_rate": 4.309046445276867e-05, "loss": 0.0178, "step": 21366 }, { "epoch": 0.41, "learning_rate": 4.3089817492511436e-05, "loss": 0.0001, "step": 21368 }, { "epoch": 0.41, "learning_rate": 4.308917053225421e-05, "loss": 0.0004, "step": 21370 }, { "epoch": 0.41, "learning_rate": 4.3088523571996974e-05, "loss": 0.0001, "step": 21372 }, { "epoch": 0.41, "learning_rate": 4.308787661173974e-05, "loss": 0.0002, "step": 21374 }, { "epoch": 0.41, "learning_rate": 4.308722965148251e-05, "loss": 0.0053, "step": 21376 }, { "epoch": 0.41, "learning_rate": 4.308658269122528e-05, "loss": 0.0024, "step": 21378 }, { "epoch": 0.41, "learning_rate": 4.3085935730968044e-05, "loss": 0.0154, "step": 21380 }, { "epoch": 0.41, "learning_rate": 4.308528877071082e-05, "loss": 0.0001, "step": 21382 }, { "epoch": 0.42, "learning_rate": 4.308464181045359e-05, "loss": 0.0225, "step": 21384 }, { "epoch": 0.42, "learning_rate": 4.308399485019635e-05, "loss": 0.0067, "step": 21386 }, { "epoch": 0.42, "learning_rate": 4.308334788993913e-05, "loss": 0.0067, "step": 21388 }, { "epoch": 0.42, "learning_rate": 4.308270092968189e-05, "loss": 0.0051, "step": 21390 }, { "epoch": 0.42, "learning_rate": 4.3082053969424665e-05, "loss": 0.0048, "step": 21392 }, { "epoch": 0.42, "learning_rate": 4.308140700916743e-05, "loss": 0.0048, "step": 21394 }, { "epoch": 0.42, "learning_rate": 4.3080760048910196e-05, "loss": 0.0001, "step": 21396 }, { "epoch": 0.42, "learning_rate": 4.3080113088652965e-05, "loss": 0.0001, "step": 21398 }, { "epoch": 0.42, "learning_rate": 4.3079466128395734e-05, "loss": 0.0001, "step": 21400 }, { "epoch": 0.42, "learning_rate": 4.3078819168138503e-05, "loss": 0.001, "step": 21402 }, { "epoch": 0.42, "learning_rate": 4.307817220788127e-05, "loss": 0.0015, "step": 21404 }, { "epoch": 0.42, "learning_rate": 4.307752524762404e-05, "loss": 0.0142, "step": 21406 }, { "epoch": 0.42, "learning_rate": 4.3076878287366804e-05, "loss": 0.006, "step": 21408 }, { "epoch": 0.42, "learning_rate": 4.307623132710958e-05, "loss": 0.0094, "step": 21410 }, { "epoch": 0.42, "learning_rate": 4.307558436685234e-05, "loss": 0.0024, "step": 21412 }, { "epoch": 0.42, "learning_rate": 4.307493740659512e-05, "loss": 0.0004, "step": 21414 }, { "epoch": 0.42, "learning_rate": 4.307429044633789e-05, "loss": 0.0, "step": 21416 }, { "epoch": 0.42, "learning_rate": 4.307364348608065e-05, "loss": 0.0013, "step": 21418 }, { "epoch": 0.42, "learning_rate": 4.3072996525823425e-05, "loss": 0.0025, "step": 21420 }, { "epoch": 0.42, "learning_rate": 4.307234956556619e-05, "loss": 0.0001, "step": 21422 }, { "epoch": 0.42, "learning_rate": 4.307170260530896e-05, "loss": 0.0011, "step": 21424 }, { "epoch": 0.42, "learning_rate": 4.3071055645051726e-05, "loss": 0.0002, "step": 21426 }, { "epoch": 0.42, "learning_rate": 4.3070408684794495e-05, "loss": 0.0046, "step": 21428 }, { "epoch": 0.42, "learning_rate": 4.3069761724537264e-05, "loss": 0.0002, "step": 21430 }, { "epoch": 0.42, "learning_rate": 4.306911476428003e-05, "loss": 0.0096, "step": 21432 }, { "epoch": 0.42, "learning_rate": 4.30684678040228e-05, "loss": 0.0071, "step": 21434 }, { "epoch": 0.42, "learning_rate": 4.306782084376557e-05, "loss": 0.0001, "step": 21436 }, { "epoch": 0.42, "learning_rate": 4.306717388350834e-05, "loss": 0.002, "step": 21438 }, { "epoch": 0.42, "learning_rate": 4.30665269232511e-05, "loss": 0.0012, "step": 21440 }, { "epoch": 0.42, "learning_rate": 4.306587996299388e-05, "loss": 0.0019, "step": 21442 }, { "epoch": 0.42, "learning_rate": 4.306523300273664e-05, "loss": 0.0001, "step": 21444 }, { "epoch": 0.42, "learning_rate": 4.306458604247941e-05, "loss": 0.0012, "step": 21446 }, { "epoch": 0.42, "learning_rate": 4.306393908222218e-05, "loss": 0.0004, "step": 21448 }, { "epoch": 0.42, "learning_rate": 4.306329212196495e-05, "loss": 0.0239, "step": 21450 }, { "epoch": 0.42, "learning_rate": 4.3062645161707724e-05, "loss": 0.0027, "step": 21452 }, { "epoch": 0.42, "learning_rate": 4.3061998201450486e-05, "loss": 0.0001, "step": 21454 }, { "epoch": 0.42, "learning_rate": 4.3061351241193255e-05, "loss": 0.0134, "step": 21456 }, { "epoch": 0.42, "learning_rate": 4.3060704280936024e-05, "loss": 0.0003, "step": 21458 }, { "epoch": 0.42, "learning_rate": 4.306005732067879e-05, "loss": 0.0001, "step": 21460 }, { "epoch": 0.42, "learning_rate": 4.305941036042156e-05, "loss": 0.0005, "step": 21462 }, { "epoch": 0.42, "learning_rate": 4.305876340016433e-05, "loss": 0.0001, "step": 21464 }, { "epoch": 0.42, "learning_rate": 4.30581164399071e-05, "loss": 0.0014, "step": 21466 }, { "epoch": 0.42, "learning_rate": 4.305746947964986e-05, "loss": 0.0001, "step": 21468 }, { "epoch": 0.42, "learning_rate": 4.305682251939264e-05, "loss": 0.0002, "step": 21470 }, { "epoch": 0.42, "learning_rate": 4.30561755591354e-05, "loss": 0.0043, "step": 21472 }, { "epoch": 0.42, "learning_rate": 4.305552859887818e-05, "loss": 0.0074, "step": 21474 }, { "epoch": 0.42, "learning_rate": 4.305488163862094e-05, "loss": 0.0074, "step": 21476 }, { "epoch": 0.42, "learning_rate": 4.305423467836371e-05, "loss": 0.0155, "step": 21478 }, { "epoch": 0.42, "learning_rate": 4.305358771810648e-05, "loss": 0.0005, "step": 21480 }, { "epoch": 0.42, "learning_rate": 4.3052940757849247e-05, "loss": 0.0154, "step": 21482 }, { "epoch": 0.42, "learning_rate": 4.3052293797592016e-05, "loss": 0.0001, "step": 21484 }, { "epoch": 0.42, "learning_rate": 4.3051646837334785e-05, "loss": 0.0002, "step": 21486 }, { "epoch": 0.42, "learning_rate": 4.3050999877077554e-05, "loss": 0.0004, "step": 21488 }, { "epoch": 0.42, "learning_rate": 4.305035291682032e-05, "loss": 0.0047, "step": 21490 }, { "epoch": 0.42, "learning_rate": 4.304970595656309e-05, "loss": 0.0013, "step": 21492 }, { "epoch": 0.42, "learning_rate": 4.304905899630586e-05, "loss": 0.0008, "step": 21494 }, { "epoch": 0.42, "learning_rate": 4.304841203604863e-05, "loss": 0.0075, "step": 21496 }, { "epoch": 0.42, "learning_rate": 4.30477650757914e-05, "loss": 0.0003, "step": 21498 }, { "epoch": 0.42, "learning_rate": 4.304711811553416e-05, "loss": 0.0072, "step": 21500 }, { "epoch": 0.42, "learning_rate": 4.304647115527694e-05, "loss": 0.0003, "step": 21502 }, { "epoch": 0.42, "learning_rate": 4.30458241950197e-05, "loss": 0.001, "step": 21504 }, { "epoch": 0.42, "learning_rate": 4.304517723476247e-05, "loss": 0.0003, "step": 21506 }, { "epoch": 0.42, "learning_rate": 4.304453027450524e-05, "loss": 0.0001, "step": 21508 }, { "epoch": 0.42, "learning_rate": 4.304388331424801e-05, "loss": 0.0017, "step": 21510 }, { "epoch": 0.42, "learning_rate": 4.3043236353990776e-05, "loss": 0.0062, "step": 21512 }, { "epoch": 0.42, "learning_rate": 4.3042589393733545e-05, "loss": 0.002, "step": 21514 }, { "epoch": 0.42, "learning_rate": 4.3041942433476314e-05, "loss": 0.0032, "step": 21516 }, { "epoch": 0.42, "learning_rate": 4.304129547321908e-05, "loss": 0.0004, "step": 21518 }, { "epoch": 0.42, "learning_rate": 4.304064851296185e-05, "loss": 0.0037, "step": 21520 }, { "epoch": 0.42, "learning_rate": 4.3040001552704615e-05, "loss": 0.0092, "step": 21522 }, { "epoch": 0.42, "learning_rate": 4.303935459244739e-05, "loss": 0.0137, "step": 21524 }, { "epoch": 0.42, "learning_rate": 4.303870763219015e-05, "loss": 0.0001, "step": 21526 }, { "epoch": 0.42, "learning_rate": 4.303806067193292e-05, "loss": 0.0047, "step": 21528 }, { "epoch": 0.42, "learning_rate": 4.30374137116757e-05, "loss": 0.008, "step": 21530 }, { "epoch": 0.42, "learning_rate": 4.303676675141846e-05, "loss": 0.0004, "step": 21532 }, { "epoch": 0.42, "learning_rate": 4.3036119791161236e-05, "loss": 0.0072, "step": 21534 }, { "epoch": 0.42, "learning_rate": 4.3035472830904e-05, "loss": 0.0005, "step": 21536 }, { "epoch": 0.42, "learning_rate": 4.303482587064677e-05, "loss": 0.0243, "step": 21538 }, { "epoch": 0.42, "learning_rate": 4.3034178910389536e-05, "loss": 0.0002, "step": 21540 }, { "epoch": 0.42, "learning_rate": 4.3033531950132305e-05, "loss": 0.0006, "step": 21542 }, { "epoch": 0.42, "learning_rate": 4.3032884989875075e-05, "loss": 0.012, "step": 21544 }, { "epoch": 0.42, "learning_rate": 4.3032238029617844e-05, "loss": 0.0002, "step": 21546 }, { "epoch": 0.42, "learning_rate": 4.303159106936061e-05, "loss": 0.0018, "step": 21548 }, { "epoch": 0.42, "learning_rate": 4.3030944109103375e-05, "loss": 0.009, "step": 21550 }, { "epoch": 0.42, "learning_rate": 4.303029714884615e-05, "loss": 0.0001, "step": 21552 }, { "epoch": 0.42, "learning_rate": 4.302965018858891e-05, "loss": 0.0003, "step": 21554 }, { "epoch": 0.42, "learning_rate": 4.302900322833169e-05, "loss": 0.0086, "step": 21556 }, { "epoch": 0.42, "learning_rate": 4.302835626807445e-05, "loss": 0.0001, "step": 21558 }, { "epoch": 0.42, "learning_rate": 4.302770930781722e-05, "loss": 0.0004, "step": 21560 }, { "epoch": 0.42, "learning_rate": 4.3027062347559996e-05, "loss": 0.0135, "step": 21562 }, { "epoch": 0.42, "learning_rate": 4.302641538730276e-05, "loss": 0.0001, "step": 21564 }, { "epoch": 0.42, "learning_rate": 4.302576842704553e-05, "loss": 0.0007, "step": 21566 }, { "epoch": 0.42, "learning_rate": 4.30251214667883e-05, "loss": 0.008, "step": 21568 }, { "epoch": 0.42, "learning_rate": 4.3024474506531066e-05, "loss": 0.0001, "step": 21570 }, { "epoch": 0.42, "learning_rate": 4.3023827546273835e-05, "loss": 0.0008, "step": 21572 }, { "epoch": 0.42, "learning_rate": 4.3023180586016604e-05, "loss": 0.0019, "step": 21574 }, { "epoch": 0.42, "learning_rate": 4.302253362575937e-05, "loss": 0.0001, "step": 21576 }, { "epoch": 0.42, "learning_rate": 4.302188666550214e-05, "loss": 0.0072, "step": 21578 }, { "epoch": 0.42, "learning_rate": 4.302123970524491e-05, "loss": 0.0013, "step": 21580 }, { "epoch": 0.42, "learning_rate": 4.3020592744987674e-05, "loss": 0.0004, "step": 21582 }, { "epoch": 0.42, "learning_rate": 4.301994578473045e-05, "loss": 0.0003, "step": 21584 }, { "epoch": 0.42, "learning_rate": 4.301929882447321e-05, "loss": 0.006, "step": 21586 }, { "epoch": 0.42, "learning_rate": 4.301865186421598e-05, "loss": 0.0002, "step": 21588 }, { "epoch": 0.42, "learning_rate": 4.301800490395875e-05, "loss": 0.0029, "step": 21590 }, { "epoch": 0.42, "learning_rate": 4.301735794370152e-05, "loss": 0.0038, "step": 21592 }, { "epoch": 0.42, "learning_rate": 4.3016710983444295e-05, "loss": 0.0137, "step": 21594 }, { "epoch": 0.42, "learning_rate": 4.301606402318706e-05, "loss": 0.0083, "step": 21596 }, { "epoch": 0.42, "learning_rate": 4.3015417062929826e-05, "loss": 0.0043, "step": 21598 }, { "epoch": 0.42, "learning_rate": 4.3014770102672595e-05, "loss": 0.0002, "step": 21600 }, { "epoch": 0.42, "learning_rate": 4.3014123142415364e-05, "loss": 0.0001, "step": 21602 }, { "epoch": 0.42, "learning_rate": 4.301347618215813e-05, "loss": 0.0004, "step": 21604 }, { "epoch": 0.42, "learning_rate": 4.30128292219009e-05, "loss": 0.0004, "step": 21606 }, { "epoch": 0.42, "learning_rate": 4.301218226164367e-05, "loss": 0.0027, "step": 21608 }, { "epoch": 0.42, "learning_rate": 4.3011535301386434e-05, "loss": 0.0039, "step": 21610 }, { "epoch": 0.42, "learning_rate": 4.301088834112921e-05, "loss": 0.0072, "step": 21612 }, { "epoch": 0.42, "learning_rate": 4.301024138087197e-05, "loss": 0.0528, "step": 21614 }, { "epoch": 0.42, "learning_rate": 4.300959442061475e-05, "loss": 0.0002, "step": 21616 }, { "epoch": 0.42, "learning_rate": 4.300894746035751e-05, "loss": 0.0046, "step": 21618 }, { "epoch": 0.42, "learning_rate": 4.300830050010028e-05, "loss": 0.0001, "step": 21620 }, { "epoch": 0.42, "learning_rate": 4.300765353984305e-05, "loss": 0.0012, "step": 21622 }, { "epoch": 0.42, "learning_rate": 4.300700657958582e-05, "loss": 0.0002, "step": 21624 }, { "epoch": 0.42, "learning_rate": 4.300635961932859e-05, "loss": 0.0258, "step": 21626 }, { "epoch": 0.42, "learning_rate": 4.3005712659071356e-05, "loss": 0.0082, "step": 21628 }, { "epoch": 0.42, "learning_rate": 4.3005065698814125e-05, "loss": 0.0063, "step": 21630 }, { "epoch": 0.42, "learning_rate": 4.300441873855689e-05, "loss": 0.0038, "step": 21632 }, { "epoch": 0.42, "learning_rate": 4.300377177829966e-05, "loss": 0.0001, "step": 21634 }, { "epoch": 0.42, "learning_rate": 4.3003124818042425e-05, "loss": 0.0002, "step": 21636 }, { "epoch": 0.42, "learning_rate": 4.30024778577852e-05, "loss": 0.0049, "step": 21638 }, { "epoch": 0.42, "learning_rate": 4.300183089752797e-05, "loss": 0.0439, "step": 21640 }, { "epoch": 0.42, "learning_rate": 4.300118393727073e-05, "loss": 0.0013, "step": 21642 }, { "epoch": 0.42, "learning_rate": 4.300053697701351e-05, "loss": 0.0059, "step": 21644 }, { "epoch": 0.42, "learning_rate": 4.299989001675627e-05, "loss": 0.0003, "step": 21646 }, { "epoch": 0.42, "learning_rate": 4.299924305649904e-05, "loss": 0.011, "step": 21648 }, { "epoch": 0.42, "learning_rate": 4.299859609624181e-05, "loss": 0.0023, "step": 21650 }, { "epoch": 0.42, "learning_rate": 4.299794913598458e-05, "loss": 0.0006, "step": 21652 }, { "epoch": 0.42, "learning_rate": 4.299730217572735e-05, "loss": 0.0285, "step": 21654 }, { "epoch": 0.42, "learning_rate": 4.2996655215470116e-05, "loss": 0.0003, "step": 21656 }, { "epoch": 0.42, "learning_rate": 4.2996008255212885e-05, "loss": 0.0047, "step": 21658 }, { "epoch": 0.42, "learning_rate": 4.2995361294955654e-05, "loss": 0.0081, "step": 21660 }, { "epoch": 0.42, "learning_rate": 4.2994714334698423e-05, "loss": 0.0, "step": 21662 }, { "epoch": 0.42, "learning_rate": 4.2994067374441186e-05, "loss": 0.0126, "step": 21664 }, { "epoch": 0.42, "learning_rate": 4.299342041418396e-05, "loss": 0.0092, "step": 21666 }, { "epoch": 0.42, "learning_rate": 4.2992773453926724e-05, "loss": 0.0029, "step": 21668 }, { "epoch": 0.42, "learning_rate": 4.299212649366949e-05, "loss": 0.007, "step": 21670 }, { "epoch": 0.42, "learning_rate": 4.299147953341227e-05, "loss": 0.0001, "step": 21672 }, { "epoch": 0.42, "learning_rate": 4.299083257315503e-05, "loss": 0.0017, "step": 21674 }, { "epoch": 0.42, "learning_rate": 4.299018561289781e-05, "loss": 0.0082, "step": 21676 }, { "epoch": 0.42, "learning_rate": 4.298953865264057e-05, "loss": 0.0045, "step": 21678 }, { "epoch": 0.42, "learning_rate": 4.298889169238334e-05, "loss": 0.0007, "step": 21680 }, { "epoch": 0.42, "learning_rate": 4.298824473212611e-05, "loss": 0.0017, "step": 21682 }, { "epoch": 0.42, "learning_rate": 4.2987597771868877e-05, "loss": 0.0017, "step": 21684 }, { "epoch": 0.42, "learning_rate": 4.2986950811611646e-05, "loss": 0.0064, "step": 21686 }, { "epoch": 0.42, "learning_rate": 4.2986303851354415e-05, "loss": 0.0118, "step": 21688 }, { "epoch": 0.42, "learning_rate": 4.2985656891097184e-05, "loss": 0.0016, "step": 21690 }, { "epoch": 0.42, "learning_rate": 4.2985009930839946e-05, "loss": 0.0031, "step": 21692 }, { "epoch": 0.42, "learning_rate": 4.298436297058272e-05, "loss": 0.0052, "step": 21694 }, { "epoch": 0.42, "learning_rate": 4.2983716010325484e-05, "loss": 0.0001, "step": 21696 }, { "epoch": 0.42, "learning_rate": 4.298306905006826e-05, "loss": 0.0049, "step": 21698 }, { "epoch": 0.42, "learning_rate": 4.298242208981102e-05, "loss": 0.0005, "step": 21700 }, { "epoch": 0.42, "learning_rate": 4.298177512955379e-05, "loss": 0.0023, "step": 21702 }, { "epoch": 0.42, "learning_rate": 4.298112816929656e-05, "loss": 0.0009, "step": 21704 }, { "epoch": 0.42, "learning_rate": 4.298048120903933e-05, "loss": 0.0017, "step": 21706 }, { "epoch": 0.42, "learning_rate": 4.29798342487821e-05, "loss": 0.0001, "step": 21708 }, { "epoch": 0.42, "learning_rate": 4.297918728852487e-05, "loss": 0.0018, "step": 21710 }, { "epoch": 0.42, "learning_rate": 4.297854032826764e-05, "loss": 0.0067, "step": 21712 }, { "epoch": 0.42, "learning_rate": 4.2977893368010406e-05, "loss": 0.0003, "step": 21714 }, { "epoch": 0.42, "learning_rate": 4.2977246407753175e-05, "loss": 0.0114, "step": 21716 }, { "epoch": 0.42, "learning_rate": 4.2976599447495944e-05, "loss": 0.0, "step": 21718 }, { "epoch": 0.42, "learning_rate": 4.297595248723871e-05, "loss": 0.0084, "step": 21720 }, { "epoch": 0.42, "learning_rate": 4.297530552698148e-05, "loss": 0.0, "step": 21722 }, { "epoch": 0.42, "learning_rate": 4.2974658566724245e-05, "loss": 0.0002, "step": 21724 }, { "epoch": 0.42, "learning_rate": 4.297401160646702e-05, "loss": 0.0119, "step": 21726 }, { "epoch": 0.42, "learning_rate": 4.297336464620978e-05, "loss": 0.0095, "step": 21728 }, { "epoch": 0.42, "learning_rate": 4.297271768595255e-05, "loss": 0.0009, "step": 21730 }, { "epoch": 0.42, "learning_rate": 4.297207072569532e-05, "loss": 0.0068, "step": 21732 }, { "epoch": 0.42, "learning_rate": 4.297142376543809e-05, "loss": 0.0014, "step": 21734 }, { "epoch": 0.42, "learning_rate": 4.297077680518086e-05, "loss": 0.0004, "step": 21736 }, { "epoch": 0.42, "learning_rate": 4.297012984492363e-05, "loss": 0.0001, "step": 21738 }, { "epoch": 0.42, "learning_rate": 4.29694828846664e-05, "loss": 0.0111, "step": 21740 }, { "epoch": 0.42, "learning_rate": 4.2968835924409166e-05, "loss": 0.0013, "step": 21742 }, { "epoch": 0.42, "learning_rate": 4.2968188964151936e-05, "loss": 0.0029, "step": 21744 }, { "epoch": 0.42, "learning_rate": 4.29675420038947e-05, "loss": 0.0037, "step": 21746 }, { "epoch": 0.42, "learning_rate": 4.2966895043637474e-05, "loss": 0.008, "step": 21748 }, { "epoch": 0.42, "learning_rate": 4.2966248083380236e-05, "loss": 0.011, "step": 21750 }, { "epoch": 0.42, "learning_rate": 4.2965601123123005e-05, "loss": 0.0002, "step": 21752 }, { "epoch": 0.42, "learning_rate": 4.296495416286578e-05, "loss": 0.0001, "step": 21754 }, { "epoch": 0.42, "learning_rate": 4.296430720260854e-05, "loss": 0.0045, "step": 21756 }, { "epoch": 0.42, "learning_rate": 4.296366024235132e-05, "loss": 0.0001, "step": 21758 }, { "epoch": 0.42, "learning_rate": 4.296301328209408e-05, "loss": 0.0068, "step": 21760 }, { "epoch": 0.42, "learning_rate": 4.296236632183685e-05, "loss": 0.0336, "step": 21762 }, { "epoch": 0.42, "learning_rate": 4.296171936157962e-05, "loss": 0.0002, "step": 21764 }, { "epoch": 0.42, "learning_rate": 4.296107240132239e-05, "loss": 0.0, "step": 21766 }, { "epoch": 0.42, "learning_rate": 4.296042544106516e-05, "loss": 0.009, "step": 21768 }, { "epoch": 0.42, "learning_rate": 4.295977848080793e-05, "loss": 0.0035, "step": 21770 }, { "epoch": 0.42, "learning_rate": 4.2959131520550696e-05, "loss": 0.0028, "step": 21772 }, { "epoch": 0.42, "learning_rate": 4.295848456029346e-05, "loss": 0.0001, "step": 21774 }, { "epoch": 0.42, "learning_rate": 4.2957837600036234e-05, "loss": 0.0212, "step": 21776 }, { "epoch": 0.42, "learning_rate": 4.2957190639778996e-05, "loss": 0.0094, "step": 21778 }, { "epoch": 0.42, "learning_rate": 4.295654367952177e-05, "loss": 0.0052, "step": 21780 }, { "epoch": 0.42, "learning_rate": 4.2955896719264535e-05, "loss": 0.0005, "step": 21782 }, { "epoch": 0.42, "learning_rate": 4.2955249759007304e-05, "loss": 0.0001, "step": 21784 }, { "epoch": 0.42, "learning_rate": 4.295460279875008e-05, "loss": 0.0104, "step": 21786 }, { "epoch": 0.42, "learning_rate": 4.295395583849284e-05, "loss": 0.0049, "step": 21788 }, { "epoch": 0.42, "learning_rate": 4.295330887823561e-05, "loss": 0.0001, "step": 21790 }, { "epoch": 0.42, "learning_rate": 4.295266191797838e-05, "loss": 0.0143, "step": 21792 }, { "epoch": 0.42, "learning_rate": 4.295201495772115e-05, "loss": 0.0, "step": 21794 }, { "epoch": 0.42, "learning_rate": 4.295136799746392e-05, "loss": 0.0064, "step": 21796 }, { "epoch": 0.42, "learning_rate": 4.295072103720669e-05, "loss": 0.0032, "step": 21798 }, { "epoch": 0.42, "learning_rate": 4.2950074076949456e-05, "loss": 0.0001, "step": 21800 }, { "epoch": 0.42, "learning_rate": 4.2949427116692225e-05, "loss": 0.0011, "step": 21802 }, { "epoch": 0.42, "learning_rate": 4.2948780156434995e-05, "loss": 0.0001, "step": 21804 }, { "epoch": 0.42, "learning_rate": 4.294813319617776e-05, "loss": 0.0021, "step": 21806 }, { "epoch": 0.42, "learning_rate": 4.294748623592053e-05, "loss": 0.0046, "step": 21808 }, { "epoch": 0.42, "learning_rate": 4.2946839275663295e-05, "loss": 0.0271, "step": 21810 }, { "epoch": 0.42, "learning_rate": 4.2946192315406064e-05, "loss": 0.0001, "step": 21812 }, { "epoch": 0.42, "learning_rate": 4.294554535514883e-05, "loss": 0.0016, "step": 21814 }, { "epoch": 0.42, "learning_rate": 4.29448983948916e-05, "loss": 0.0016, "step": 21816 }, { "epoch": 0.42, "learning_rate": 4.294425143463438e-05, "loss": 0.0108, "step": 21818 }, { "epoch": 0.42, "learning_rate": 4.294360447437714e-05, "loss": 0.0002, "step": 21820 }, { "epoch": 0.42, "learning_rate": 4.294295751411991e-05, "loss": 0.0, "step": 21822 }, { "epoch": 0.42, "learning_rate": 4.294231055386268e-05, "loss": 0.0039, "step": 21824 }, { "epoch": 0.42, "learning_rate": 4.294166359360545e-05, "loss": 0.0008, "step": 21826 }, { "epoch": 0.42, "learning_rate": 4.294101663334821e-05, "loss": 0.0147, "step": 21828 }, { "epoch": 0.42, "learning_rate": 4.2940369673090986e-05, "loss": 0.005, "step": 21830 }, { "epoch": 0.42, "learning_rate": 4.2939722712833755e-05, "loss": 0.0563, "step": 21832 }, { "epoch": 0.42, "learning_rate": 4.293907575257652e-05, "loss": 0.0787, "step": 21834 }, { "epoch": 0.42, "learning_rate": 4.293842879231929e-05, "loss": 0.0595, "step": 21836 }, { "epoch": 0.42, "learning_rate": 4.2937781832062055e-05, "loss": 0.0049, "step": 21838 }, { "epoch": 0.42, "learning_rate": 4.293713487180483e-05, "loss": 0.0001, "step": 21840 }, { "epoch": 0.42, "learning_rate": 4.2936487911547594e-05, "loss": 0.044, "step": 21842 }, { "epoch": 0.42, "learning_rate": 4.293584095129036e-05, "loss": 0.0001, "step": 21844 }, { "epoch": 0.42, "learning_rate": 4.293519399103313e-05, "loss": 0.0002, "step": 21846 }, { "epoch": 0.42, "learning_rate": 4.29345470307759e-05, "loss": 0.0087, "step": 21848 }, { "epoch": 0.42, "learning_rate": 4.293390007051867e-05, "loss": 0.0003, "step": 21850 }, { "epoch": 0.42, "learning_rate": 4.293325311026144e-05, "loss": 0.0119, "step": 21852 }, { "epoch": 0.42, "learning_rate": 4.293260615000421e-05, "loss": 0.0001, "step": 21854 }, { "epoch": 0.42, "learning_rate": 4.293195918974698e-05, "loss": 0.0046, "step": 21856 }, { "epoch": 0.42, "learning_rate": 4.2931312229489746e-05, "loss": 0.0002, "step": 21858 }, { "epoch": 0.42, "learning_rate": 4.293066526923251e-05, "loss": 0.0053, "step": 21860 }, { "epoch": 0.42, "learning_rate": 4.2930018308975284e-05, "loss": 0.0021, "step": 21862 }, { "epoch": 0.42, "learning_rate": 4.2929371348718054e-05, "loss": 0.0004, "step": 21864 }, { "epoch": 0.42, "learning_rate": 4.2928724388460816e-05, "loss": 0.0004, "step": 21866 }, { "epoch": 0.42, "learning_rate": 4.292807742820359e-05, "loss": 0.0001, "step": 21868 }, { "epoch": 0.42, "learning_rate": 4.2927430467946354e-05, "loss": 0.0027, "step": 21870 }, { "epoch": 0.42, "learning_rate": 4.292678350768912e-05, "loss": 0.0001, "step": 21872 }, { "epoch": 0.42, "learning_rate": 4.292613654743189e-05, "loss": 0.0, "step": 21874 }, { "epoch": 0.42, "learning_rate": 4.292548958717466e-05, "loss": 0.0001, "step": 21876 }, { "epoch": 0.42, "learning_rate": 4.292484262691743e-05, "loss": 0.0, "step": 21878 }, { "epoch": 0.42, "learning_rate": 4.29241956666602e-05, "loss": 0.0003, "step": 21880 }, { "epoch": 0.42, "learning_rate": 4.292354870640297e-05, "loss": 0.0001, "step": 21882 }, { "epoch": 0.42, "learning_rate": 4.292290174614574e-05, "loss": 0.0103, "step": 21884 }, { "epoch": 0.42, "learning_rate": 4.292225478588851e-05, "loss": 0.0002, "step": 21886 }, { "epoch": 0.42, "learning_rate": 4.292160782563127e-05, "loss": 0.004, "step": 21888 }, { "epoch": 0.42, "learning_rate": 4.2920960865374045e-05, "loss": 0.0002, "step": 21890 }, { "epoch": 0.42, "learning_rate": 4.292031390511681e-05, "loss": 0.0076, "step": 21892 }, { "epoch": 0.42, "learning_rate": 4.2919666944859576e-05, "loss": 0.0002, "step": 21894 }, { "epoch": 0.42, "learning_rate": 4.291901998460235e-05, "loss": 0.0113, "step": 21896 }, { "epoch": 0.43, "learning_rate": 4.2918373024345114e-05, "loss": 0.0006, "step": 21898 }, { "epoch": 0.43, "learning_rate": 4.291772606408789e-05, "loss": 0.0053, "step": 21900 }, { "epoch": 0.43, "learning_rate": 4.291707910383065e-05, "loss": 0.0, "step": 21902 }, { "epoch": 0.43, "learning_rate": 4.291643214357342e-05, "loss": 0.0001, "step": 21904 }, { "epoch": 0.43, "learning_rate": 4.291578518331619e-05, "loss": 0.0042, "step": 21906 }, { "epoch": 0.43, "learning_rate": 4.291513822305896e-05, "loss": 0.0051, "step": 21908 }, { "epoch": 0.43, "learning_rate": 4.291449126280173e-05, "loss": 0.0006, "step": 21910 }, { "epoch": 0.43, "learning_rate": 4.29138443025445e-05, "loss": 0.0018, "step": 21912 }, { "epoch": 0.43, "learning_rate": 4.291319734228727e-05, "loss": 0.001, "step": 21914 }, { "epoch": 0.43, "learning_rate": 4.291255038203003e-05, "loss": 0.0004, "step": 21916 }, { "epoch": 0.43, "learning_rate": 4.2911903421772805e-05, "loss": 0.0057, "step": 21918 }, { "epoch": 0.43, "learning_rate": 4.291125646151557e-05, "loss": 0.0156, "step": 21920 }, { "epoch": 0.43, "learning_rate": 4.2910609501258343e-05, "loss": 0.0007, "step": 21922 }, { "epoch": 0.43, "learning_rate": 4.2909962541001106e-05, "loss": 0.0, "step": 21924 }, { "epoch": 0.43, "learning_rate": 4.2909315580743875e-05, "loss": 0.0003, "step": 21926 }, { "epoch": 0.43, "learning_rate": 4.2908668620486644e-05, "loss": 0.0001, "step": 21928 }, { "epoch": 0.43, "learning_rate": 4.290802166022941e-05, "loss": 0.0009, "step": 21930 }, { "epoch": 0.43, "learning_rate": 4.290737469997218e-05, "loss": 0.0152, "step": 21932 }, { "epoch": 0.43, "learning_rate": 4.290672773971495e-05, "loss": 0.0002, "step": 21934 }, { "epoch": 0.43, "learning_rate": 4.290608077945772e-05, "loss": 0.0025, "step": 21936 }, { "epoch": 0.43, "learning_rate": 4.290543381920049e-05, "loss": 0.1246, "step": 21938 }, { "epoch": 0.43, "learning_rate": 4.290478685894326e-05, "loss": 0.0056, "step": 21940 }, { "epoch": 0.43, "learning_rate": 4.290413989868603e-05, "loss": 0.0079, "step": 21942 }, { "epoch": 0.43, "learning_rate": 4.2903492938428797e-05, "loss": 0.0004, "step": 21944 }, { "epoch": 0.43, "learning_rate": 4.2902845978171566e-05, "loss": 0.0027, "step": 21946 }, { "epoch": 0.43, "learning_rate": 4.290219901791433e-05, "loss": 0.0005, "step": 21948 }, { "epoch": 0.43, "learning_rate": 4.2901552057657104e-05, "loss": 0.0049, "step": 21950 }, { "epoch": 0.43, "learning_rate": 4.2900905097399866e-05, "loss": 0.0006, "step": 21952 }, { "epoch": 0.43, "learning_rate": 4.2900258137142635e-05, "loss": 0.001, "step": 21954 }, { "epoch": 0.43, "learning_rate": 4.2899611176885404e-05, "loss": 0.0007, "step": 21956 }, { "epoch": 0.43, "learning_rate": 4.2898964216628173e-05, "loss": 0.0015, "step": 21958 }, { "epoch": 0.43, "learning_rate": 4.289831725637094e-05, "loss": 0.002, "step": 21960 }, { "epoch": 0.43, "learning_rate": 4.289767029611371e-05, "loss": 0.0001, "step": 21962 }, { "epoch": 0.43, "learning_rate": 4.289702333585648e-05, "loss": 0.0003, "step": 21964 }, { "epoch": 0.43, "learning_rate": 4.289637637559925e-05, "loss": 0.0078, "step": 21966 }, { "epoch": 0.43, "learning_rate": 4.289572941534202e-05, "loss": 0.0032, "step": 21968 }, { "epoch": 0.43, "learning_rate": 4.289508245508478e-05, "loss": 0.0026, "step": 21970 }, { "epoch": 0.43, "learning_rate": 4.289443549482756e-05, "loss": 0.0009, "step": 21972 }, { "epoch": 0.43, "learning_rate": 4.2893788534570326e-05, "loss": 0.0029, "step": 21974 }, { "epoch": 0.43, "learning_rate": 4.289314157431309e-05, "loss": 0.0124, "step": 21976 }, { "epoch": 0.43, "learning_rate": 4.2892494614055864e-05, "loss": 0.0004, "step": 21978 }, { "epoch": 0.43, "learning_rate": 4.2891847653798627e-05, "loss": 0.0017, "step": 21980 }, { "epoch": 0.43, "learning_rate": 4.28912006935414e-05, "loss": 0.0002, "step": 21982 }, { "epoch": 0.43, "learning_rate": 4.2890553733284165e-05, "loss": 0.0, "step": 21984 }, { "epoch": 0.43, "learning_rate": 4.2889906773026934e-05, "loss": 0.0063, "step": 21986 }, { "epoch": 0.43, "learning_rate": 4.28892598127697e-05, "loss": 0.0008, "step": 21988 }, { "epoch": 0.43, "learning_rate": 4.288861285251247e-05, "loss": 0.0047, "step": 21990 }, { "epoch": 0.43, "learning_rate": 4.288796589225524e-05, "loss": 0.0003, "step": 21992 }, { "epoch": 0.43, "learning_rate": 4.288731893199801e-05, "loss": 0.006, "step": 21994 }, { "epoch": 0.43, "learning_rate": 4.288667197174078e-05, "loss": 0.0001, "step": 21996 }, { "epoch": 0.43, "learning_rate": 4.288602501148355e-05, "loss": 0.0428, "step": 21998 }, { "epoch": 0.43, "learning_rate": 4.288537805122632e-05, "loss": 0.0001, "step": 22000 }, { "epoch": 0.43, "learning_rate": 4.288473109096908e-05, "loss": 0.0134, "step": 22002 }, { "epoch": 0.43, "learning_rate": 4.2884084130711856e-05, "loss": 0.0096, "step": 22004 }, { "epoch": 0.43, "learning_rate": 4.288343717045462e-05, "loss": 0.0182, "step": 22006 }, { "epoch": 0.43, "learning_rate": 4.288279021019739e-05, "loss": 0.0001, "step": 22008 }, { "epoch": 0.43, "learning_rate": 4.288214324994016e-05, "loss": 0.0003, "step": 22010 }, { "epoch": 0.43, "learning_rate": 4.2881496289682925e-05, "loss": 0.0001, "step": 22012 }, { "epoch": 0.43, "learning_rate": 4.2880849329425694e-05, "loss": 0.0155, "step": 22014 }, { "epoch": 0.43, "learning_rate": 4.288020236916846e-05, "loss": 0.0017, "step": 22016 }, { "epoch": 0.43, "learning_rate": 4.287955540891123e-05, "loss": 0.0007, "step": 22018 }, { "epoch": 0.43, "learning_rate": 4.2878908448654e-05, "loss": 0.0029, "step": 22020 }, { "epoch": 0.43, "learning_rate": 4.287826148839677e-05, "loss": 0.0044, "step": 22022 }, { "epoch": 0.43, "learning_rate": 4.287761452813954e-05, "loss": 0.0024, "step": 22024 }, { "epoch": 0.43, "learning_rate": 4.287696756788231e-05, "loss": 0.0063, "step": 22026 }, { "epoch": 0.43, "learning_rate": 4.287632060762508e-05, "loss": 0.0, "step": 22028 }, { "epoch": 0.43, "learning_rate": 4.287567364736784e-05, "loss": 0.0094, "step": 22030 }, { "epoch": 0.43, "learning_rate": 4.2875026687110616e-05, "loss": 0.0015, "step": 22032 }, { "epoch": 0.43, "learning_rate": 4.287437972685338e-05, "loss": 0.0049, "step": 22034 }, { "epoch": 0.43, "learning_rate": 4.287373276659615e-05, "loss": 0.0001, "step": 22036 }, { "epoch": 0.43, "learning_rate": 4.2873085806338916e-05, "loss": 0.0006, "step": 22038 }, { "epoch": 0.43, "learning_rate": 4.2872438846081686e-05, "loss": 0.0108, "step": 22040 }, { "epoch": 0.43, "learning_rate": 4.287179188582446e-05, "loss": 0.0067, "step": 22042 }, { "epoch": 0.43, "learning_rate": 4.2871144925567224e-05, "loss": 0.0, "step": 22044 }, { "epoch": 0.43, "learning_rate": 4.287049796530999e-05, "loss": 0.0022, "step": 22046 }, { "epoch": 0.43, "learning_rate": 4.286985100505276e-05, "loss": 0.008, "step": 22048 }, { "epoch": 0.43, "learning_rate": 4.286920404479553e-05, "loss": 0.0001, "step": 22050 }, { "epoch": 0.43, "learning_rate": 4.28685570845383e-05, "loss": 0.001, "step": 22052 }, { "epoch": 0.43, "learning_rate": 4.286791012428107e-05, "loss": 0.0256, "step": 22054 }, { "epoch": 0.43, "learning_rate": 4.286726316402384e-05, "loss": 0.0001, "step": 22056 }, { "epoch": 0.43, "learning_rate": 4.28666162037666e-05, "loss": 0.0098, "step": 22058 }, { "epoch": 0.43, "learning_rate": 4.2865969243509376e-05, "loss": 0.0014, "step": 22060 }, { "epoch": 0.43, "learning_rate": 4.286532228325214e-05, "loss": 0.0002, "step": 22062 }, { "epoch": 0.43, "learning_rate": 4.2864675322994915e-05, "loss": 0.0004, "step": 22064 }, { "epoch": 0.43, "learning_rate": 4.286402836273768e-05, "loss": 0.0003, "step": 22066 }, { "epoch": 0.43, "learning_rate": 4.2863381402480446e-05, "loss": 0.0005, "step": 22068 }, { "epoch": 0.43, "learning_rate": 4.2862734442223215e-05, "loss": 0.0005, "step": 22070 }, { "epoch": 0.43, "learning_rate": 4.2862087481965984e-05, "loss": 0.0001, "step": 22072 }, { "epoch": 0.43, "learning_rate": 4.286144052170875e-05, "loss": 0.0005, "step": 22074 }, { "epoch": 0.43, "learning_rate": 4.286079356145152e-05, "loss": 0.0107, "step": 22076 }, { "epoch": 0.43, "learning_rate": 4.286014660119429e-05, "loss": 0.0073, "step": 22078 }, { "epoch": 0.43, "learning_rate": 4.285949964093706e-05, "loss": 0.0081, "step": 22080 }, { "epoch": 0.43, "learning_rate": 4.285885268067983e-05, "loss": 0.0011, "step": 22082 }, { "epoch": 0.43, "learning_rate": 4.285820572042259e-05, "loss": 0.0046, "step": 22084 }, { "epoch": 0.43, "learning_rate": 4.285755876016537e-05, "loss": 0.0004, "step": 22086 }, { "epoch": 0.43, "learning_rate": 4.285691179990814e-05, "loss": 0.0001, "step": 22088 }, { "epoch": 0.43, "learning_rate": 4.28562648396509e-05, "loss": 0.0002, "step": 22090 }, { "epoch": 0.43, "learning_rate": 4.2855617879393675e-05, "loss": 0.0001, "step": 22092 }, { "epoch": 0.43, "learning_rate": 4.285497091913644e-05, "loss": 0.0007, "step": 22094 }, { "epoch": 0.43, "learning_rate": 4.2854323958879206e-05, "loss": 0.039, "step": 22096 }, { "epoch": 0.43, "learning_rate": 4.2853676998621975e-05, "loss": 0.0002, "step": 22098 }, { "epoch": 0.43, "learning_rate": 4.2853030038364745e-05, "loss": 0.0004, "step": 22100 }, { "epoch": 0.43, "learning_rate": 4.2852383078107514e-05, "loss": 0.0093, "step": 22102 }, { "epoch": 0.43, "learning_rate": 4.285173611785028e-05, "loss": 0.0006, "step": 22104 }, { "epoch": 0.43, "learning_rate": 4.285108915759305e-05, "loss": 0.045, "step": 22106 }, { "epoch": 0.43, "learning_rate": 4.285044219733582e-05, "loss": 0.0146, "step": 22108 }, { "epoch": 0.43, "learning_rate": 4.284979523707859e-05, "loss": 0.0003, "step": 22110 }, { "epoch": 0.43, "learning_rate": 4.284914827682135e-05, "loss": 0.0001, "step": 22112 }, { "epoch": 0.43, "learning_rate": 4.284850131656413e-05, "loss": 0.0038, "step": 22114 }, { "epoch": 0.43, "learning_rate": 4.284785435630689e-05, "loss": 0.0033, "step": 22116 }, { "epoch": 0.43, "learning_rate": 4.284720739604966e-05, "loss": 0.0178, "step": 22118 }, { "epoch": 0.43, "learning_rate": 4.2846560435792435e-05, "loss": 0.0013, "step": 22120 }, { "epoch": 0.43, "learning_rate": 4.28459134755352e-05, "loss": 0.0001, "step": 22122 }, { "epoch": 0.43, "learning_rate": 4.2845266515277974e-05, "loss": 0.0091, "step": 22124 }, { "epoch": 0.43, "learning_rate": 4.2844619555020736e-05, "loss": 0.0012, "step": 22126 }, { "epoch": 0.43, "learning_rate": 4.2843972594763505e-05, "loss": 0.0002, "step": 22128 }, { "epoch": 0.43, "learning_rate": 4.2843325634506274e-05, "loss": 0.0072, "step": 22130 }, { "epoch": 0.43, "learning_rate": 4.284267867424904e-05, "loss": 0.0018, "step": 22132 }, { "epoch": 0.43, "learning_rate": 4.284203171399181e-05, "loss": 0.0, "step": 22134 }, { "epoch": 0.43, "learning_rate": 4.284138475373458e-05, "loss": 0.007, "step": 22136 }, { "epoch": 0.43, "learning_rate": 4.284073779347735e-05, "loss": 0.013, "step": 22138 }, { "epoch": 0.43, "learning_rate": 4.284009083322012e-05, "loss": 0.0064, "step": 22140 }, { "epoch": 0.43, "learning_rate": 4.283944387296289e-05, "loss": 0.0042, "step": 22142 }, { "epoch": 0.43, "learning_rate": 4.283879691270565e-05, "loss": 0.0003, "step": 22144 }, { "epoch": 0.43, "learning_rate": 4.283814995244843e-05, "loss": 0.0037, "step": 22146 }, { "epoch": 0.43, "learning_rate": 4.283750299219119e-05, "loss": 0.0002, "step": 22148 }, { "epoch": 0.43, "learning_rate": 4.283685603193396e-05, "loss": 0.0081, "step": 22150 }, { "epoch": 0.43, "learning_rate": 4.2836209071676734e-05, "loss": 0.0014, "step": 22152 }, { "epoch": 0.43, "learning_rate": 4.2835562111419496e-05, "loss": 0.0045, "step": 22154 }, { "epoch": 0.43, "learning_rate": 4.2834915151162265e-05, "loss": 0.0013, "step": 22156 }, { "epoch": 0.43, "learning_rate": 4.2834268190905034e-05, "loss": 0.0032, "step": 22158 }, { "epoch": 0.43, "learning_rate": 4.2833621230647803e-05, "loss": 0.0041, "step": 22160 }, { "epoch": 0.43, "learning_rate": 4.283297427039057e-05, "loss": 0.0024, "step": 22162 }, { "epoch": 0.43, "learning_rate": 4.283232731013334e-05, "loss": 0.0001, "step": 22164 }, { "epoch": 0.43, "learning_rate": 4.283168034987611e-05, "loss": 0.0096, "step": 22166 }, { "epoch": 0.43, "learning_rate": 4.283103338961888e-05, "loss": 0.0021, "step": 22168 }, { "epoch": 0.43, "learning_rate": 4.283038642936165e-05, "loss": 0.0005, "step": 22170 }, { "epoch": 0.43, "learning_rate": 4.282973946910441e-05, "loss": 0.0001, "step": 22172 }, { "epoch": 0.43, "learning_rate": 4.282909250884719e-05, "loss": 0.0065, "step": 22174 }, { "epoch": 0.43, "learning_rate": 4.282844554858995e-05, "loss": 0.0005, "step": 22176 }, { "epoch": 0.43, "learning_rate": 4.282779858833272e-05, "loss": 0.0005, "step": 22178 }, { "epoch": 0.43, "learning_rate": 4.282715162807549e-05, "loss": 0.0011, "step": 22180 }, { "epoch": 0.43, "learning_rate": 4.282650466781826e-05, "loss": 0.002, "step": 22182 }, { "epoch": 0.43, "learning_rate": 4.2825857707561026e-05, "loss": 0.0123, "step": 22184 }, { "epoch": 0.43, "learning_rate": 4.2825210747303795e-05, "loss": 0.0001, "step": 22186 }, { "epoch": 0.43, "learning_rate": 4.2824563787046564e-05, "loss": 0.0074, "step": 22188 }, { "epoch": 0.43, "learning_rate": 4.282391682678933e-05, "loss": 0.0001, "step": 22190 }, { "epoch": 0.43, "learning_rate": 4.28232698665321e-05, "loss": 0.0017, "step": 22192 }, { "epoch": 0.43, "learning_rate": 4.2822622906274864e-05, "loss": 0.0001, "step": 22194 }, { "epoch": 0.43, "learning_rate": 4.282197594601764e-05, "loss": 0.0005, "step": 22196 }, { "epoch": 0.43, "learning_rate": 4.282132898576041e-05, "loss": 0.0096, "step": 22198 }, { "epoch": 0.43, "learning_rate": 4.282068202550317e-05, "loss": 0.0021, "step": 22200 }, { "epoch": 0.43, "learning_rate": 4.282003506524595e-05, "loss": 0.0012, "step": 22202 }, { "epoch": 0.43, "learning_rate": 4.281938810498871e-05, "loss": 0.0005, "step": 22204 }, { "epoch": 0.43, "learning_rate": 4.2818741144731486e-05, "loss": 0.0096, "step": 22206 }, { "epoch": 0.43, "learning_rate": 4.281809418447425e-05, "loss": 0.0008, "step": 22208 }, { "epoch": 0.43, "learning_rate": 4.281744722421702e-05, "loss": 0.0008, "step": 22210 }, { "epoch": 0.43, "learning_rate": 4.2816800263959786e-05, "loss": 0.0086, "step": 22212 }, { "epoch": 0.43, "learning_rate": 4.2816153303702555e-05, "loss": 0.0184, "step": 22214 }, { "epoch": 0.43, "learning_rate": 4.2815506343445324e-05, "loss": 0.0127, "step": 22216 }, { "epoch": 0.43, "learning_rate": 4.281485938318809e-05, "loss": 0.0, "step": 22218 }, { "epoch": 0.43, "learning_rate": 4.281421242293086e-05, "loss": 0.0002, "step": 22220 }, { "epoch": 0.43, "learning_rate": 4.281356546267363e-05, "loss": 0.0001, "step": 22222 }, { "epoch": 0.43, "learning_rate": 4.28129185024164e-05, "loss": 0.0008, "step": 22224 }, { "epoch": 0.43, "learning_rate": 4.281227154215916e-05, "loss": 0.001, "step": 22226 }, { "epoch": 0.43, "learning_rate": 4.281162458190194e-05, "loss": 0.0008, "step": 22228 }, { "epoch": 0.43, "learning_rate": 4.28109776216447e-05, "loss": 0.0138, "step": 22230 }, { "epoch": 0.43, "learning_rate": 4.281033066138747e-05, "loss": 0.0001, "step": 22232 }, { "epoch": 0.43, "learning_rate": 4.2809683701130246e-05, "loss": 0.0007, "step": 22234 }, { "epoch": 0.43, "learning_rate": 4.280903674087301e-05, "loss": 0.015, "step": 22236 }, { "epoch": 0.43, "learning_rate": 4.280838978061578e-05, "loss": 0.0099, "step": 22238 }, { "epoch": 0.43, "learning_rate": 4.2807742820358547e-05, "loss": 0.0011, "step": 22240 }, { "epoch": 0.43, "learning_rate": 4.2807095860101316e-05, "loss": 0.007, "step": 22242 }, { "epoch": 0.43, "learning_rate": 4.2806448899844085e-05, "loss": 0.0008, "step": 22244 }, { "epoch": 0.43, "learning_rate": 4.2805801939586854e-05, "loss": 0.0002, "step": 22246 }, { "epoch": 0.43, "learning_rate": 4.280515497932962e-05, "loss": 0.0147, "step": 22248 }, { "epoch": 0.43, "learning_rate": 4.280450801907239e-05, "loss": 0.0059, "step": 22250 }, { "epoch": 0.43, "learning_rate": 4.280386105881516e-05, "loss": 0.0047, "step": 22252 }, { "epoch": 0.43, "learning_rate": 4.280321409855792e-05, "loss": 0.0002, "step": 22254 }, { "epoch": 0.43, "learning_rate": 4.28025671383007e-05, "loss": 0.0092, "step": 22256 }, { "epoch": 0.43, "learning_rate": 4.280192017804346e-05, "loss": 0.0006, "step": 22258 }, { "epoch": 0.43, "learning_rate": 4.280127321778623e-05, "loss": 0.0001, "step": 22260 }, { "epoch": 0.43, "learning_rate": 4.2800626257529e-05, "loss": 0.0086, "step": 22262 }, { "epoch": 0.43, "learning_rate": 4.279997929727177e-05, "loss": 0.0003, "step": 22264 }, { "epoch": 0.43, "learning_rate": 4.2799332337014545e-05, "loss": 0.0011, "step": 22266 }, { "epoch": 0.43, "learning_rate": 4.279868537675731e-05, "loss": 0.0127, "step": 22268 }, { "epoch": 0.43, "learning_rate": 4.2798038416500076e-05, "loss": 0.0008, "step": 22270 }, { "epoch": 0.43, "learning_rate": 4.2797391456242845e-05, "loss": 0.0097, "step": 22272 }, { "epoch": 0.43, "learning_rate": 4.2796744495985614e-05, "loss": 0.0083, "step": 22274 }, { "epoch": 0.43, "learning_rate": 4.279609753572838e-05, "loss": 0.0002, "step": 22276 }, { "epoch": 0.43, "learning_rate": 4.279545057547115e-05, "loss": 0.0001, "step": 22278 }, { "epoch": 0.43, "learning_rate": 4.279480361521392e-05, "loss": 0.0099, "step": 22280 }, { "epoch": 0.43, "learning_rate": 4.279415665495669e-05, "loss": 0.0341, "step": 22282 }, { "epoch": 0.43, "learning_rate": 4.279350969469946e-05, "loss": 0.0096, "step": 22284 }, { "epoch": 0.43, "learning_rate": 4.279286273444222e-05, "loss": 0.0042, "step": 22286 }, { "epoch": 0.43, "learning_rate": 4.2792215774185e-05, "loss": 0.0154, "step": 22288 }, { "epoch": 0.43, "learning_rate": 4.279156881392776e-05, "loss": 0.0057, "step": 22290 }, { "epoch": 0.43, "learning_rate": 4.279092185367053e-05, "loss": 0.0002, "step": 22292 }, { "epoch": 0.43, "learning_rate": 4.27902748934133e-05, "loss": 0.0002, "step": 22294 }, { "epoch": 0.43, "learning_rate": 4.278962793315607e-05, "loss": 0.0006, "step": 22296 }, { "epoch": 0.43, "learning_rate": 4.2788980972898836e-05, "loss": 0.0002, "step": 22298 }, { "epoch": 0.43, "learning_rate": 4.2788334012641606e-05, "loss": 0.0, "step": 22300 }, { "epoch": 0.43, "learning_rate": 4.2787687052384375e-05, "loss": 0.0046, "step": 22302 }, { "epoch": 0.43, "learning_rate": 4.2787040092127144e-05, "loss": 0.0013, "step": 22304 }, { "epoch": 0.43, "learning_rate": 4.278639313186991e-05, "loss": 0.0101, "step": 22306 }, { "epoch": 0.43, "learning_rate": 4.2785746171612675e-05, "loss": 0.0003, "step": 22308 }, { "epoch": 0.43, "learning_rate": 4.278509921135545e-05, "loss": 0.0006, "step": 22310 }, { "epoch": 0.43, "learning_rate": 4.278445225109822e-05, "loss": 0.0005, "step": 22312 }, { "epoch": 0.43, "learning_rate": 4.278380529084098e-05, "loss": 0.0006, "step": 22314 }, { "epoch": 0.43, "learning_rate": 4.278315833058376e-05, "loss": 0.0, "step": 22316 }, { "epoch": 0.43, "learning_rate": 4.278251137032652e-05, "loss": 0.0075, "step": 22318 }, { "epoch": 0.43, "learning_rate": 4.278186441006929e-05, "loss": 0.0001, "step": 22320 }, { "epoch": 0.43, "learning_rate": 4.278121744981206e-05, "loss": 0.0137, "step": 22322 }, { "epoch": 0.43, "learning_rate": 4.278057048955483e-05, "loss": 0.0052, "step": 22324 }, { "epoch": 0.43, "learning_rate": 4.27799235292976e-05, "loss": 0.0003, "step": 22326 }, { "epoch": 0.43, "learning_rate": 4.2779276569040366e-05, "loss": 0.002, "step": 22328 }, { "epoch": 0.43, "learning_rate": 4.2778629608783135e-05, "loss": 0.024, "step": 22330 }, { "epoch": 0.43, "learning_rate": 4.2777982648525904e-05, "loss": 0.0003, "step": 22332 }, { "epoch": 0.43, "learning_rate": 4.277733568826867e-05, "loss": 0.0002, "step": 22334 }, { "epoch": 0.43, "learning_rate": 4.2776688728011435e-05, "loss": 0.0003, "step": 22336 }, { "epoch": 0.43, "learning_rate": 4.277604176775421e-05, "loss": 0.0109, "step": 22338 }, { "epoch": 0.43, "learning_rate": 4.2775394807496974e-05, "loss": 0.0026, "step": 22340 }, { "epoch": 0.43, "learning_rate": 4.277474784723974e-05, "loss": 0.0125, "step": 22342 }, { "epoch": 0.43, "learning_rate": 4.277410088698252e-05, "loss": 0.0062, "step": 22344 }, { "epoch": 0.43, "learning_rate": 4.277345392672528e-05, "loss": 0.0005, "step": 22346 }, { "epoch": 0.43, "learning_rate": 4.277280696646806e-05, "loss": 0.0077, "step": 22348 }, { "epoch": 0.43, "learning_rate": 4.277216000621082e-05, "loss": 0.0022, "step": 22350 }, { "epoch": 0.43, "learning_rate": 4.277151304595359e-05, "loss": 0.0048, "step": 22352 }, { "epoch": 0.43, "learning_rate": 4.277086608569636e-05, "loss": 0.0001, "step": 22354 }, { "epoch": 0.43, "learning_rate": 4.2770219125439126e-05, "loss": 0.0001, "step": 22356 }, { "epoch": 0.43, "learning_rate": 4.2769572165181895e-05, "loss": 0.0001, "step": 22358 }, { "epoch": 0.43, "learning_rate": 4.2768925204924664e-05, "loss": 0.0203, "step": 22360 }, { "epoch": 0.43, "learning_rate": 4.2768278244667434e-05, "loss": 0.0026, "step": 22362 }, { "epoch": 0.43, "learning_rate": 4.27676312844102e-05, "loss": 0.0204, "step": 22364 }, { "epoch": 0.43, "learning_rate": 4.276698432415297e-05, "loss": 0.0021, "step": 22366 }, { "epoch": 0.43, "learning_rate": 4.2766337363895734e-05, "loss": 0.0013, "step": 22368 }, { "epoch": 0.43, "learning_rate": 4.276569040363851e-05, "loss": 0.0044, "step": 22370 }, { "epoch": 0.43, "learning_rate": 4.276504344338127e-05, "loss": 0.0001, "step": 22372 }, { "epoch": 0.43, "learning_rate": 4.276439648312404e-05, "loss": 0.0092, "step": 22374 }, { "epoch": 0.43, "learning_rate": 4.276374952286682e-05, "loss": 0.0067, "step": 22376 }, { "epoch": 0.43, "learning_rate": 4.276310256260958e-05, "loss": 0.0003, "step": 22378 }, { "epoch": 0.43, "learning_rate": 4.276245560235235e-05, "loss": 0.0065, "step": 22380 }, { "epoch": 0.43, "learning_rate": 4.276180864209512e-05, "loss": 0.0012, "step": 22382 }, { "epoch": 0.43, "learning_rate": 4.276116168183789e-05, "loss": 0.0284, "step": 22384 }, { "epoch": 0.43, "learning_rate": 4.2760514721580656e-05, "loss": 0.001, "step": 22386 }, { "epoch": 0.43, "learning_rate": 4.2759867761323425e-05, "loss": 0.0004, "step": 22388 }, { "epoch": 0.43, "learning_rate": 4.2759220801066194e-05, "loss": 0.0005, "step": 22390 }, { "epoch": 0.43, "learning_rate": 4.275857384080896e-05, "loss": 0.0026, "step": 22392 }, { "epoch": 0.43, "learning_rate": 4.275792688055173e-05, "loss": 0.0001, "step": 22394 }, { "epoch": 0.43, "learning_rate": 4.2757279920294494e-05, "loss": 0.0001, "step": 22396 }, { "epoch": 0.43, "learning_rate": 4.275663296003727e-05, "loss": 0.0001, "step": 22398 }, { "epoch": 0.43, "learning_rate": 4.275598599978003e-05, "loss": 0.0018, "step": 22400 }, { "epoch": 0.43, "learning_rate": 4.27553390395228e-05, "loss": 0.0004, "step": 22402 }, { "epoch": 0.43, "learning_rate": 4.275469207926557e-05, "loss": 0.0179, "step": 22404 }, { "epoch": 0.43, "learning_rate": 4.275404511900834e-05, "loss": 0.0003, "step": 22406 }, { "epoch": 0.43, "learning_rate": 4.275339815875111e-05, "loss": 0.0324, "step": 22408 }, { "epoch": 0.43, "learning_rate": 4.275275119849388e-05, "loss": 0.1238, "step": 22410 }, { "epoch": 0.43, "learning_rate": 4.275210423823665e-05, "loss": 0.0002, "step": 22412 }, { "epoch": 0.44, "learning_rate": 4.2751457277979416e-05, "loss": 0.0004, "step": 22414 }, { "epoch": 0.44, "learning_rate": 4.2750810317722185e-05, "loss": 0.0005, "step": 22416 }, { "epoch": 0.44, "learning_rate": 4.275016335746495e-05, "loss": 0.0047, "step": 22418 }, { "epoch": 0.44, "learning_rate": 4.2749516397207723e-05, "loss": 0.0007, "step": 22420 }, { "epoch": 0.44, "learning_rate": 4.274886943695049e-05, "loss": 0.0003, "step": 22422 }, { "epoch": 0.44, "learning_rate": 4.274822247669326e-05, "loss": 0.001, "step": 22424 }, { "epoch": 0.44, "learning_rate": 4.274757551643603e-05, "loss": 0.0081, "step": 22426 }, { "epoch": 0.44, "learning_rate": 4.274692855617879e-05, "loss": 0.0062, "step": 22428 }, { "epoch": 0.44, "learning_rate": 4.274628159592157e-05, "loss": 0.004, "step": 22430 }, { "epoch": 0.44, "learning_rate": 4.274563463566433e-05, "loss": 0.1289, "step": 22432 }, { "epoch": 0.44, "learning_rate": 4.27449876754071e-05, "loss": 0.0377, "step": 22434 }, { "epoch": 0.44, "learning_rate": 4.274434071514987e-05, "loss": 0.0008, "step": 22436 }, { "epoch": 0.44, "learning_rate": 4.274369375489264e-05, "loss": 0.0271, "step": 22438 }, { "epoch": 0.44, "learning_rate": 4.274304679463541e-05, "loss": 0.0018, "step": 22440 }, { "epoch": 0.44, "learning_rate": 4.274239983437818e-05, "loss": 0.0002, "step": 22442 }, { "epoch": 0.44, "learning_rate": 4.2741752874120946e-05, "loss": 0.0017, "step": 22444 }, { "epoch": 0.44, "learning_rate": 4.2741105913863715e-05, "loss": 0.0001, "step": 22446 }, { "epoch": 0.44, "learning_rate": 4.2740458953606484e-05, "loss": 0.0008, "step": 22448 }, { "epoch": 0.44, "learning_rate": 4.2739811993349246e-05, "loss": 0.0005, "step": 22450 }, { "epoch": 0.44, "learning_rate": 4.273916503309202e-05, "loss": 0.0001, "step": 22452 }, { "epoch": 0.44, "learning_rate": 4.273851807283479e-05, "loss": 0.0001, "step": 22454 }, { "epoch": 0.44, "learning_rate": 4.2737871112577553e-05, "loss": 0.0, "step": 22456 }, { "epoch": 0.44, "learning_rate": 4.273722415232033e-05, "loss": 0.0002, "step": 22458 }, { "epoch": 0.44, "learning_rate": 4.273657719206309e-05, "loss": 0.0045, "step": 22460 }, { "epoch": 0.44, "learning_rate": 4.273593023180586e-05, "loss": 0.0002, "step": 22462 }, { "epoch": 0.44, "learning_rate": 4.273528327154863e-05, "loss": 0.0002, "step": 22464 }, { "epoch": 0.44, "learning_rate": 4.27346363112914e-05, "loss": 0.0008, "step": 22466 }, { "epoch": 0.44, "learning_rate": 4.273398935103417e-05, "loss": 0.0029, "step": 22468 }, { "epoch": 0.44, "learning_rate": 4.273334239077694e-05, "loss": 0.0029, "step": 22470 }, { "epoch": 0.44, "learning_rate": 4.2732695430519706e-05, "loss": 0.0004, "step": 22472 }, { "epoch": 0.44, "learning_rate": 4.2732048470262475e-05, "loss": 0.0025, "step": 22474 }, { "epoch": 0.44, "learning_rate": 4.2731401510005244e-05, "loss": 0.0002, "step": 22476 }, { "epoch": 0.44, "learning_rate": 4.2730754549748007e-05, "loss": 0.0041, "step": 22478 }, { "epoch": 0.44, "learning_rate": 4.273010758949078e-05, "loss": 0.0196, "step": 22480 }, { "epoch": 0.44, "learning_rate": 4.2729460629233545e-05, "loss": 0.0001, "step": 22482 }, { "epoch": 0.44, "learning_rate": 4.2728813668976314e-05, "loss": 0.009, "step": 22484 }, { "epoch": 0.44, "learning_rate": 4.272816670871908e-05, "loss": 0.0001, "step": 22486 }, { "epoch": 0.44, "learning_rate": 4.272751974846185e-05, "loss": 0.0028, "step": 22488 }, { "epoch": 0.44, "learning_rate": 4.272687278820463e-05, "loss": 0.0106, "step": 22490 }, { "epoch": 0.44, "learning_rate": 4.272622582794739e-05, "loss": 0.0003, "step": 22492 }, { "epoch": 0.44, "learning_rate": 4.272557886769016e-05, "loss": 0.0063, "step": 22494 }, { "epoch": 0.44, "learning_rate": 4.272493190743293e-05, "loss": 0.0214, "step": 22496 }, { "epoch": 0.44, "learning_rate": 4.27242849471757e-05, "loss": 0.0052, "step": 22498 }, { "epoch": 0.44, "learning_rate": 4.2723637986918467e-05, "loss": 0.0013, "step": 22500 }, { "epoch": 0.44, "learning_rate": 4.2722991026661236e-05, "loss": 0.0006, "step": 22502 }, { "epoch": 0.44, "learning_rate": 4.2722344066404005e-05, "loss": 0.0152, "step": 22504 }, { "epoch": 0.44, "learning_rate": 4.2721697106146774e-05, "loss": 0.004, "step": 22506 }, { "epoch": 0.44, "learning_rate": 4.272105014588954e-05, "loss": 0.0051, "step": 22508 }, { "epoch": 0.44, "learning_rate": 4.2720403185632305e-05, "loss": 0.0, "step": 22510 }, { "epoch": 0.44, "learning_rate": 4.271975622537508e-05, "loss": 0.0102, "step": 22512 }, { "epoch": 0.44, "learning_rate": 4.271910926511784e-05, "loss": 0.0011, "step": 22514 }, { "epoch": 0.44, "learning_rate": 4.271846230486061e-05, "loss": 0.0005, "step": 22516 }, { "epoch": 0.44, "learning_rate": 4.271781534460338e-05, "loss": 0.0002, "step": 22518 }, { "epoch": 0.44, "learning_rate": 4.271716838434615e-05, "loss": 0.0001, "step": 22520 }, { "epoch": 0.44, "learning_rate": 4.271652142408892e-05, "loss": 0.0063, "step": 22522 }, { "epoch": 0.44, "learning_rate": 4.271587446383169e-05, "loss": 0.0072, "step": 22524 }, { "epoch": 0.44, "learning_rate": 4.271522750357446e-05, "loss": 0.0027, "step": 22526 }, { "epoch": 0.44, "learning_rate": 4.271458054331723e-05, "loss": 0.0005, "step": 22528 }, { "epoch": 0.44, "learning_rate": 4.2713933583059996e-05, "loss": 0.0015, "step": 22530 }, { "epoch": 0.44, "learning_rate": 4.2713286622802765e-05, "loss": 0.0015, "step": 22532 }, { "epoch": 0.44, "learning_rate": 4.2712639662545534e-05, "loss": 0.0017, "step": 22534 }, { "epoch": 0.44, "learning_rate": 4.27119927022883e-05, "loss": 0.0001, "step": 22536 }, { "epoch": 0.44, "learning_rate": 4.2711345742031066e-05, "loss": 0.0115, "step": 22538 }, { "epoch": 0.44, "learning_rate": 4.271069878177384e-05, "loss": 0.0024, "step": 22540 }, { "epoch": 0.44, "learning_rate": 4.2710051821516604e-05, "loss": 0.0003, "step": 22542 }, { "epoch": 0.44, "learning_rate": 4.270940486125937e-05, "loss": 0.0021, "step": 22544 }, { "epoch": 0.44, "learning_rate": 4.270875790100214e-05, "loss": 0.0001, "step": 22546 }, { "epoch": 0.44, "learning_rate": 4.270811094074491e-05, "loss": 0.0111, "step": 22548 }, { "epoch": 0.44, "learning_rate": 4.270746398048768e-05, "loss": 0.0007, "step": 22550 }, { "epoch": 0.44, "learning_rate": 4.270681702023045e-05, "loss": 0.0021, "step": 22552 }, { "epoch": 0.44, "learning_rate": 4.270617005997322e-05, "loss": 0.0139, "step": 22554 }, { "epoch": 0.44, "learning_rate": 4.270552309971599e-05, "loss": 0.0006, "step": 22556 }, { "epoch": 0.44, "learning_rate": 4.2704876139458756e-05, "loss": 0.0001, "step": 22558 }, { "epoch": 0.44, "learning_rate": 4.270422917920152e-05, "loss": 0.0019, "step": 22560 }, { "epoch": 0.44, "learning_rate": 4.2703582218944295e-05, "loss": 0.0014, "step": 22562 }, { "epoch": 0.44, "learning_rate": 4.270293525868706e-05, "loss": 0.0108, "step": 22564 }, { "epoch": 0.44, "learning_rate": 4.270228829842983e-05, "loss": 0.019, "step": 22566 }, { "epoch": 0.44, "learning_rate": 4.27016413381726e-05, "loss": 0.0012, "step": 22568 }, { "epoch": 0.44, "learning_rate": 4.2700994377915364e-05, "loss": 0.0007, "step": 22570 }, { "epoch": 0.44, "learning_rate": 4.270034741765814e-05, "loss": 0.0017, "step": 22572 }, { "epoch": 0.44, "learning_rate": 4.26997004574009e-05, "loss": 0.0071, "step": 22574 }, { "epoch": 0.44, "learning_rate": 4.269905349714367e-05, "loss": 0.0002, "step": 22576 }, { "epoch": 0.44, "learning_rate": 4.269840653688644e-05, "loss": 0.0, "step": 22578 }, { "epoch": 0.44, "learning_rate": 4.269775957662921e-05, "loss": 0.0194, "step": 22580 }, { "epoch": 0.44, "learning_rate": 4.269711261637198e-05, "loss": 0.001, "step": 22582 }, { "epoch": 0.44, "learning_rate": 4.269646565611475e-05, "loss": 0.0006, "step": 22584 }, { "epoch": 0.44, "learning_rate": 4.269581869585752e-05, "loss": 0.001, "step": 22586 }, { "epoch": 0.44, "learning_rate": 4.2695171735600286e-05, "loss": 0.0003, "step": 22588 }, { "epoch": 0.44, "learning_rate": 4.2694524775343055e-05, "loss": 0.0001, "step": 22590 }, { "epoch": 0.44, "learning_rate": 4.269387781508582e-05, "loss": 0.0002, "step": 22592 }, { "epoch": 0.44, "learning_rate": 4.269323085482859e-05, "loss": 0.074, "step": 22594 }, { "epoch": 0.44, "learning_rate": 4.2692583894571355e-05, "loss": 0.0003, "step": 22596 }, { "epoch": 0.44, "learning_rate": 4.2691936934314125e-05, "loss": 0.0, "step": 22598 }, { "epoch": 0.44, "learning_rate": 4.26912899740569e-05, "loss": 0.0066, "step": 22600 }, { "epoch": 0.44, "learning_rate": 4.269064301379966e-05, "loss": 0.01, "step": 22602 }, { "epoch": 0.44, "learning_rate": 4.268999605354243e-05, "loss": 0.0004, "step": 22604 }, { "epoch": 0.44, "learning_rate": 4.26893490932852e-05, "loss": 0.0, "step": 22606 }, { "epoch": 0.44, "learning_rate": 4.268870213302797e-05, "loss": 0.0005, "step": 22608 }, { "epoch": 0.44, "learning_rate": 4.268805517277074e-05, "loss": 0.0001, "step": 22610 }, { "epoch": 0.44, "learning_rate": 4.268740821251351e-05, "loss": 0.0073, "step": 22612 }, { "epoch": 0.44, "learning_rate": 4.268676125225628e-05, "loss": 0.0014, "step": 22614 }, { "epoch": 0.44, "learning_rate": 4.2686114291999046e-05, "loss": 0.0052, "step": 22616 }, { "epoch": 0.44, "learning_rate": 4.2685467331741815e-05, "loss": 0.0, "step": 22618 }, { "epoch": 0.44, "learning_rate": 4.268482037148458e-05, "loss": 0.0, "step": 22620 }, { "epoch": 0.44, "learning_rate": 4.2684173411227354e-05, "loss": 0.0002, "step": 22622 }, { "epoch": 0.44, "learning_rate": 4.2683526450970116e-05, "loss": 0.0002, "step": 22624 }, { "epoch": 0.44, "learning_rate": 4.2682879490712885e-05, "loss": 0.0, "step": 22626 }, { "epoch": 0.44, "learning_rate": 4.2682232530455654e-05, "loss": 0.0077, "step": 22628 }, { "epoch": 0.44, "learning_rate": 4.268158557019842e-05, "loss": 0.0046, "step": 22630 }, { "epoch": 0.44, "learning_rate": 4.26809386099412e-05, "loss": 0.0075, "step": 22632 }, { "epoch": 0.44, "learning_rate": 4.268029164968396e-05, "loss": 0.0003, "step": 22634 }, { "epoch": 0.44, "learning_rate": 4.267964468942673e-05, "loss": 0.0019, "step": 22636 }, { "epoch": 0.44, "learning_rate": 4.26789977291695e-05, "loss": 0.0005, "step": 22638 }, { "epoch": 0.44, "learning_rate": 4.267835076891227e-05, "loss": 0.0034, "step": 22640 }, { "epoch": 0.44, "learning_rate": 4.267770380865503e-05, "loss": 0.0001, "step": 22642 }, { "epoch": 0.44, "learning_rate": 4.267705684839781e-05, "loss": 0.0229, "step": 22644 }, { "epoch": 0.44, "learning_rate": 4.2676409888140576e-05, "loss": 0.0007, "step": 22646 }, { "epoch": 0.44, "learning_rate": 4.2675762927883345e-05, "loss": 0.0001, "step": 22648 }, { "epoch": 0.44, "learning_rate": 4.2675115967626114e-05, "loss": 0.0006, "step": 22650 }, { "epoch": 0.44, "learning_rate": 4.2674469007368876e-05, "loss": 0.0039, "step": 22652 }, { "epoch": 0.44, "learning_rate": 4.267382204711165e-05, "loss": 0.0126, "step": 22654 }, { "epoch": 0.44, "learning_rate": 4.2673175086854414e-05, "loss": 0.0002, "step": 22656 }, { "epoch": 0.44, "learning_rate": 4.2672528126597184e-05, "loss": 0.0001, "step": 22658 }, { "epoch": 0.44, "learning_rate": 4.267188116633995e-05, "loss": 0.0055, "step": 22660 }, { "epoch": 0.44, "learning_rate": 4.267123420608272e-05, "loss": 0.1088, "step": 22662 }, { "epoch": 0.44, "learning_rate": 4.267058724582549e-05, "loss": 0.0105, "step": 22664 }, { "epoch": 0.44, "learning_rate": 4.266994028556826e-05, "loss": 0.0021, "step": 22666 }, { "epoch": 0.44, "learning_rate": 4.266929332531103e-05, "loss": 0.0051, "step": 22668 }, { "epoch": 0.44, "learning_rate": 4.26686463650538e-05, "loss": 0.0006, "step": 22670 }, { "epoch": 0.44, "learning_rate": 4.266799940479657e-05, "loss": 0.0003, "step": 22672 }, { "epoch": 0.44, "learning_rate": 4.266735244453933e-05, "loss": 0.0141, "step": 22674 }, { "epoch": 0.44, "learning_rate": 4.2666705484282105e-05, "loss": 0.0008, "step": 22676 }, { "epoch": 0.44, "learning_rate": 4.2666058524024874e-05, "loss": 0.0086, "step": 22678 }, { "epoch": 0.44, "learning_rate": 4.266541156376764e-05, "loss": 0.0045, "step": 22680 }, { "epoch": 0.44, "learning_rate": 4.266476460351041e-05, "loss": 0.0002, "step": 22682 }, { "epoch": 0.44, "learning_rate": 4.2664117643253175e-05, "loss": 0.0001, "step": 22684 }, { "epoch": 0.44, "learning_rate": 4.2663470682995944e-05, "loss": 0.0002, "step": 22686 }, { "epoch": 0.44, "learning_rate": 4.266282372273871e-05, "loss": 0.0003, "step": 22688 }, { "epoch": 0.44, "learning_rate": 4.266217676248148e-05, "loss": 0.0002, "step": 22690 }, { "epoch": 0.44, "learning_rate": 4.266152980222425e-05, "loss": 0.0034, "step": 22692 }, { "epoch": 0.44, "learning_rate": 4.266088284196702e-05, "loss": 0.0006, "step": 22694 }, { "epoch": 0.44, "learning_rate": 4.266023588170979e-05, "loss": 0.0002, "step": 22696 }, { "epoch": 0.44, "learning_rate": 4.265958892145256e-05, "loss": 0.0024, "step": 22698 }, { "epoch": 0.44, "learning_rate": 4.265894196119533e-05, "loss": 0.0003, "step": 22700 }, { "epoch": 0.44, "learning_rate": 4.265829500093809e-05, "loss": 0.0038, "step": 22702 }, { "epoch": 0.44, "learning_rate": 4.2657648040680866e-05, "loss": 0.0002, "step": 22704 }, { "epoch": 0.44, "learning_rate": 4.265700108042363e-05, "loss": 0.0047, "step": 22706 }, { "epoch": 0.44, "learning_rate": 4.2656354120166404e-05, "loss": 0.0013, "step": 22708 }, { "epoch": 0.44, "learning_rate": 4.265570715990917e-05, "loss": 0.0091, "step": 22710 }, { "epoch": 0.44, "learning_rate": 4.2655060199651935e-05, "loss": 0.0002, "step": 22712 }, { "epoch": 0.44, "learning_rate": 4.265441323939471e-05, "loss": 0.0084, "step": 22714 }, { "epoch": 0.44, "learning_rate": 4.2653766279137473e-05, "loss": 0.0381, "step": 22716 }, { "epoch": 0.44, "learning_rate": 4.265311931888024e-05, "loss": 0.0008, "step": 22718 }, { "epoch": 0.44, "learning_rate": 4.265247235862301e-05, "loss": 0.0014, "step": 22720 }, { "epoch": 0.44, "learning_rate": 4.265182539836578e-05, "loss": 0.0001, "step": 22722 }, { "epoch": 0.44, "learning_rate": 4.265117843810855e-05, "loss": 0.003, "step": 22724 }, { "epoch": 0.44, "learning_rate": 4.265053147785132e-05, "loss": 0.0001, "step": 22726 }, { "epoch": 0.44, "learning_rate": 4.264988451759409e-05, "loss": 0.0001, "step": 22728 }, { "epoch": 0.44, "learning_rate": 4.264923755733686e-05, "loss": 0.0045, "step": 22730 }, { "epoch": 0.44, "learning_rate": 4.2648590597079626e-05, "loss": 0.0006, "step": 22732 }, { "epoch": 0.44, "learning_rate": 4.264794363682239e-05, "loss": 0.0001, "step": 22734 }, { "epoch": 0.44, "learning_rate": 4.2647296676565164e-05, "loss": 0.0005, "step": 22736 }, { "epoch": 0.44, "learning_rate": 4.2646649716307927e-05, "loss": 0.0001, "step": 22738 }, { "epoch": 0.44, "learning_rate": 4.2646002756050696e-05, "loss": 0.0002, "step": 22740 }, { "epoch": 0.44, "learning_rate": 4.2645355795793465e-05, "loss": 0.0037, "step": 22742 }, { "epoch": 0.44, "learning_rate": 4.2644708835536234e-05, "loss": 0.0057, "step": 22744 }, { "epoch": 0.44, "learning_rate": 4.2644061875279e-05, "loss": 0.0088, "step": 22746 }, { "epoch": 0.44, "learning_rate": 4.264341491502177e-05, "loss": 0.0031, "step": 22748 }, { "epoch": 0.44, "learning_rate": 4.264276795476454e-05, "loss": 0.0032, "step": 22750 }, { "epoch": 0.44, "learning_rate": 4.264212099450731e-05, "loss": 0.0085, "step": 22752 }, { "epoch": 0.44, "learning_rate": 4.264147403425008e-05, "loss": 0.0012, "step": 22754 }, { "epoch": 0.44, "learning_rate": 4.264082707399285e-05, "loss": 0.0018, "step": 22756 }, { "epoch": 0.44, "learning_rate": 4.264018011373562e-05, "loss": 0.0001, "step": 22758 }, { "epoch": 0.44, "learning_rate": 4.2639533153478386e-05, "loss": 0.0001, "step": 22760 }, { "epoch": 0.44, "learning_rate": 4.263888619322115e-05, "loss": 0.0004, "step": 22762 }, { "epoch": 0.44, "learning_rate": 4.2638239232963925e-05, "loss": 0.0001, "step": 22764 }, { "epoch": 0.44, "learning_rate": 4.263759227270669e-05, "loss": 0.002, "step": 22766 }, { "epoch": 0.44, "learning_rate": 4.2636945312449456e-05, "loss": 0.0, "step": 22768 }, { "epoch": 0.44, "learning_rate": 4.2636298352192225e-05, "loss": 0.0003, "step": 22770 }, { "epoch": 0.44, "learning_rate": 4.2635651391934994e-05, "loss": 0.0002, "step": 22772 }, { "epoch": 0.44, "learning_rate": 4.263500443167776e-05, "loss": 0.0001, "step": 22774 }, { "epoch": 0.44, "learning_rate": 4.263435747142053e-05, "loss": 0.0047, "step": 22776 }, { "epoch": 0.44, "learning_rate": 4.26337105111633e-05, "loss": 0.0007, "step": 22778 }, { "epoch": 0.44, "learning_rate": 4.263306355090607e-05, "loss": 0.0013, "step": 22780 }, { "epoch": 0.44, "learning_rate": 4.263241659064884e-05, "loss": 0.0155, "step": 22782 }, { "epoch": 0.44, "learning_rate": 4.26317696303916e-05, "loss": 0.0019, "step": 22784 }, { "epoch": 0.44, "learning_rate": 4.263112267013438e-05, "loss": 0.0042, "step": 22786 }, { "epoch": 0.44, "learning_rate": 4.263047570987714e-05, "loss": 0.0018, "step": 22788 }, { "epoch": 0.44, "learning_rate": 4.2629828749619916e-05, "loss": 0.0062, "step": 22790 }, { "epoch": 0.44, "learning_rate": 4.2629181789362685e-05, "loss": 0.0002, "step": 22792 }, { "epoch": 0.44, "learning_rate": 4.262853482910545e-05, "loss": 0.0111, "step": 22794 }, { "epoch": 0.44, "learning_rate": 4.262788786884822e-05, "loss": 0.0042, "step": 22796 }, { "epoch": 0.44, "learning_rate": 4.2627240908590986e-05, "loss": 0.0008, "step": 22798 }, { "epoch": 0.44, "learning_rate": 4.2626593948333755e-05, "loss": 0.0003, "step": 22800 }, { "epoch": 0.44, "learning_rate": 4.2625946988076524e-05, "loss": 0.0002, "step": 22802 }, { "epoch": 0.44, "learning_rate": 4.262530002781929e-05, "loss": 0.007, "step": 22804 }, { "epoch": 0.44, "learning_rate": 4.262465306756206e-05, "loss": 0.0001, "step": 22806 }, { "epoch": 0.44, "learning_rate": 4.262400610730483e-05, "loss": 0.0007, "step": 22808 }, { "epoch": 0.44, "learning_rate": 4.26233591470476e-05, "loss": 0.0002, "step": 22810 }, { "epoch": 0.44, "learning_rate": 4.262271218679037e-05, "loss": 0.0001, "step": 22812 }, { "epoch": 0.44, "learning_rate": 4.262206522653314e-05, "loss": 0.0041, "step": 22814 }, { "epoch": 0.44, "learning_rate": 4.26214182662759e-05, "loss": 0.0001, "step": 22816 }, { "epoch": 0.44, "learning_rate": 4.2620771306018676e-05, "loss": 0.0025, "step": 22818 }, { "epoch": 0.44, "learning_rate": 4.262012434576144e-05, "loss": 0.0002, "step": 22820 }, { "epoch": 0.44, "learning_rate": 4.261947738550421e-05, "loss": 0.0, "step": 22822 }, { "epoch": 0.44, "learning_rate": 4.2618830425246984e-05, "loss": 0.0017, "step": 22824 }, { "epoch": 0.44, "learning_rate": 4.2618183464989746e-05, "loss": 0.0544, "step": 22826 }, { "epoch": 0.44, "learning_rate": 4.2617536504732515e-05, "loss": 0.0001, "step": 22828 }, { "epoch": 0.44, "learning_rate": 4.2616889544475284e-05, "loss": 0.0021, "step": 22830 }, { "epoch": 0.44, "learning_rate": 4.261624258421805e-05, "loss": 0.0066, "step": 22832 }, { "epoch": 0.44, "learning_rate": 4.261559562396082e-05, "loss": 0.0001, "step": 22834 }, { "epoch": 0.44, "learning_rate": 4.261494866370359e-05, "loss": 0.0003, "step": 22836 }, { "epoch": 0.44, "learning_rate": 4.261430170344636e-05, "loss": 0.0001, "step": 22838 }, { "epoch": 0.44, "learning_rate": 4.261365474318913e-05, "loss": 0.0004, "step": 22840 }, { "epoch": 0.44, "learning_rate": 4.26130077829319e-05, "loss": 0.013, "step": 22842 }, { "epoch": 0.44, "learning_rate": 4.261236082267466e-05, "loss": 0.0001, "step": 22844 }, { "epoch": 0.44, "learning_rate": 4.261171386241744e-05, "loss": 0.0002, "step": 22846 }, { "epoch": 0.44, "learning_rate": 4.26110669021602e-05, "loss": 0.0004, "step": 22848 }, { "epoch": 0.44, "learning_rate": 4.2610419941902975e-05, "loss": 0.0038, "step": 22850 }, { "epoch": 0.44, "learning_rate": 4.260977298164574e-05, "loss": 0.0037, "step": 22852 }, { "epoch": 0.44, "learning_rate": 4.2609126021388506e-05, "loss": 0.0316, "step": 22854 }, { "epoch": 0.44, "learning_rate": 4.260847906113128e-05, "loss": 0.0006, "step": 22856 }, { "epoch": 0.44, "learning_rate": 4.2607832100874045e-05, "loss": 0.0022, "step": 22858 }, { "epoch": 0.44, "learning_rate": 4.2607185140616814e-05, "loss": 0.0003, "step": 22860 }, { "epoch": 0.44, "learning_rate": 4.260653818035958e-05, "loss": 0.0006, "step": 22862 }, { "epoch": 0.44, "learning_rate": 4.260589122010235e-05, "loss": 0.006, "step": 22864 }, { "epoch": 0.44, "learning_rate": 4.2605244259845114e-05, "loss": 0.0001, "step": 22866 }, { "epoch": 0.44, "learning_rate": 4.260459729958789e-05, "loss": 0.0003, "step": 22868 }, { "epoch": 0.44, "learning_rate": 4.260395033933066e-05, "loss": 0.0001, "step": 22870 }, { "epoch": 0.44, "learning_rate": 4.260330337907343e-05, "loss": 0.0001, "step": 22872 }, { "epoch": 0.44, "learning_rate": 4.26026564188162e-05, "loss": 0.0, "step": 22874 }, { "epoch": 0.44, "learning_rate": 4.260200945855896e-05, "loss": 0.0001, "step": 22876 }, { "epoch": 0.44, "learning_rate": 4.2601362498301735e-05, "loss": 0.0048, "step": 22878 }, { "epoch": 0.44, "learning_rate": 4.26007155380445e-05, "loss": 0.0246, "step": 22880 }, { "epoch": 0.44, "learning_rate": 4.260006857778727e-05, "loss": 0.0, "step": 22882 }, { "epoch": 0.44, "learning_rate": 4.2599421617530036e-05, "loss": 0.0067, "step": 22884 }, { "epoch": 0.44, "learning_rate": 4.2598774657272805e-05, "loss": 0.0045, "step": 22886 }, { "epoch": 0.44, "learning_rate": 4.2598127697015574e-05, "loss": 0.0001, "step": 22888 }, { "epoch": 0.44, "learning_rate": 4.259748073675834e-05, "loss": 0.0059, "step": 22890 }, { "epoch": 0.44, "learning_rate": 4.259683377650111e-05, "loss": 0.0001, "step": 22892 }, { "epoch": 0.44, "learning_rate": 4.259618681624388e-05, "loss": 0.0001, "step": 22894 }, { "epoch": 0.44, "learning_rate": 4.259553985598665e-05, "loss": 0.0019, "step": 22896 }, { "epoch": 0.44, "learning_rate": 4.259489289572941e-05, "loss": 0.0079, "step": 22898 }, { "epoch": 0.44, "learning_rate": 4.259424593547219e-05, "loss": 0.0002, "step": 22900 }, { "epoch": 0.44, "learning_rate": 4.259359897521496e-05, "loss": 0.0, "step": 22902 }, { "epoch": 0.44, "learning_rate": 4.259295201495772e-05, "loss": 0.0, "step": 22904 }, { "epoch": 0.44, "learning_rate": 4.2592305054700496e-05, "loss": 0.0002, "step": 22906 }, { "epoch": 0.44, "learning_rate": 4.259165809444326e-05, "loss": 0.0, "step": 22908 }, { "epoch": 0.44, "learning_rate": 4.259101113418603e-05, "loss": 0.011, "step": 22910 }, { "epoch": 0.44, "learning_rate": 4.2590364173928796e-05, "loss": 0.0026, "step": 22912 }, { "epoch": 0.44, "learning_rate": 4.2589717213671565e-05, "loss": 0.0001, "step": 22914 }, { "epoch": 0.44, "learning_rate": 4.2589070253414334e-05, "loss": 0.0005, "step": 22916 }, { "epoch": 0.44, "learning_rate": 4.2588423293157103e-05, "loss": 0.0176, "step": 22918 }, { "epoch": 0.44, "learning_rate": 4.258777633289987e-05, "loss": 0.0033, "step": 22920 }, { "epoch": 0.44, "learning_rate": 4.258712937264264e-05, "loss": 0.0034, "step": 22922 }, { "epoch": 0.44, "learning_rate": 4.258648241238541e-05, "loss": 0.0031, "step": 22924 }, { "epoch": 0.44, "learning_rate": 4.258583545212817e-05, "loss": 0.0103, "step": 22926 }, { "epoch": 0.45, "learning_rate": 4.258518849187095e-05, "loss": 0.0006, "step": 22928 }, { "epoch": 0.45, "learning_rate": 4.258454153161371e-05, "loss": 0.0012, "step": 22930 }, { "epoch": 0.45, "learning_rate": 4.258389457135649e-05, "loss": 0.0002, "step": 22932 }, { "epoch": 0.45, "learning_rate": 4.2583247611099256e-05, "loss": 0.003, "step": 22934 }, { "epoch": 0.45, "learning_rate": 4.258260065084202e-05, "loss": 0.0005, "step": 22936 }, { "epoch": 0.45, "learning_rate": 4.2581953690584794e-05, "loss": 0.0001, "step": 22938 }, { "epoch": 0.45, "learning_rate": 4.258130673032756e-05, "loss": 0.0032, "step": 22940 }, { "epoch": 0.45, "learning_rate": 4.2580659770070326e-05, "loss": 0.0001, "step": 22942 }, { "epoch": 0.45, "learning_rate": 4.2580012809813095e-05, "loss": 0.0219, "step": 22944 }, { "epoch": 0.45, "learning_rate": 4.2579365849555864e-05, "loss": 0.0003, "step": 22946 }, { "epoch": 0.45, "learning_rate": 4.257871888929863e-05, "loss": 0.022, "step": 22948 }, { "epoch": 0.45, "learning_rate": 4.25780719290414e-05, "loss": 0.0082, "step": 22950 }, { "epoch": 0.45, "learning_rate": 4.257742496878417e-05, "loss": 0.0019, "step": 22952 }, { "epoch": 0.45, "learning_rate": 4.257677800852694e-05, "loss": 0.0078, "step": 22954 }, { "epoch": 0.45, "learning_rate": 4.257613104826971e-05, "loss": 0.01, "step": 22956 }, { "epoch": 0.45, "learning_rate": 4.257548408801247e-05, "loss": 0.0002, "step": 22958 }, { "epoch": 0.45, "learning_rate": 4.257483712775525e-05, "loss": 0.0001, "step": 22960 }, { "epoch": 0.45, "learning_rate": 4.257419016749801e-05, "loss": 0.0056, "step": 22962 }, { "epoch": 0.45, "learning_rate": 4.257354320724078e-05, "loss": 0.0125, "step": 22964 }, { "epoch": 0.45, "learning_rate": 4.257289624698355e-05, "loss": 0.0013, "step": 22966 }, { "epoch": 0.45, "learning_rate": 4.257224928672632e-05, "loss": 0.0006, "step": 22968 }, { "epoch": 0.45, "learning_rate": 4.2571602326469086e-05, "loss": 0.0, "step": 22970 }, { "epoch": 0.45, "learning_rate": 4.2570955366211855e-05, "loss": 0.0001, "step": 22972 }, { "epoch": 0.45, "learning_rate": 4.2570308405954624e-05, "loss": 0.0036, "step": 22974 }, { "epoch": 0.45, "learning_rate": 4.2569661445697393e-05, "loss": 0.0043, "step": 22976 }, { "epoch": 0.45, "learning_rate": 4.256901448544016e-05, "loss": 0.0221, "step": 22978 }, { "epoch": 0.45, "learning_rate": 4.256836752518293e-05, "loss": 0.0018, "step": 22980 }, { "epoch": 0.45, "learning_rate": 4.25677205649257e-05, "loss": 0.0119, "step": 22982 }, { "epoch": 0.45, "learning_rate": 4.256707360466847e-05, "loss": 0.0001, "step": 22984 }, { "epoch": 0.45, "learning_rate": 4.256642664441123e-05, "loss": 0.0026, "step": 22986 }, { "epoch": 0.45, "learning_rate": 4.256577968415401e-05, "loss": 0.0001, "step": 22988 }, { "epoch": 0.45, "learning_rate": 4.256513272389677e-05, "loss": 0.0156, "step": 22990 }, { "epoch": 0.45, "learning_rate": 4.2564485763639546e-05, "loss": 0.0002, "step": 22992 }, { "epoch": 0.45, "learning_rate": 4.256383880338231e-05, "loss": 0.0001, "step": 22994 }, { "epoch": 0.45, "learning_rate": 4.256319184312508e-05, "loss": 0.0003, "step": 22996 }, { "epoch": 0.45, "learning_rate": 4.2562544882867847e-05, "loss": 0.0129, "step": 22998 }, { "epoch": 0.45, "learning_rate": 4.2561897922610616e-05, "loss": 0.0049, "step": 23000 }, { "epoch": 0.45, "learning_rate": 4.2561250962353385e-05, "loss": 0.0, "step": 23002 }, { "epoch": 0.45, "learning_rate": 4.2560604002096154e-05, "loss": 0.0001, "step": 23004 }, { "epoch": 0.45, "learning_rate": 4.255995704183892e-05, "loss": 0.0009, "step": 23006 }, { "epoch": 0.45, "learning_rate": 4.2559310081581685e-05, "loss": 0.0001, "step": 23008 }, { "epoch": 0.45, "learning_rate": 4.255866312132446e-05, "loss": 0.0014, "step": 23010 }, { "epoch": 0.45, "learning_rate": 4.255801616106723e-05, "loss": 0.0002, "step": 23012 }, { "epoch": 0.45, "learning_rate": 4.255736920081e-05, "loss": 0.0002, "step": 23014 }, { "epoch": 0.45, "learning_rate": 4.255672224055277e-05, "loss": 0.008, "step": 23016 }, { "epoch": 0.45, "learning_rate": 4.255607528029553e-05, "loss": 0.0001, "step": 23018 }, { "epoch": 0.45, "learning_rate": 4.2555428320038306e-05, "loss": 0.0026, "step": 23020 }, { "epoch": 0.45, "learning_rate": 4.255478135978107e-05, "loss": 0.0, "step": 23022 }, { "epoch": 0.45, "learning_rate": 4.255413439952384e-05, "loss": 0.0003, "step": 23024 }, { "epoch": 0.45, "learning_rate": 4.255348743926661e-05, "loss": 0.0064, "step": 23026 }, { "epoch": 0.45, "learning_rate": 4.2552840479009376e-05, "loss": 0.0001, "step": 23028 }, { "epoch": 0.45, "learning_rate": 4.2552193518752145e-05, "loss": 0.0376, "step": 23030 }, { "epoch": 0.45, "learning_rate": 4.2551546558494914e-05, "loss": 0.0022, "step": 23032 }, { "epoch": 0.45, "learning_rate": 4.255089959823768e-05, "loss": 0.0001, "step": 23034 }, { "epoch": 0.45, "learning_rate": 4.255025263798045e-05, "loss": 0.0002, "step": 23036 }, { "epoch": 0.45, "learning_rate": 4.254960567772322e-05, "loss": 0.0001, "step": 23038 }, { "epoch": 0.45, "learning_rate": 4.2548958717465984e-05, "loss": 0.0003, "step": 23040 }, { "epoch": 0.45, "learning_rate": 4.254831175720876e-05, "loss": 0.0002, "step": 23042 }, { "epoch": 0.45, "learning_rate": 4.254766479695152e-05, "loss": 0.0032, "step": 23044 }, { "epoch": 0.45, "learning_rate": 4.254701783669429e-05, "loss": 0.0014, "step": 23046 }, { "epoch": 0.45, "learning_rate": 4.254637087643707e-05, "loss": 0.0012, "step": 23048 }, { "epoch": 0.45, "learning_rate": 4.254572391617983e-05, "loss": 0.0, "step": 23050 }, { "epoch": 0.45, "learning_rate": 4.25450769559226e-05, "loss": 0.001, "step": 23052 }, { "epoch": 0.45, "learning_rate": 4.254442999566537e-05, "loss": 0.0001, "step": 23054 }, { "epoch": 0.45, "learning_rate": 4.2543783035408136e-05, "loss": 0.0044, "step": 23056 }, { "epoch": 0.45, "learning_rate": 4.2543136075150906e-05, "loss": 0.0001, "step": 23058 }, { "epoch": 0.45, "learning_rate": 4.2542489114893675e-05, "loss": 0.0103, "step": 23060 }, { "epoch": 0.45, "learning_rate": 4.2541842154636444e-05, "loss": 0.0004, "step": 23062 }, { "epoch": 0.45, "learning_rate": 4.254119519437921e-05, "loss": 0.0002, "step": 23064 }, { "epoch": 0.45, "learning_rate": 4.254054823412198e-05, "loss": 0.0002, "step": 23066 }, { "epoch": 0.45, "learning_rate": 4.2539901273864744e-05, "loss": 0.0, "step": 23068 }, { "epoch": 0.45, "learning_rate": 4.253925431360752e-05, "loss": 0.005, "step": 23070 }, { "epoch": 0.45, "learning_rate": 4.253860735335028e-05, "loss": 0.0002, "step": 23072 }, { "epoch": 0.45, "learning_rate": 4.253796039309306e-05, "loss": 0.0003, "step": 23074 }, { "epoch": 0.45, "learning_rate": 4.253731343283582e-05, "loss": 0.0001, "step": 23076 }, { "epoch": 0.45, "learning_rate": 4.253666647257859e-05, "loss": 0.0022, "step": 23078 }, { "epoch": 0.45, "learning_rate": 4.2536019512321365e-05, "loss": 0.0004, "step": 23080 }, { "epoch": 0.45, "learning_rate": 4.253537255206413e-05, "loss": 0.0001, "step": 23082 }, { "epoch": 0.45, "learning_rate": 4.25347255918069e-05, "loss": 0.0002, "step": 23084 }, { "epoch": 0.45, "learning_rate": 4.2534078631549666e-05, "loss": 0.0002, "step": 23086 }, { "epoch": 0.45, "learning_rate": 4.2533431671292435e-05, "loss": 0.0001, "step": 23088 }, { "epoch": 0.45, "learning_rate": 4.253310819116382e-05, "loss": 0.0278, "step": 23090 }, { "epoch": 0.45, "learning_rate": 4.253246123090659e-05, "loss": 0.0033, "step": 23092 }, { "epoch": 0.45, "learning_rate": 4.253181427064936e-05, "loss": 0.001, "step": 23094 }, { "epoch": 0.45, "learning_rate": 4.253116731039212e-05, "loss": 0.0003, "step": 23096 }, { "epoch": 0.45, "learning_rate": 4.2530520350134896e-05, "loss": 0.0067, "step": 23098 }, { "epoch": 0.45, "learning_rate": 4.252987338987766e-05, "loss": 0.0002, "step": 23100 }, { "epoch": 0.45, "learning_rate": 4.252922642962043e-05, "loss": 0.0002, "step": 23102 }, { "epoch": 0.45, "learning_rate": 4.25285794693632e-05, "loss": 0.0, "step": 23104 }, { "epoch": 0.45, "learning_rate": 4.2527932509105965e-05, "loss": 0.0002, "step": 23106 }, { "epoch": 0.45, "learning_rate": 4.252728554884874e-05, "loss": 0.0, "step": 23108 }, { "epoch": 0.45, "learning_rate": 4.2526638588591504e-05, "loss": 0.0012, "step": 23110 }, { "epoch": 0.45, "learning_rate": 4.252599162833427e-05, "loss": 0.0, "step": 23112 }, { "epoch": 0.45, "learning_rate": 4.252534466807704e-05, "loss": 0.0, "step": 23114 }, { "epoch": 0.45, "learning_rate": 4.252469770781981e-05, "loss": 0.0125, "step": 23116 }, { "epoch": 0.45, "learning_rate": 4.252405074756258e-05, "loss": 0.0058, "step": 23118 }, { "epoch": 0.45, "learning_rate": 4.252340378730535e-05, "loss": 0.0017, "step": 23120 }, { "epoch": 0.45, "learning_rate": 4.252275682704812e-05, "loss": 0.0021, "step": 23122 }, { "epoch": 0.45, "learning_rate": 4.252210986679088e-05, "loss": 0.0084, "step": 23124 }, { "epoch": 0.45, "learning_rate": 4.2521462906533656e-05, "loss": 0.0001, "step": 23126 }, { "epoch": 0.45, "learning_rate": 4.252081594627642e-05, "loss": 0.0002, "step": 23128 }, { "epoch": 0.45, "learning_rate": 4.2520168986019194e-05, "loss": 0.0006, "step": 23130 }, { "epoch": 0.45, "learning_rate": 4.251952202576196e-05, "loss": 0.0022, "step": 23132 }, { "epoch": 0.45, "learning_rate": 4.2518875065504726e-05, "loss": 0.0, "step": 23134 }, { "epoch": 0.45, "learning_rate": 4.25182281052475e-05, "loss": 0.0074, "step": 23136 }, { "epoch": 0.45, "learning_rate": 4.2517581144990264e-05, "loss": 0.0005, "step": 23138 }, { "epoch": 0.45, "learning_rate": 4.251693418473303e-05, "loss": 0.0003, "step": 23140 }, { "epoch": 0.45, "learning_rate": 4.25162872244758e-05, "loss": 0.0, "step": 23142 }, { "epoch": 0.45, "learning_rate": 4.251564026421857e-05, "loss": 0.0001, "step": 23144 }, { "epoch": 0.45, "learning_rate": 4.251499330396134e-05, "loss": 0.0001, "step": 23146 }, { "epoch": 0.45, "learning_rate": 4.251434634370411e-05, "loss": 0.0, "step": 23148 }, { "epoch": 0.45, "learning_rate": 4.251369938344688e-05, "loss": 0.0023, "step": 23150 }, { "epoch": 0.45, "learning_rate": 4.251305242318965e-05, "loss": 0.0001, "step": 23152 }, { "epoch": 0.45, "learning_rate": 4.251240546293242e-05, "loss": 0.0004, "step": 23154 }, { "epoch": 0.45, "learning_rate": 4.251175850267518e-05, "loss": 0.0013, "step": 23156 }, { "epoch": 0.45, "learning_rate": 4.2511111542417955e-05, "loss": 0.0002, "step": 23158 }, { "epoch": 0.45, "learning_rate": 4.251046458216072e-05, "loss": 0.0121, "step": 23160 }, { "epoch": 0.45, "learning_rate": 4.2509817621903486e-05, "loss": 0.0002, "step": 23162 }, { "epoch": 0.45, "learning_rate": 4.2509170661646255e-05, "loss": 0.0001, "step": 23164 }, { "epoch": 0.45, "learning_rate": 4.2508523701389024e-05, "loss": 0.005, "step": 23166 }, { "epoch": 0.45, "learning_rate": 4.2507876741131794e-05, "loss": 0.0022, "step": 23168 }, { "epoch": 0.45, "learning_rate": 4.250722978087456e-05, "loss": 0.0006, "step": 23170 }, { "epoch": 0.45, "learning_rate": 4.250658282061733e-05, "loss": 0.0007, "step": 23172 }, { "epoch": 0.45, "learning_rate": 4.25059358603601e-05, "loss": 0.0011, "step": 23174 }, { "epoch": 0.45, "learning_rate": 4.250528890010287e-05, "loss": 0.0009, "step": 23176 }, { "epoch": 0.45, "learning_rate": 4.250464193984563e-05, "loss": 0.0003, "step": 23178 }, { "epoch": 0.45, "learning_rate": 4.250399497958841e-05, "loss": 0.0006, "step": 23180 }, { "epoch": 0.45, "learning_rate": 4.250334801933118e-05, "loss": 0.0056, "step": 23182 }, { "epoch": 0.45, "learning_rate": 4.250270105907394e-05, "loss": 0.0, "step": 23184 }, { "epoch": 0.45, "learning_rate": 4.2502054098816715e-05, "loss": 0.0002, "step": 23186 }, { "epoch": 0.45, "learning_rate": 4.250140713855948e-05, "loss": 0.0008, "step": 23188 }, { "epoch": 0.45, "learning_rate": 4.2500760178302253e-05, "loss": 0.0142, "step": 23190 }, { "epoch": 0.45, "learning_rate": 4.2500113218045016e-05, "loss": 0.0, "step": 23192 }, { "epoch": 0.45, "learning_rate": 4.2499466257787785e-05, "loss": 0.0005, "step": 23194 }, { "epoch": 0.45, "learning_rate": 4.2498819297530554e-05, "loss": 0.0005, "step": 23196 }, { "epoch": 0.45, "learning_rate": 4.249817233727332e-05, "loss": 0.0004, "step": 23198 }, { "epoch": 0.45, "learning_rate": 4.249752537701609e-05, "loss": 0.0033, "step": 23200 }, { "epoch": 0.45, "learning_rate": 4.249687841675886e-05, "loss": 0.0001, "step": 23202 }, { "epoch": 0.45, "learning_rate": 4.249623145650163e-05, "loss": 0.0038, "step": 23204 }, { "epoch": 0.45, "learning_rate": 4.249558449624439e-05, "loss": 0.0011, "step": 23206 }, { "epoch": 0.45, "learning_rate": 4.249493753598717e-05, "loss": 0.002, "step": 23208 }, { "epoch": 0.45, "learning_rate": 4.249429057572993e-05, "loss": 0.0001, "step": 23210 }, { "epoch": 0.45, "learning_rate": 4.2493643615472707e-05, "loss": 0.0037, "step": 23212 }, { "epoch": 0.45, "learning_rate": 4.2492996655215476e-05, "loss": 0.0002, "step": 23214 }, { "epoch": 0.45, "learning_rate": 4.249234969495824e-05, "loss": 0.0001, "step": 23216 }, { "epoch": 0.45, "learning_rate": 4.2491702734701014e-05, "loss": 0.0004, "step": 23218 }, { "epoch": 0.45, "learning_rate": 4.2491055774443776e-05, "loss": 0.0161, "step": 23220 }, { "epoch": 0.45, "learning_rate": 4.2490408814186545e-05, "loss": 0.0052, "step": 23222 }, { "epoch": 0.45, "learning_rate": 4.2489761853929314e-05, "loss": 0.0094, "step": 23224 }, { "epoch": 0.45, "learning_rate": 4.2489114893672083e-05, "loss": 0.0001, "step": 23226 }, { "epoch": 0.45, "learning_rate": 4.248846793341485e-05, "loss": 0.0001, "step": 23228 }, { "epoch": 0.45, "learning_rate": 4.248782097315762e-05, "loss": 0.0001, "step": 23230 }, { "epoch": 0.45, "learning_rate": 4.248717401290039e-05, "loss": 0.0048, "step": 23232 }, { "epoch": 0.45, "learning_rate": 4.248652705264316e-05, "loss": 0.0011, "step": 23234 }, { "epoch": 0.45, "learning_rate": 4.248588009238593e-05, "loss": 0.0003, "step": 23236 }, { "epoch": 0.45, "learning_rate": 4.248523313212869e-05, "loss": 0.0023, "step": 23238 }, { "epoch": 0.45, "learning_rate": 4.248458617187147e-05, "loss": 0.0005, "step": 23240 }, { "epoch": 0.45, "learning_rate": 4.248393921161423e-05, "loss": 0.0003, "step": 23242 }, { "epoch": 0.45, "learning_rate": 4.2483292251357e-05, "loss": 0.0147, "step": 23244 }, { "epoch": 0.45, "learning_rate": 4.248264529109977e-05, "loss": 0.0001, "step": 23246 }, { "epoch": 0.45, "learning_rate": 4.2481998330842537e-05, "loss": 0.0001, "step": 23248 }, { "epoch": 0.45, "learning_rate": 4.248135137058531e-05, "loss": 0.0231, "step": 23250 }, { "epoch": 0.45, "learning_rate": 4.2480704410328075e-05, "loss": 0.0023, "step": 23252 }, { "epoch": 0.45, "learning_rate": 4.2480057450070844e-05, "loss": 0.0009, "step": 23254 }, { "epoch": 0.45, "learning_rate": 4.247941048981361e-05, "loss": 0.008, "step": 23256 }, { "epoch": 0.45, "learning_rate": 4.247876352955638e-05, "loss": 0.0001, "step": 23258 }, { "epoch": 0.45, "learning_rate": 4.247811656929915e-05, "loss": 0.0001, "step": 23260 }, { "epoch": 0.45, "learning_rate": 4.247746960904192e-05, "loss": 0.0146, "step": 23262 }, { "epoch": 0.45, "learning_rate": 4.247682264878469e-05, "loss": 0.0031, "step": 23264 }, { "epoch": 0.45, "learning_rate": 4.247617568852745e-05, "loss": 0.0014, "step": 23266 }, { "epoch": 0.45, "learning_rate": 4.247552872827023e-05, "loss": 0.0005, "step": 23268 }, { "epoch": 0.45, "learning_rate": 4.247488176801299e-05, "loss": 0.01, "step": 23270 }, { "epoch": 0.45, "learning_rate": 4.2474234807755766e-05, "loss": 0.0, "step": 23272 }, { "epoch": 0.45, "learning_rate": 4.247358784749853e-05, "loss": 0.0, "step": 23274 }, { "epoch": 0.45, "learning_rate": 4.24729408872413e-05, "loss": 0.0565, "step": 23276 }, { "epoch": 0.45, "learning_rate": 4.2472293926984066e-05, "loss": 0.0078, "step": 23278 }, { "epoch": 0.45, "learning_rate": 4.2471646966726835e-05, "loss": 0.0044, "step": 23280 }, { "epoch": 0.45, "learning_rate": 4.2471000006469604e-05, "loss": 0.0002, "step": 23282 }, { "epoch": 0.45, "learning_rate": 4.247035304621237e-05, "loss": 0.0003, "step": 23284 }, { "epoch": 0.45, "learning_rate": 4.246970608595514e-05, "loss": 0.0001, "step": 23286 }, { "epoch": 0.45, "learning_rate": 4.246905912569791e-05, "loss": 0.0035, "step": 23288 }, { "epoch": 0.45, "learning_rate": 4.246841216544068e-05, "loss": 0.0002, "step": 23290 }, { "epoch": 0.45, "learning_rate": 4.246776520518345e-05, "loss": 0.0002, "step": 23292 }, { "epoch": 0.45, "learning_rate": 4.246711824492622e-05, "loss": 0.0001, "step": 23294 }, { "epoch": 0.45, "learning_rate": 4.246647128466899e-05, "loss": 0.0005, "step": 23296 }, { "epoch": 0.45, "learning_rate": 4.246582432441175e-05, "loss": 0.0116, "step": 23298 }, { "epoch": 0.45, "learning_rate": 4.2465177364154526e-05, "loss": 0.0002, "step": 23300 }, { "epoch": 0.45, "learning_rate": 4.246453040389729e-05, "loss": 0.0218, "step": 23302 }, { "epoch": 0.45, "learning_rate": 4.246388344364006e-05, "loss": 0.0, "step": 23304 }, { "epoch": 0.45, "learning_rate": 4.2463236483382826e-05, "loss": 0.0001, "step": 23306 }, { "epoch": 0.45, "learning_rate": 4.2462589523125596e-05, "loss": 0.0079, "step": 23308 }, { "epoch": 0.45, "learning_rate": 4.2461942562868365e-05, "loss": 0.0, "step": 23310 }, { "epoch": 0.45, "learning_rate": 4.2461295602611134e-05, "loss": 0.0005, "step": 23312 }, { "epoch": 0.45, "learning_rate": 4.24606486423539e-05, "loss": 0.0387, "step": 23314 }, { "epoch": 0.45, "learning_rate": 4.246000168209667e-05, "loss": 0.0002, "step": 23316 }, { "epoch": 0.45, "learning_rate": 4.245935472183944e-05, "loss": 0.0086, "step": 23318 }, { "epoch": 0.45, "learning_rate": 4.24587077615822e-05, "loss": 0.001, "step": 23320 }, { "epoch": 0.45, "learning_rate": 4.245806080132498e-05, "loss": 0.0101, "step": 23322 }, { "epoch": 0.45, "learning_rate": 4.245741384106774e-05, "loss": 0.0005, "step": 23324 }, { "epoch": 0.45, "learning_rate": 4.245676688081051e-05, "loss": 0.0006, "step": 23326 }, { "epoch": 0.45, "learning_rate": 4.2456119920553286e-05, "loss": 0.0002, "step": 23328 }, { "epoch": 0.45, "learning_rate": 4.245547296029605e-05, "loss": 0.0029, "step": 23330 }, { "epoch": 0.45, "learning_rate": 4.2454826000038825e-05, "loss": 0.0, "step": 23332 }, { "epoch": 0.45, "learning_rate": 4.245417903978159e-05, "loss": 0.0022, "step": 23334 }, { "epoch": 0.45, "learning_rate": 4.2453532079524356e-05, "loss": 0.0081, "step": 23336 }, { "epoch": 0.45, "learning_rate": 4.2452885119267125e-05, "loss": 0.0001, "step": 23338 }, { "epoch": 0.45, "learning_rate": 4.2452238159009894e-05, "loss": 0.0001, "step": 23340 }, { "epoch": 0.45, "learning_rate": 4.245159119875266e-05, "loss": 0.0023, "step": 23342 }, { "epoch": 0.45, "learning_rate": 4.245094423849543e-05, "loss": 0.0002, "step": 23344 }, { "epoch": 0.45, "learning_rate": 4.24502972782382e-05, "loss": 0.0, "step": 23346 }, { "epoch": 0.45, "learning_rate": 4.2449650317980964e-05, "loss": 0.0001, "step": 23348 }, { "epoch": 0.45, "learning_rate": 4.244900335772374e-05, "loss": 0.0029, "step": 23350 }, { "epoch": 0.45, "learning_rate": 4.24483563974665e-05, "loss": 0.0008, "step": 23352 }, { "epoch": 0.45, "learning_rate": 4.244770943720928e-05, "loss": 0.0006, "step": 23354 }, { "epoch": 0.45, "learning_rate": 4.244706247695204e-05, "loss": 0.0003, "step": 23356 }, { "epoch": 0.45, "learning_rate": 4.244641551669481e-05, "loss": 0.004, "step": 23358 }, { "epoch": 0.45, "learning_rate": 4.2445768556437585e-05, "loss": 0.0001, "step": 23360 }, { "epoch": 0.45, "learning_rate": 4.244512159618035e-05, "loss": 0.0018, "step": 23362 }, { "epoch": 0.45, "learning_rate": 4.2444474635923116e-05, "loss": 0.0051, "step": 23364 }, { "epoch": 0.45, "learning_rate": 4.2443827675665885e-05, "loss": 0.013, "step": 23366 }, { "epoch": 0.45, "learning_rate": 4.2443180715408655e-05, "loss": 0.0041, "step": 23368 }, { "epoch": 0.45, "learning_rate": 4.2442533755151424e-05, "loss": 0.0001, "step": 23370 }, { "epoch": 0.45, "learning_rate": 4.244188679489419e-05, "loss": 0.0154, "step": 23372 }, { "epoch": 0.45, "learning_rate": 4.244123983463696e-05, "loss": 0.0055, "step": 23374 }, { "epoch": 0.45, "learning_rate": 4.244059287437973e-05, "loss": 0.0012, "step": 23376 }, { "epoch": 0.45, "learning_rate": 4.24399459141225e-05, "loss": 0.0348, "step": 23378 }, { "epoch": 0.45, "learning_rate": 4.243929895386526e-05, "loss": 0.0, "step": 23380 }, { "epoch": 0.45, "learning_rate": 4.243865199360804e-05, "loss": 0.0001, "step": 23382 }, { "epoch": 0.45, "learning_rate": 4.24380050333508e-05, "loss": 0.0017, "step": 23384 }, { "epoch": 0.45, "learning_rate": 4.243735807309357e-05, "loss": 0.0002, "step": 23386 }, { "epoch": 0.45, "learning_rate": 4.243671111283634e-05, "loss": 0.0015, "step": 23388 }, { "epoch": 0.45, "learning_rate": 4.243606415257911e-05, "loss": 0.0076, "step": 23390 }, { "epoch": 0.45, "learning_rate": 4.2435417192321884e-05, "loss": 0.0101, "step": 23392 }, { "epoch": 0.45, "learning_rate": 4.2434770232064646e-05, "loss": 0.0017, "step": 23394 }, { "epoch": 0.45, "learning_rate": 4.2434123271807415e-05, "loss": 0.0025, "step": 23396 }, { "epoch": 0.45, "learning_rate": 4.2433476311550184e-05, "loss": 0.0001, "step": 23398 }, { "epoch": 0.45, "learning_rate": 4.243282935129295e-05, "loss": 0.0023, "step": 23400 }, { "epoch": 0.45, "learning_rate": 4.2432182391035715e-05, "loss": 0.0005, "step": 23402 }, { "epoch": 0.45, "learning_rate": 4.243153543077849e-05, "loss": 0.0001, "step": 23404 }, { "epoch": 0.45, "learning_rate": 4.243088847052126e-05, "loss": 0.0001, "step": 23406 }, { "epoch": 0.45, "learning_rate": 4.243024151026402e-05, "loss": 0.0002, "step": 23408 }, { "epoch": 0.45, "learning_rate": 4.24295945500068e-05, "loss": 0.0057, "step": 23410 }, { "epoch": 0.45, "learning_rate": 4.242894758974956e-05, "loss": 0.0418, "step": 23412 }, { "epoch": 0.45, "learning_rate": 4.242830062949234e-05, "loss": 0.0001, "step": 23414 }, { "epoch": 0.45, "learning_rate": 4.24276536692351e-05, "loss": 0.0012, "step": 23416 }, { "epoch": 0.45, "learning_rate": 4.242700670897787e-05, "loss": 0.0001, "step": 23418 }, { "epoch": 0.45, "learning_rate": 4.242635974872064e-05, "loss": 0.0006, "step": 23420 }, { "epoch": 0.45, "learning_rate": 4.2425712788463406e-05, "loss": 0.0001, "step": 23422 }, { "epoch": 0.45, "learning_rate": 4.2425065828206175e-05, "loss": 0.0005, "step": 23424 }, { "epoch": 0.45, "learning_rate": 4.2424418867948944e-05, "loss": 0.0043, "step": 23426 }, { "epoch": 0.45, "learning_rate": 4.2423771907691713e-05, "loss": 0.0014, "step": 23428 }, { "epoch": 0.45, "learning_rate": 4.242312494743448e-05, "loss": 0.0002, "step": 23430 }, { "epoch": 0.45, "learning_rate": 4.242247798717725e-05, "loss": 0.0072, "step": 23432 }, { "epoch": 0.45, "learning_rate": 4.2421831026920014e-05, "loss": 0.0007, "step": 23434 }, { "epoch": 0.45, "learning_rate": 4.242118406666279e-05, "loss": 0.0, "step": 23436 }, { "epoch": 0.45, "learning_rate": 4.242053710640556e-05, "loss": 0.0024, "step": 23438 }, { "epoch": 0.45, "learning_rate": 4.241989014614832e-05, "loss": 0.0007, "step": 23440 }, { "epoch": 0.45, "learning_rate": 4.24192431858911e-05, "loss": 0.0005, "step": 23442 }, { "epoch": 0.46, "learning_rate": 4.241859622563386e-05, "loss": 0.0023, "step": 23444 }, { "epoch": 0.46, "learning_rate": 4.241794926537663e-05, "loss": 0.0007, "step": 23446 }, { "epoch": 0.46, "learning_rate": 4.24173023051194e-05, "loss": 0.0002, "step": 23448 }, { "epoch": 0.46, "learning_rate": 4.241665534486217e-05, "loss": 0.0, "step": 23450 }, { "epoch": 0.46, "learning_rate": 4.2416008384604936e-05, "loss": 0.0093, "step": 23452 }, { "epoch": 0.46, "learning_rate": 4.2415361424347705e-05, "loss": 0.0002, "step": 23454 }, { "epoch": 0.46, "learning_rate": 4.2414714464090474e-05, "loss": 0.0168, "step": 23456 }, { "epoch": 0.46, "learning_rate": 4.241406750383324e-05, "loss": 0.0022, "step": 23458 }, { "epoch": 0.46, "learning_rate": 4.241342054357601e-05, "loss": 0.0001, "step": 23460 }, { "epoch": 0.46, "learning_rate": 4.2412773583318774e-05, "loss": 0.003, "step": 23462 }, { "epoch": 0.46, "learning_rate": 4.241212662306155e-05, "loss": 0.0008, "step": 23464 }, { "epoch": 0.46, "learning_rate": 4.241147966280431e-05, "loss": 0.0001, "step": 23466 }, { "epoch": 0.46, "learning_rate": 4.241083270254708e-05, "loss": 0.0001, "step": 23468 }, { "epoch": 0.46, "learning_rate": 4.241018574228985e-05, "loss": 0.0003, "step": 23470 }, { "epoch": 0.46, "learning_rate": 4.240953878203262e-05, "loss": 0.0018, "step": 23472 }, { "epoch": 0.46, "learning_rate": 4.2408891821775396e-05, "loss": 0.0004, "step": 23474 }, { "epoch": 0.46, "learning_rate": 4.240824486151816e-05, "loss": 0.0044, "step": 23476 }, { "epoch": 0.46, "learning_rate": 4.240759790126093e-05, "loss": 0.0003, "step": 23478 }, { "epoch": 0.46, "learning_rate": 4.2406950941003696e-05, "loss": 0.0042, "step": 23480 }, { "epoch": 0.46, "learning_rate": 4.2406303980746465e-05, "loss": 0.007, "step": 23482 }, { "epoch": 0.46, "learning_rate": 4.2405657020489234e-05, "loss": 0.0047, "step": 23484 }, { "epoch": 0.46, "learning_rate": 4.2405010060232003e-05, "loss": 0.0099, "step": 23486 }, { "epoch": 0.46, "learning_rate": 4.240436309997477e-05, "loss": 0.0042, "step": 23488 }, { "epoch": 0.46, "learning_rate": 4.2403716139717535e-05, "loss": 0.0024, "step": 23490 }, { "epoch": 0.46, "learning_rate": 4.240306917946031e-05, "loss": 0.0002, "step": 23492 }, { "epoch": 0.46, "learning_rate": 4.240242221920307e-05, "loss": 0.0148, "step": 23494 }, { "epoch": 0.46, "learning_rate": 4.240177525894585e-05, "loss": 0.0001, "step": 23496 }, { "epoch": 0.46, "learning_rate": 4.240112829868861e-05, "loss": 0.0009, "step": 23498 }, { "epoch": 0.46, "learning_rate": 4.240048133843138e-05, "loss": 0.0113, "step": 23500 }, { "epoch": 0.46, "learning_rate": 4.239983437817415e-05, "loss": 0.0011, "step": 23502 }, { "epoch": 0.46, "learning_rate": 4.239918741791692e-05, "loss": 0.0003, "step": 23504 }, { "epoch": 0.46, "learning_rate": 4.239854045765969e-05, "loss": 0.0001, "step": 23506 }, { "epoch": 0.46, "learning_rate": 4.2397893497402457e-05, "loss": 0.0184, "step": 23508 }, { "epoch": 0.46, "learning_rate": 4.2397246537145226e-05, "loss": 0.0006, "step": 23510 }, { "epoch": 0.46, "learning_rate": 4.2396599576887995e-05, "loss": 0.0001, "step": 23512 }, { "epoch": 0.46, "learning_rate": 4.2395952616630764e-05, "loss": 0.0053, "step": 23514 }, { "epoch": 0.46, "learning_rate": 4.239530565637353e-05, "loss": 0.0001, "step": 23516 }, { "epoch": 0.46, "learning_rate": 4.23946586961163e-05, "loss": 0.0017, "step": 23518 }, { "epoch": 0.46, "learning_rate": 4.239401173585907e-05, "loss": 0.0022, "step": 23520 }, { "epoch": 0.46, "learning_rate": 4.239336477560183e-05, "loss": 0.0032, "step": 23522 }, { "epoch": 0.46, "learning_rate": 4.239271781534461e-05, "loss": 0.0001, "step": 23524 }, { "epoch": 0.46, "learning_rate": 4.239207085508737e-05, "loss": 0.0013, "step": 23526 }, { "epoch": 0.46, "learning_rate": 4.239142389483014e-05, "loss": 0.0007, "step": 23528 }, { "epoch": 0.46, "learning_rate": 4.239077693457291e-05, "loss": 0.0001, "step": 23530 }, { "epoch": 0.46, "learning_rate": 4.239012997431568e-05, "loss": 0.0, "step": 23532 }, { "epoch": 0.46, "learning_rate": 4.238948301405845e-05, "loss": 0.0, "step": 23534 }, { "epoch": 0.46, "learning_rate": 4.238883605380122e-05, "loss": 0.0007, "step": 23536 }, { "epoch": 0.46, "learning_rate": 4.2388189093543986e-05, "loss": 0.0006, "step": 23538 }, { "epoch": 0.46, "learning_rate": 4.2387542133286755e-05, "loss": 0.0151, "step": 23540 }, { "epoch": 0.46, "learning_rate": 4.2386895173029524e-05, "loss": 0.0006, "step": 23542 }, { "epoch": 0.46, "learning_rate": 4.2386248212772287e-05, "loss": 0.0002, "step": 23544 }, { "epoch": 0.46, "learning_rate": 4.238560125251506e-05, "loss": 0.0197, "step": 23546 }, { "epoch": 0.46, "learning_rate": 4.2384954292257825e-05, "loss": 0.012, "step": 23548 }, { "epoch": 0.46, "learning_rate": 4.2384307332000594e-05, "loss": 0.0066, "step": 23550 }, { "epoch": 0.46, "learning_rate": 4.238366037174337e-05, "loss": 0.0004, "step": 23552 }, { "epoch": 0.46, "learning_rate": 4.238301341148613e-05, "loss": 0.0092, "step": 23554 }, { "epoch": 0.46, "learning_rate": 4.238236645122891e-05, "loss": 0.0004, "step": 23556 }, { "epoch": 0.46, "learning_rate": 4.238171949097167e-05, "loss": 0.0177, "step": 23558 }, { "epoch": 0.46, "learning_rate": 4.238107253071444e-05, "loss": 0.0011, "step": 23560 }, { "epoch": 0.46, "learning_rate": 4.238042557045721e-05, "loss": 0.0218, "step": 23562 }, { "epoch": 0.46, "learning_rate": 4.237977861019998e-05, "loss": 0.0001, "step": 23564 }, { "epoch": 0.46, "learning_rate": 4.2379131649942746e-05, "loss": 0.0, "step": 23566 }, { "epoch": 0.46, "learning_rate": 4.2378484689685516e-05, "loss": 0.0231, "step": 23568 }, { "epoch": 0.46, "learning_rate": 4.2377837729428285e-05, "loss": 0.0025, "step": 23570 }, { "epoch": 0.46, "learning_rate": 4.2377190769171054e-05, "loss": 0.0027, "step": 23572 }, { "epoch": 0.46, "learning_rate": 4.237654380891382e-05, "loss": 0.0001, "step": 23574 }, { "epoch": 0.46, "learning_rate": 4.2375896848656585e-05, "loss": 0.0115, "step": 23576 }, { "epoch": 0.46, "learning_rate": 4.237524988839936e-05, "loss": 0.0004, "step": 23578 }, { "epoch": 0.46, "learning_rate": 4.237460292814212e-05, "loss": 0.0001, "step": 23580 }, { "epoch": 0.46, "learning_rate": 4.237395596788489e-05, "loss": 0.0035, "step": 23582 }, { "epoch": 0.46, "learning_rate": 4.237330900762767e-05, "loss": 0.0011, "step": 23584 }, { "epoch": 0.46, "learning_rate": 4.237266204737043e-05, "loss": 0.0093, "step": 23586 }, { "epoch": 0.46, "learning_rate": 4.23720150871132e-05, "loss": 0.0003, "step": 23588 }, { "epoch": 0.46, "learning_rate": 4.237136812685597e-05, "loss": 0.0004, "step": 23590 }, { "epoch": 0.46, "learning_rate": 4.237072116659874e-05, "loss": 0.0032, "step": 23592 }, { "epoch": 0.46, "learning_rate": 4.237007420634151e-05, "loss": 0.0001, "step": 23594 }, { "epoch": 0.46, "learning_rate": 4.2369427246084276e-05, "loss": 0.005, "step": 23596 }, { "epoch": 0.46, "learning_rate": 4.2368780285827045e-05, "loss": 0.0002, "step": 23598 }, { "epoch": 0.46, "learning_rate": 4.2368133325569814e-05, "loss": 0.0009, "step": 23600 }, { "epoch": 0.46, "learning_rate": 4.236748636531258e-05, "loss": 0.0004, "step": 23602 }, { "epoch": 0.46, "learning_rate": 4.2366839405055345e-05, "loss": 0.0044, "step": 23604 }, { "epoch": 0.46, "learning_rate": 4.236619244479812e-05, "loss": 0.0005, "step": 23606 }, { "epoch": 0.46, "learning_rate": 4.2365545484540884e-05, "loss": 0.0049, "step": 23608 }, { "epoch": 0.46, "learning_rate": 4.236489852428365e-05, "loss": 0.0002, "step": 23610 }, { "epoch": 0.46, "learning_rate": 4.236425156402642e-05, "loss": 0.0086, "step": 23612 }, { "epoch": 0.46, "learning_rate": 4.236360460376919e-05, "loss": 0.0002, "step": 23614 }, { "epoch": 0.46, "learning_rate": 4.236295764351197e-05, "loss": 0.0, "step": 23616 }, { "epoch": 0.46, "learning_rate": 4.236231068325473e-05, "loss": 0.0008, "step": 23618 }, { "epoch": 0.46, "learning_rate": 4.23616637229975e-05, "loss": 0.0, "step": 23620 }, { "epoch": 0.46, "learning_rate": 4.236101676274027e-05, "loss": 0.0003, "step": 23622 }, { "epoch": 0.46, "learning_rate": 4.2360369802483036e-05, "loss": 0.0022, "step": 23624 }, { "epoch": 0.46, "learning_rate": 4.23597228422258e-05, "loss": 0.0036, "step": 23626 }, { "epoch": 0.46, "learning_rate": 4.2359075881968575e-05, "loss": 0.0001, "step": 23628 }, { "epoch": 0.46, "learning_rate": 4.2358428921711344e-05, "loss": 0.0001, "step": 23630 }, { "epoch": 0.46, "learning_rate": 4.2357781961454106e-05, "loss": 0.0107, "step": 23632 }, { "epoch": 0.46, "learning_rate": 4.235713500119688e-05, "loss": 0.0001, "step": 23634 }, { "epoch": 0.46, "learning_rate": 4.2356488040939644e-05, "loss": 0.0004, "step": 23636 }, { "epoch": 0.46, "learning_rate": 4.235584108068242e-05, "loss": 0.0002, "step": 23638 }, { "epoch": 0.46, "learning_rate": 4.235519412042518e-05, "loss": 0.0001, "step": 23640 }, { "epoch": 0.46, "learning_rate": 4.235454716016795e-05, "loss": 0.0001, "step": 23642 }, { "epoch": 0.46, "learning_rate": 4.235390019991072e-05, "loss": 0.0039, "step": 23644 }, { "epoch": 0.46, "learning_rate": 4.235325323965349e-05, "loss": 0.0033, "step": 23646 }, { "epoch": 0.46, "learning_rate": 4.235260627939626e-05, "loss": 0.0, "step": 23648 }, { "epoch": 0.46, "learning_rate": 4.235195931913903e-05, "loss": 0.0001, "step": 23650 }, { "epoch": 0.46, "learning_rate": 4.23513123588818e-05, "loss": 0.0138, "step": 23652 }, { "epoch": 0.46, "learning_rate": 4.2350665398624566e-05, "loss": 0.0002, "step": 23654 }, { "epoch": 0.46, "learning_rate": 4.2350018438367335e-05, "loss": 0.0011, "step": 23656 }, { "epoch": 0.46, "learning_rate": 4.23493714781101e-05, "loss": 0.0023, "step": 23658 }, { "epoch": 0.46, "learning_rate": 4.234872451785287e-05, "loss": 0.002, "step": 23660 }, { "epoch": 0.46, "learning_rate": 4.234807755759564e-05, "loss": 0.0002, "step": 23662 }, { "epoch": 0.46, "learning_rate": 4.2347430597338404e-05, "loss": 0.0017, "step": 23664 }, { "epoch": 0.46, "learning_rate": 4.234678363708118e-05, "loss": 0.0, "step": 23666 }, { "epoch": 0.46, "learning_rate": 4.234613667682394e-05, "loss": 0.0001, "step": 23668 }, { "epoch": 0.46, "learning_rate": 4.234548971656671e-05, "loss": 0.004, "step": 23670 }, { "epoch": 0.46, "learning_rate": 4.234484275630948e-05, "loss": 0.0003, "step": 23672 }, { "epoch": 0.46, "learning_rate": 4.234419579605225e-05, "loss": 0.0024, "step": 23674 }, { "epoch": 0.46, "learning_rate": 4.234354883579502e-05, "loss": 0.0096, "step": 23676 }, { "epoch": 0.46, "learning_rate": 4.234290187553779e-05, "loss": 0.0, "step": 23678 }, { "epoch": 0.46, "learning_rate": 4.234225491528056e-05, "loss": 0.0052, "step": 23680 }, { "epoch": 0.46, "learning_rate": 4.2341607955023326e-05, "loss": 0.0001, "step": 23682 }, { "epoch": 0.46, "learning_rate": 4.2340960994766095e-05, "loss": 0.0019, "step": 23684 }, { "epoch": 0.46, "learning_rate": 4.234031403450886e-05, "loss": 0.0052, "step": 23686 }, { "epoch": 0.46, "learning_rate": 4.2339667074251633e-05, "loss": 0.0069, "step": 23688 }, { "epoch": 0.46, "learning_rate": 4.2339020113994396e-05, "loss": 0.0001, "step": 23690 }, { "epoch": 0.46, "learning_rate": 4.2338373153737165e-05, "loss": 0.0046, "step": 23692 }, { "epoch": 0.46, "learning_rate": 4.233772619347994e-05, "loss": 0.0003, "step": 23694 }, { "epoch": 0.46, "learning_rate": 4.23370792332227e-05, "loss": 0.0001, "step": 23696 }, { "epoch": 0.46, "learning_rate": 4.233643227296548e-05, "loss": 0.0008, "step": 23698 }, { "epoch": 0.46, "learning_rate": 4.233578531270824e-05, "loss": 0.001, "step": 23700 }, { "epoch": 0.46, "learning_rate": 4.233513835245101e-05, "loss": 0.0002, "step": 23702 }, { "epoch": 0.46, "learning_rate": 4.233449139219378e-05, "loss": 0.0027, "step": 23704 }, { "epoch": 0.46, "learning_rate": 4.233384443193655e-05, "loss": 0.0033, "step": 23706 }, { "epoch": 0.46, "learning_rate": 4.233319747167932e-05, "loss": 0.0002, "step": 23708 }, { "epoch": 0.46, "learning_rate": 4.233255051142209e-05, "loss": 0.0006, "step": 23710 }, { "epoch": 0.46, "learning_rate": 4.2331903551164856e-05, "loss": 0.0072, "step": 23712 }, { "epoch": 0.46, "learning_rate": 4.2331256590907625e-05, "loss": 0.0013, "step": 23714 }, { "epoch": 0.46, "learning_rate": 4.2330609630650394e-05, "loss": 0.0, "step": 23716 }, { "epoch": 0.46, "learning_rate": 4.2329962670393156e-05, "loss": 0.0, "step": 23718 }, { "epoch": 0.46, "learning_rate": 4.232931571013593e-05, "loss": 0.0074, "step": 23720 }, { "epoch": 0.46, "learning_rate": 4.2328668749878694e-05, "loss": 0.0003, "step": 23722 }, { "epoch": 0.46, "learning_rate": 4.2328021789621463e-05, "loss": 0.0004, "step": 23724 }, { "epoch": 0.46, "learning_rate": 4.232737482936423e-05, "loss": 0.0019, "step": 23726 }, { "epoch": 0.46, "learning_rate": 4.2326727869107e-05, "loss": 0.0003, "step": 23728 }, { "epoch": 0.46, "learning_rate": 4.232608090884977e-05, "loss": 0.0009, "step": 23730 }, { "epoch": 0.46, "learning_rate": 4.232543394859254e-05, "loss": 0.0017, "step": 23732 }, { "epoch": 0.46, "learning_rate": 4.232478698833531e-05, "loss": 0.0008, "step": 23734 }, { "epoch": 0.46, "learning_rate": 4.232414002807808e-05, "loss": 0.0004, "step": 23736 }, { "epoch": 0.46, "learning_rate": 4.232349306782085e-05, "loss": 0.0, "step": 23738 }, { "epoch": 0.46, "learning_rate": 4.2322846107563616e-05, "loss": 0.0005, "step": 23740 }, { "epoch": 0.46, "learning_rate": 4.2322199147306385e-05, "loss": 0.0001, "step": 23742 }, { "epoch": 0.46, "learning_rate": 4.2321552187049154e-05, "loss": 0.0005, "step": 23744 }, { "epoch": 0.46, "learning_rate": 4.2320905226791917e-05, "loss": 0.0007, "step": 23746 }, { "epoch": 0.46, "learning_rate": 4.232025826653469e-05, "loss": 0.0047, "step": 23748 }, { "epoch": 0.46, "learning_rate": 4.2319611306277455e-05, "loss": 0.0092, "step": 23750 }, { "epoch": 0.46, "learning_rate": 4.2318964346020224e-05, "loss": 0.0037, "step": 23752 }, { "epoch": 0.46, "learning_rate": 4.231831738576299e-05, "loss": 0.0002, "step": 23754 }, { "epoch": 0.46, "learning_rate": 4.231767042550576e-05, "loss": 0.0041, "step": 23756 }, { "epoch": 0.46, "learning_rate": 4.231702346524853e-05, "loss": 0.0035, "step": 23758 }, { "epoch": 0.46, "learning_rate": 4.23163765049913e-05, "loss": 0.0004, "step": 23760 }, { "epoch": 0.46, "learning_rate": 4.231572954473407e-05, "loss": 0.0001, "step": 23762 }, { "epoch": 0.46, "learning_rate": 4.231508258447684e-05, "loss": 0.0014, "step": 23764 }, { "epoch": 0.46, "learning_rate": 4.231443562421961e-05, "loss": 0.012, "step": 23766 }, { "epoch": 0.46, "learning_rate": 4.231378866396237e-05, "loss": 0.0012, "step": 23768 }, { "epoch": 0.46, "learning_rate": 4.2313141703705146e-05, "loss": 0.0, "step": 23770 }, { "epoch": 0.46, "learning_rate": 4.2312494743447915e-05, "loss": 0.0043, "step": 23772 }, { "epoch": 0.46, "learning_rate": 4.231184778319068e-05, "loss": 0.0002, "step": 23774 }, { "epoch": 0.46, "learning_rate": 4.231120082293345e-05, "loss": 0.0016, "step": 23776 }, { "epoch": 0.46, "learning_rate": 4.2310553862676215e-05, "loss": 0.0, "step": 23778 }, { "epoch": 0.46, "learning_rate": 4.230990690241899e-05, "loss": 0.0005, "step": 23780 }, { "epoch": 0.46, "learning_rate": 4.230925994216175e-05, "loss": 0.0002, "step": 23782 }, { "epoch": 0.46, "learning_rate": 4.230861298190452e-05, "loss": 0.0, "step": 23784 }, { "epoch": 0.46, "learning_rate": 4.230796602164729e-05, "loss": 0.0001, "step": 23786 }, { "epoch": 0.46, "learning_rate": 4.230731906139006e-05, "loss": 0.0001, "step": 23788 }, { "epoch": 0.46, "learning_rate": 4.230667210113283e-05, "loss": 0.0056, "step": 23790 }, { "epoch": 0.46, "learning_rate": 4.23060251408756e-05, "loss": 0.0001, "step": 23792 }, { "epoch": 0.46, "learning_rate": 4.230537818061837e-05, "loss": 0.0001, "step": 23794 }, { "epoch": 0.46, "learning_rate": 4.230473122036114e-05, "loss": 0.0008, "step": 23796 }, { "epoch": 0.46, "learning_rate": 4.2304084260103906e-05, "loss": 0.0008, "step": 23798 }, { "epoch": 0.46, "learning_rate": 4.230343729984667e-05, "loss": 0.0004, "step": 23800 }, { "epoch": 0.46, "learning_rate": 4.2302790339589444e-05, "loss": 0.0022, "step": 23802 }, { "epoch": 0.46, "learning_rate": 4.2302143379332206e-05, "loss": 0.0442, "step": 23804 }, { "epoch": 0.46, "learning_rate": 4.2301496419074976e-05, "loss": 0.0002, "step": 23806 }, { "epoch": 0.46, "learning_rate": 4.230084945881775e-05, "loss": 0.0, "step": 23808 }, { "epoch": 0.46, "learning_rate": 4.2300202498560514e-05, "loss": 0.0007, "step": 23810 }, { "epoch": 0.46, "learning_rate": 4.229955553830328e-05, "loss": 0.0011, "step": 23812 }, { "epoch": 0.46, "learning_rate": 4.229890857804605e-05, "loss": 0.002, "step": 23814 }, { "epoch": 0.46, "learning_rate": 4.229826161778882e-05, "loss": 0.0134, "step": 23816 }, { "epoch": 0.46, "learning_rate": 4.229761465753159e-05, "loss": 0.0063, "step": 23818 }, { "epoch": 0.46, "learning_rate": 4.229696769727436e-05, "loss": 0.0015, "step": 23820 }, { "epoch": 0.46, "learning_rate": 4.229632073701713e-05, "loss": 0.0003, "step": 23822 }, { "epoch": 0.46, "learning_rate": 4.22956737767599e-05, "loss": 0.0048, "step": 23824 }, { "epoch": 0.46, "learning_rate": 4.2295026816502666e-05, "loss": 0.0049, "step": 23826 }, { "epoch": 0.46, "learning_rate": 4.229437985624543e-05, "loss": 0.0004, "step": 23828 }, { "epoch": 0.46, "learning_rate": 4.2293732895988205e-05, "loss": 0.0109, "step": 23830 }, { "epoch": 0.46, "learning_rate": 4.229308593573097e-05, "loss": 0.0002, "step": 23832 }, { "epoch": 0.46, "learning_rate": 4.2292438975473736e-05, "loss": 0.0001, "step": 23834 }, { "epoch": 0.46, "learning_rate": 4.2291792015216505e-05, "loss": 0.0232, "step": 23836 }, { "epoch": 0.46, "learning_rate": 4.2291145054959274e-05, "loss": 0.0006, "step": 23838 }, { "epoch": 0.46, "learning_rate": 4.229049809470205e-05, "loss": 0.0002, "step": 23840 }, { "epoch": 0.46, "learning_rate": 4.228985113444481e-05, "loss": 0.011, "step": 23842 }, { "epoch": 0.46, "learning_rate": 4.228920417418758e-05, "loss": 0.0001, "step": 23844 }, { "epoch": 0.46, "learning_rate": 4.228855721393035e-05, "loss": 0.0189, "step": 23846 }, { "epoch": 0.46, "learning_rate": 4.228791025367312e-05, "loss": 0.0001, "step": 23848 }, { "epoch": 0.46, "learning_rate": 4.228726329341588e-05, "loss": 0.0933, "step": 23850 }, { "epoch": 0.46, "learning_rate": 4.228661633315866e-05, "loss": 0.0002, "step": 23852 }, { "epoch": 0.46, "learning_rate": 4.228596937290143e-05, "loss": 0.0352, "step": 23854 }, { "epoch": 0.46, "learning_rate": 4.2285322412644196e-05, "loss": 0.01, "step": 23856 }, { "epoch": 0.46, "learning_rate": 4.2284675452386965e-05, "loss": 0.0025, "step": 23858 }, { "epoch": 0.46, "learning_rate": 4.228402849212973e-05, "loss": 0.0001, "step": 23860 }, { "epoch": 0.46, "learning_rate": 4.22833815318725e-05, "loss": 0.0002, "step": 23862 }, { "epoch": 0.46, "learning_rate": 4.2282734571615265e-05, "loss": 0.0002, "step": 23864 }, { "epoch": 0.46, "learning_rate": 4.2282087611358035e-05, "loss": 0.0034, "step": 23866 }, { "epoch": 0.46, "learning_rate": 4.2281440651100804e-05, "loss": 0.0001, "step": 23868 }, { "epoch": 0.46, "learning_rate": 4.228079369084357e-05, "loss": 0.0029, "step": 23870 }, { "epoch": 0.46, "learning_rate": 4.228014673058634e-05, "loss": 0.0094, "step": 23872 }, { "epoch": 0.46, "learning_rate": 4.227949977032911e-05, "loss": 0.0012, "step": 23874 }, { "epoch": 0.46, "learning_rate": 4.227885281007188e-05, "loss": 0.0001, "step": 23876 }, { "epoch": 0.46, "learning_rate": 4.227820584981465e-05, "loss": 0.0, "step": 23878 }, { "epoch": 0.46, "learning_rate": 4.227755888955742e-05, "loss": 0.0028, "step": 23880 }, { "epoch": 0.46, "learning_rate": 4.227691192930018e-05, "loss": 0.0002, "step": 23882 }, { "epoch": 0.46, "learning_rate": 4.2276264969042956e-05, "loss": 0.0029, "step": 23884 }, { "epoch": 0.46, "learning_rate": 4.2275618008785725e-05, "loss": 0.0027, "step": 23886 }, { "epoch": 0.46, "learning_rate": 4.227497104852849e-05, "loss": 0.0003, "step": 23888 }, { "epoch": 0.46, "learning_rate": 4.2274324088271264e-05, "loss": 0.0051, "step": 23890 }, { "epoch": 0.46, "learning_rate": 4.2273677128014026e-05, "loss": 0.0044, "step": 23892 }, { "epoch": 0.46, "learning_rate": 4.2273030167756795e-05, "loss": 0.0005, "step": 23894 }, { "epoch": 0.46, "learning_rate": 4.2272383207499564e-05, "loss": 0.0, "step": 23896 }, { "epoch": 0.46, "learning_rate": 4.227173624724233e-05, "loss": 0.0001, "step": 23898 }, { "epoch": 0.46, "learning_rate": 4.22710892869851e-05, "loss": 0.0, "step": 23900 }, { "epoch": 0.46, "learning_rate": 4.227044232672787e-05, "loss": 0.0018, "step": 23902 }, { "epoch": 0.46, "learning_rate": 4.226979536647064e-05, "loss": 0.0015, "step": 23904 }, { "epoch": 0.46, "learning_rate": 4.226914840621341e-05, "loss": 0.0, "step": 23906 }, { "epoch": 0.46, "learning_rate": 4.226850144595618e-05, "loss": 0.0004, "step": 23908 }, { "epoch": 0.46, "learning_rate": 4.226785448569894e-05, "loss": 0.0002, "step": 23910 }, { "epoch": 0.46, "learning_rate": 4.226720752544172e-05, "loss": 0.0, "step": 23912 }, { "epoch": 0.46, "learning_rate": 4.226656056518448e-05, "loss": 0.0003, "step": 23914 }, { "epoch": 0.46, "learning_rate": 4.226591360492725e-05, "loss": 0.0002, "step": 23916 }, { "epoch": 0.46, "learning_rate": 4.2265266644670024e-05, "loss": 0.0, "step": 23918 }, { "epoch": 0.46, "learning_rate": 4.2264619684412786e-05, "loss": 0.0116, "step": 23920 }, { "epoch": 0.46, "learning_rate": 4.226397272415556e-05, "loss": 0.0007, "step": 23922 }, { "epoch": 0.46, "learning_rate": 4.2263325763898324e-05, "loss": 0.0155, "step": 23924 }, { "epoch": 0.46, "learning_rate": 4.2262678803641094e-05, "loss": 0.0071, "step": 23926 }, { "epoch": 0.46, "learning_rate": 4.226203184338386e-05, "loss": 0.0001, "step": 23928 }, { "epoch": 0.46, "learning_rate": 4.226138488312663e-05, "loss": 0.0001, "step": 23930 }, { "epoch": 0.46, "learning_rate": 4.22607379228694e-05, "loss": 0.0007, "step": 23932 }, { "epoch": 0.46, "learning_rate": 4.226009096261217e-05, "loss": 0.0003, "step": 23934 }, { "epoch": 0.46, "learning_rate": 4.225944400235494e-05, "loss": 0.0046, "step": 23936 }, { "epoch": 0.46, "learning_rate": 4.225879704209771e-05, "loss": 0.0016, "step": 23938 }, { "epoch": 0.46, "learning_rate": 4.225815008184048e-05, "loss": 0.0045, "step": 23940 }, { "epoch": 0.46, "learning_rate": 4.225750312158324e-05, "loss": 0.0001, "step": 23942 }, { "epoch": 0.46, "learning_rate": 4.2256856161326015e-05, "loss": 0.0, "step": 23944 }, { "epoch": 0.46, "learning_rate": 4.225620920106878e-05, "loss": 0.0008, "step": 23946 }, { "epoch": 0.46, "learning_rate": 4.225556224081155e-05, "loss": 0.0001, "step": 23948 }, { "epoch": 0.46, "learning_rate": 4.2254915280554316e-05, "loss": 0.0, "step": 23950 }, { "epoch": 0.46, "learning_rate": 4.2254268320297085e-05, "loss": 0.0169, "step": 23952 }, { "epoch": 0.46, "learning_rate": 4.2253621360039854e-05, "loss": 0.0018, "step": 23954 }, { "epoch": 0.46, "learning_rate": 4.225297439978262e-05, "loss": 0.0002, "step": 23956 }, { "epoch": 0.46, "learning_rate": 4.225232743952539e-05, "loss": 0.0003, "step": 23958 }, { "epoch": 0.47, "learning_rate": 4.225168047926816e-05, "loss": 0.006, "step": 23960 }, { "epoch": 0.47, "learning_rate": 4.225103351901093e-05, "loss": 0.0002, "step": 23962 }, { "epoch": 0.47, "learning_rate": 4.22503865587537e-05, "loss": 0.001, "step": 23964 }, { "epoch": 0.47, "learning_rate": 4.224973959849647e-05, "loss": 0.0102, "step": 23966 }, { "epoch": 0.47, "learning_rate": 4.224909263823924e-05, "loss": 0.0, "step": 23968 }, { "epoch": 0.47, "learning_rate": 4.2248445677982e-05, "loss": 0.0001, "step": 23970 }, { "epoch": 0.47, "learning_rate": 4.2247798717724776e-05, "loss": 0.0079, "step": 23972 }, { "epoch": 0.47, "learning_rate": 4.224715175746754e-05, "loss": 0.0006, "step": 23974 }, { "epoch": 0.47, "learning_rate": 4.224650479721031e-05, "loss": 0.0023, "step": 23976 }, { "epoch": 0.47, "learning_rate": 4.2245857836953076e-05, "loss": 0.0019, "step": 23978 }, { "epoch": 0.47, "learning_rate": 4.2245210876695845e-05, "loss": 0.0379, "step": 23980 }, { "epoch": 0.47, "learning_rate": 4.2244563916438614e-05, "loss": 0.0008, "step": 23982 }, { "epoch": 0.47, "learning_rate": 4.2243916956181383e-05, "loss": 0.0043, "step": 23984 }, { "epoch": 0.47, "learning_rate": 4.224326999592415e-05, "loss": 0.0003, "step": 23986 }, { "epoch": 0.47, "learning_rate": 4.224262303566692e-05, "loss": 0.0001, "step": 23988 }, { "epoch": 0.47, "learning_rate": 4.224197607540969e-05, "loss": 0.002, "step": 23990 }, { "epoch": 0.47, "learning_rate": 4.224132911515245e-05, "loss": 0.0002, "step": 23992 }, { "epoch": 0.47, "learning_rate": 4.224068215489523e-05, "loss": 0.0017, "step": 23994 }, { "epoch": 0.47, "learning_rate": 4.2240035194638e-05, "loss": 0.0012, "step": 23996 }, { "epoch": 0.47, "learning_rate": 4.223938823438077e-05, "loss": 0.004, "step": 23998 }, { "epoch": 0.47, "learning_rate": 4.2238741274123536e-05, "loss": 0.001, "step": 24000 }, { "epoch": 0.47, "learning_rate": 4.22380943138663e-05, "loss": 0.0001, "step": 24002 }, { "epoch": 0.47, "learning_rate": 4.2237447353609074e-05, "loss": 0.0, "step": 24004 }, { "epoch": 0.47, "learning_rate": 4.2236800393351837e-05, "loss": 0.0, "step": 24006 }, { "epoch": 0.47, "learning_rate": 4.2236153433094606e-05, "loss": 0.0, "step": 24008 }, { "epoch": 0.47, "learning_rate": 4.2235506472837375e-05, "loss": 0.0001, "step": 24010 }, { "epoch": 0.47, "learning_rate": 4.2234859512580144e-05, "loss": 0.0001, "step": 24012 }, { "epoch": 0.47, "learning_rate": 4.223421255232291e-05, "loss": 0.0063, "step": 24014 }, { "epoch": 0.47, "learning_rate": 4.223356559206568e-05, "loss": 0.0002, "step": 24016 }, { "epoch": 0.47, "learning_rate": 4.223291863180845e-05, "loss": 0.0002, "step": 24018 }, { "epoch": 0.47, "learning_rate": 4.223227167155122e-05, "loss": 0.0015, "step": 24020 }, { "epoch": 0.47, "learning_rate": 4.223162471129399e-05, "loss": 0.0, "step": 24022 }, { "epoch": 0.47, "learning_rate": 4.223097775103675e-05, "loss": 0.0001, "step": 24024 }, { "epoch": 0.47, "learning_rate": 4.223033079077953e-05, "loss": 0.0, "step": 24026 }, { "epoch": 0.47, "learning_rate": 4.222968383052229e-05, "loss": 0.0246, "step": 24028 }, { "epoch": 0.47, "learning_rate": 4.222903687026506e-05, "loss": 0.0002, "step": 24030 }, { "epoch": 0.47, "learning_rate": 4.2228389910007835e-05, "loss": 0.0001, "step": 24032 }, { "epoch": 0.47, "learning_rate": 4.22277429497506e-05, "loss": 0.001, "step": 24034 }, { "epoch": 0.47, "learning_rate": 4.2227095989493366e-05, "loss": 0.0002, "step": 24036 }, { "epoch": 0.47, "learning_rate": 4.2226449029236135e-05, "loss": 0.0011, "step": 24038 }, { "epoch": 0.47, "learning_rate": 4.2225802068978904e-05, "loss": 0.0001, "step": 24040 }, { "epoch": 0.47, "learning_rate": 4.222515510872167e-05, "loss": 0.0002, "step": 24042 }, { "epoch": 0.47, "learning_rate": 4.222450814846444e-05, "loss": 0.0019, "step": 24044 }, { "epoch": 0.47, "learning_rate": 4.222386118820721e-05, "loss": 0.0001, "step": 24046 }, { "epoch": 0.47, "learning_rate": 4.222321422794998e-05, "loss": 0.0002, "step": 24048 }, { "epoch": 0.47, "learning_rate": 4.222256726769275e-05, "loss": 0.0056, "step": 24050 }, { "epoch": 0.47, "learning_rate": 4.222192030743551e-05, "loss": 0.0, "step": 24052 }, { "epoch": 0.47, "learning_rate": 4.222127334717829e-05, "loss": 0.0021, "step": 24054 }, { "epoch": 0.47, "learning_rate": 4.222062638692105e-05, "loss": 0.0001, "step": 24056 }, { "epoch": 0.47, "learning_rate": 4.221997942666382e-05, "loss": 0.0049, "step": 24058 }, { "epoch": 0.47, "learning_rate": 4.221933246640659e-05, "loss": 0.0001, "step": 24060 }, { "epoch": 0.47, "learning_rate": 4.221868550614936e-05, "loss": 0.0, "step": 24062 }, { "epoch": 0.47, "learning_rate": 4.221803854589213e-05, "loss": 0.0082, "step": 24064 }, { "epoch": 0.47, "learning_rate": 4.2217391585634896e-05, "loss": 0.0001, "step": 24066 }, { "epoch": 0.47, "learning_rate": 4.2216744625377665e-05, "loss": 0.0027, "step": 24068 }, { "epoch": 0.47, "learning_rate": 4.2216097665120434e-05, "loss": 0.0004, "step": 24070 }, { "epoch": 0.47, "learning_rate": 4.22154507048632e-05, "loss": 0.0053, "step": 24072 }, { "epoch": 0.47, "learning_rate": 4.221480374460597e-05, "loss": 0.0001, "step": 24074 }, { "epoch": 0.47, "learning_rate": 4.221415678434874e-05, "loss": 0.0001, "step": 24076 }, { "epoch": 0.47, "learning_rate": 4.221350982409151e-05, "loss": 0.0, "step": 24078 }, { "epoch": 0.47, "learning_rate": 4.221286286383428e-05, "loss": 0.0012, "step": 24080 }, { "epoch": 0.47, "learning_rate": 4.221221590357705e-05, "loss": 0.006, "step": 24082 }, { "epoch": 0.47, "learning_rate": 4.221156894331981e-05, "loss": 0.0138, "step": 24084 }, { "epoch": 0.47, "learning_rate": 4.2210921983062586e-05, "loss": 0.0001, "step": 24086 }, { "epoch": 0.47, "learning_rate": 4.221027502280535e-05, "loss": 0.0113, "step": 24088 }, { "epoch": 0.47, "learning_rate": 4.220962806254812e-05, "loss": 0.0562, "step": 24090 }, { "epoch": 0.47, "learning_rate": 4.220898110229089e-05, "loss": 0.0008, "step": 24092 }, { "epoch": 0.47, "learning_rate": 4.2208334142033656e-05, "loss": 0.0004, "step": 24094 }, { "epoch": 0.47, "learning_rate": 4.2207687181776425e-05, "loss": 0.0002, "step": 24096 }, { "epoch": 0.47, "learning_rate": 4.2207040221519194e-05, "loss": 0.0004, "step": 24098 }, { "epoch": 0.47, "learning_rate": 4.220639326126196e-05, "loss": 0.0001, "step": 24100 }, { "epoch": 0.47, "learning_rate": 4.220574630100473e-05, "loss": 0.0001, "step": 24102 }, { "epoch": 0.47, "learning_rate": 4.22050993407475e-05, "loss": 0.0001, "step": 24104 }, { "epoch": 0.47, "learning_rate": 4.2204452380490264e-05, "loss": 0.0005, "step": 24106 }, { "epoch": 0.47, "learning_rate": 4.220380542023304e-05, "loss": 0.0006, "step": 24108 }, { "epoch": 0.47, "learning_rate": 4.220315845997581e-05, "loss": 0.0015, "step": 24110 }, { "epoch": 0.47, "learning_rate": 4.220251149971857e-05, "loss": 0.0001, "step": 24112 }, { "epoch": 0.47, "learning_rate": 4.220186453946135e-05, "loss": 0.0001, "step": 24114 }, { "epoch": 0.47, "learning_rate": 4.220121757920411e-05, "loss": 0.0313, "step": 24116 }, { "epoch": 0.47, "learning_rate": 4.220057061894688e-05, "loss": 0.0001, "step": 24118 }, { "epoch": 0.47, "learning_rate": 4.219992365868965e-05, "loss": 0.0003, "step": 24120 }, { "epoch": 0.47, "learning_rate": 4.2199276698432416e-05, "loss": 0.0221, "step": 24122 }, { "epoch": 0.47, "learning_rate": 4.2198629738175185e-05, "loss": 0.0001, "step": 24124 }, { "epoch": 0.47, "learning_rate": 4.2197982777917955e-05, "loss": 0.0141, "step": 24126 }, { "epoch": 0.47, "learning_rate": 4.2197335817660724e-05, "loss": 0.0004, "step": 24128 }, { "epoch": 0.47, "learning_rate": 4.219668885740349e-05, "loss": 0.0001, "step": 24130 }, { "epoch": 0.47, "learning_rate": 4.219604189714626e-05, "loss": 0.0051, "step": 24132 }, { "epoch": 0.47, "learning_rate": 4.2195394936889024e-05, "loss": 0.0012, "step": 24134 }, { "epoch": 0.47, "learning_rate": 4.21947479766318e-05, "loss": 0.0001, "step": 24136 }, { "epoch": 0.47, "learning_rate": 4.219410101637456e-05, "loss": 0.0004, "step": 24138 }, { "epoch": 0.47, "learning_rate": 4.219345405611734e-05, "loss": 0.0004, "step": 24140 }, { "epoch": 0.47, "learning_rate": 4.219280709586011e-05, "loss": 0.0236, "step": 24142 }, { "epoch": 0.47, "learning_rate": 4.219216013560287e-05, "loss": 0.0036, "step": 24144 }, { "epoch": 0.47, "learning_rate": 4.2191513175345645e-05, "loss": 0.0001, "step": 24146 }, { "epoch": 0.47, "learning_rate": 4.219086621508841e-05, "loss": 0.0001, "step": 24148 }, { "epoch": 0.47, "learning_rate": 4.219021925483118e-05, "loss": 0.005, "step": 24150 }, { "epoch": 0.47, "learning_rate": 4.2189572294573946e-05, "loss": 0.0044, "step": 24152 }, { "epoch": 0.47, "learning_rate": 4.2188925334316715e-05, "loss": 0.0206, "step": 24154 }, { "epoch": 0.47, "learning_rate": 4.2188278374059484e-05, "loss": 0.0032, "step": 24156 }, { "epoch": 0.47, "learning_rate": 4.218763141380225e-05, "loss": 0.0002, "step": 24158 }, { "epoch": 0.47, "learning_rate": 4.218698445354502e-05, "loss": 0.0027, "step": 24160 }, { "epoch": 0.47, "learning_rate": 4.218633749328779e-05, "loss": 0.0003, "step": 24162 }, { "epoch": 0.47, "learning_rate": 4.218569053303056e-05, "loss": 0.0009, "step": 24164 }, { "epoch": 0.47, "learning_rate": 4.218504357277332e-05, "loss": 0.0109, "step": 24166 }, { "epoch": 0.47, "learning_rate": 4.21843966125161e-05, "loss": 0.006, "step": 24168 }, { "epoch": 0.47, "learning_rate": 4.218374965225886e-05, "loss": 0.0001, "step": 24170 }, { "epoch": 0.47, "learning_rate": 4.218310269200163e-05, "loss": 0.0038, "step": 24172 }, { "epoch": 0.47, "learning_rate": 4.2182455731744406e-05, "loss": 0.0125, "step": 24174 }, { "epoch": 0.47, "learning_rate": 4.218180877148717e-05, "loss": 0.0001, "step": 24176 }, { "epoch": 0.47, "learning_rate": 4.218116181122994e-05, "loss": 0.0004, "step": 24178 }, { "epoch": 0.47, "learning_rate": 4.2180514850972706e-05, "loss": 0.0204, "step": 24180 }, { "epoch": 0.47, "learning_rate": 4.2179867890715475e-05, "loss": 0.0026, "step": 24182 }, { "epoch": 0.47, "learning_rate": 4.2179220930458244e-05, "loss": 0.0001, "step": 24184 }, { "epoch": 0.47, "learning_rate": 4.2178573970201014e-05, "loss": 0.0001, "step": 24186 }, { "epoch": 0.47, "learning_rate": 4.217792700994378e-05, "loss": 0.0005, "step": 24188 }, { "epoch": 0.47, "learning_rate": 4.217728004968655e-05, "loss": 0.0001, "step": 24190 }, { "epoch": 0.47, "learning_rate": 4.217663308942932e-05, "loss": 0.0095, "step": 24192 }, { "epoch": 0.47, "learning_rate": 4.217598612917208e-05, "loss": 0.0021, "step": 24194 }, { "epoch": 0.47, "learning_rate": 4.217533916891486e-05, "loss": 0.0024, "step": 24196 }, { "epoch": 0.47, "learning_rate": 4.217469220865762e-05, "loss": 0.0001, "step": 24198 }, { "epoch": 0.47, "learning_rate": 4.217404524840039e-05, "loss": 0.0001, "step": 24200 }, { "epoch": 0.47, "learning_rate": 4.217339828814316e-05, "loss": 0.0, "step": 24202 }, { "epoch": 0.47, "learning_rate": 4.217275132788593e-05, "loss": 0.0001, "step": 24204 }, { "epoch": 0.47, "learning_rate": 4.21721043676287e-05, "loss": 0.0086, "step": 24206 }, { "epoch": 0.47, "learning_rate": 4.217145740737147e-05, "loss": 0.0067, "step": 24208 }, { "epoch": 0.47, "learning_rate": 4.2170810447114236e-05, "loss": 0.0005, "step": 24210 }, { "epoch": 0.47, "learning_rate": 4.2170163486857005e-05, "loss": 0.0002, "step": 24212 }, { "epoch": 0.47, "learning_rate": 4.2169516526599774e-05, "loss": 0.001, "step": 24214 }, { "epoch": 0.47, "learning_rate": 4.2168869566342536e-05, "loss": 0.0166, "step": 24216 }, { "epoch": 0.47, "learning_rate": 4.216822260608531e-05, "loss": 0.0011, "step": 24218 }, { "epoch": 0.47, "learning_rate": 4.216757564582808e-05, "loss": 0.0002, "step": 24220 }, { "epoch": 0.47, "learning_rate": 4.216692868557085e-05, "loss": 0.0001, "step": 24222 }, { "epoch": 0.47, "learning_rate": 4.216628172531362e-05, "loss": 0.0001, "step": 24224 }, { "epoch": 0.47, "learning_rate": 4.216563476505638e-05, "loss": 0.0004, "step": 24226 }, { "epoch": 0.47, "learning_rate": 4.216498780479916e-05, "loss": 0.0001, "step": 24228 }, { "epoch": 0.47, "learning_rate": 4.216434084454192e-05, "loss": 0.0045, "step": 24230 }, { "epoch": 0.47, "learning_rate": 4.216369388428469e-05, "loss": 0.0002, "step": 24232 }, { "epoch": 0.47, "learning_rate": 4.216304692402746e-05, "loss": 0.0058, "step": 24234 }, { "epoch": 0.47, "learning_rate": 4.216239996377023e-05, "loss": 0.0, "step": 24236 }, { "epoch": 0.47, "learning_rate": 4.2161753003512996e-05, "loss": 0.0001, "step": 24238 }, { "epoch": 0.47, "learning_rate": 4.2161106043255765e-05, "loss": 0.0004, "step": 24240 }, { "epoch": 0.47, "learning_rate": 4.2160459082998534e-05, "loss": 0.0001, "step": 24242 }, { "epoch": 0.47, "learning_rate": 4.2159812122741303e-05, "loss": 0.0002, "step": 24244 }, { "epoch": 0.47, "learning_rate": 4.215916516248407e-05, "loss": 0.0202, "step": 24246 }, { "epoch": 0.47, "learning_rate": 4.2158518202226835e-05, "loss": 0.0058, "step": 24248 }, { "epoch": 0.47, "learning_rate": 4.215787124196961e-05, "loss": 0.0001, "step": 24250 }, { "epoch": 0.47, "learning_rate": 4.215722428171238e-05, "loss": 0.0004, "step": 24252 }, { "epoch": 0.47, "learning_rate": 4.215657732145514e-05, "loss": 0.0, "step": 24254 }, { "epoch": 0.47, "learning_rate": 4.215593036119792e-05, "loss": 0.0019, "step": 24256 }, { "epoch": 0.47, "learning_rate": 4.215528340094068e-05, "loss": 0.0, "step": 24258 }, { "epoch": 0.47, "learning_rate": 4.215463644068345e-05, "loss": 0.0008, "step": 24260 }, { "epoch": 0.47, "learning_rate": 4.215398948042622e-05, "loss": 0.0034, "step": 24262 }, { "epoch": 0.47, "learning_rate": 4.215334252016899e-05, "loss": 0.0, "step": 24264 }, { "epoch": 0.47, "learning_rate": 4.2152695559911757e-05, "loss": 0.0183, "step": 24266 }, { "epoch": 0.47, "learning_rate": 4.2152048599654526e-05, "loss": 0.0093, "step": 24268 }, { "epoch": 0.47, "learning_rate": 4.2151401639397295e-05, "loss": 0.0036, "step": 24270 }, { "epoch": 0.47, "learning_rate": 4.2150754679140064e-05, "loss": 0.0224, "step": 24272 }, { "epoch": 0.47, "learning_rate": 4.215010771888283e-05, "loss": 0.0001, "step": 24274 }, { "epoch": 0.47, "learning_rate": 4.2149460758625595e-05, "loss": 0.0001, "step": 24276 }, { "epoch": 0.47, "learning_rate": 4.214881379836837e-05, "loss": 0.0001, "step": 24278 }, { "epoch": 0.47, "learning_rate": 4.214816683811113e-05, "loss": 0.0062, "step": 24280 }, { "epoch": 0.47, "learning_rate": 4.214751987785391e-05, "loss": 0.0016, "step": 24282 }, { "epoch": 0.47, "learning_rate": 4.214687291759667e-05, "loss": 0.0005, "step": 24284 }, { "epoch": 0.47, "learning_rate": 4.214622595733944e-05, "loss": 0.0011, "step": 24286 }, { "epoch": 0.47, "learning_rate": 4.2145578997082216e-05, "loss": 0.0027, "step": 24288 }, { "epoch": 0.47, "learning_rate": 4.214493203682498e-05, "loss": 0.0033, "step": 24290 }, { "epoch": 0.47, "learning_rate": 4.214428507656775e-05, "loss": 0.0006, "step": 24292 }, { "epoch": 0.47, "learning_rate": 4.214363811631052e-05, "loss": 0.0001, "step": 24294 }, { "epoch": 0.47, "learning_rate": 4.2142991156053286e-05, "loss": 0.0026, "step": 24296 }, { "epoch": 0.47, "learning_rate": 4.2142344195796055e-05, "loss": 0.0001, "step": 24298 }, { "epoch": 0.47, "learning_rate": 4.2141697235538824e-05, "loss": 0.0067, "step": 24300 }, { "epoch": 0.47, "learning_rate": 4.214105027528159e-05, "loss": 0.0039, "step": 24302 }, { "epoch": 0.47, "learning_rate": 4.214040331502436e-05, "loss": 0.0001, "step": 24304 }, { "epoch": 0.47, "learning_rate": 4.213975635476713e-05, "loss": 0.0019, "step": 24306 }, { "epoch": 0.47, "learning_rate": 4.2139109394509894e-05, "loss": 0.0003, "step": 24308 }, { "epoch": 0.47, "learning_rate": 4.213846243425267e-05, "loss": 0.0, "step": 24310 }, { "epoch": 0.47, "learning_rate": 4.213781547399543e-05, "loss": 0.0116, "step": 24312 }, { "epoch": 0.47, "learning_rate": 4.21371685137382e-05, "loss": 0.0029, "step": 24314 }, { "epoch": 0.47, "learning_rate": 4.213652155348097e-05, "loss": 0.0001, "step": 24316 }, { "epoch": 0.47, "learning_rate": 4.213587459322374e-05, "loss": 0.0001, "step": 24318 }, { "epoch": 0.47, "learning_rate": 4.213522763296651e-05, "loss": 0.0008, "step": 24320 }, { "epoch": 0.47, "learning_rate": 4.213458067270928e-05, "loss": 0.0022, "step": 24322 }, { "epoch": 0.47, "learning_rate": 4.2133933712452046e-05, "loss": 0.0091, "step": 24324 }, { "epoch": 0.47, "learning_rate": 4.2133286752194816e-05, "loss": 0.0126, "step": 24326 }, { "epoch": 0.47, "learning_rate": 4.2132639791937585e-05, "loss": 0.0, "step": 24328 }, { "epoch": 0.47, "learning_rate": 4.213199283168035e-05, "loss": 0.0026, "step": 24330 }, { "epoch": 0.47, "learning_rate": 4.213134587142312e-05, "loss": 0.0012, "step": 24332 }, { "epoch": 0.47, "learning_rate": 4.213069891116589e-05, "loss": 0.0079, "step": 24334 }, { "epoch": 0.47, "learning_rate": 4.2130051950908654e-05, "loss": 0.001, "step": 24336 }, { "epoch": 0.47, "learning_rate": 4.212940499065143e-05, "loss": 0.0, "step": 24338 }, { "epoch": 0.47, "learning_rate": 4.212875803039419e-05, "loss": 0.0001, "step": 24340 }, { "epoch": 0.47, "learning_rate": 4.212811107013696e-05, "loss": 0.0062, "step": 24342 }, { "epoch": 0.47, "learning_rate": 4.212746410987973e-05, "loss": 0.0018, "step": 24344 }, { "epoch": 0.47, "learning_rate": 4.21268171496225e-05, "loss": 0.0, "step": 24346 }, { "epoch": 0.47, "learning_rate": 4.212617018936527e-05, "loss": 0.0002, "step": 24348 }, { "epoch": 0.47, "learning_rate": 4.212552322910804e-05, "loss": 0.0001, "step": 24350 }, { "epoch": 0.47, "learning_rate": 4.212487626885081e-05, "loss": 0.0043, "step": 24352 }, { "epoch": 0.47, "learning_rate": 4.2124229308593576e-05, "loss": 0.0019, "step": 24354 }, { "epoch": 0.47, "learning_rate": 4.2123582348336345e-05, "loss": 0.0086, "step": 24356 }, { "epoch": 0.47, "learning_rate": 4.212293538807911e-05, "loss": 0.0, "step": 24358 }, { "epoch": 0.47, "learning_rate": 4.212228842782188e-05, "loss": 0.0001, "step": 24360 }, { "epoch": 0.47, "learning_rate": 4.2121641467564646e-05, "loss": 0.0054, "step": 24362 }, { "epoch": 0.47, "learning_rate": 4.212099450730742e-05, "loss": 0.0049, "step": 24364 }, { "epoch": 0.47, "learning_rate": 4.212034754705019e-05, "loss": 0.0005, "step": 24366 }, { "epoch": 0.47, "learning_rate": 4.211970058679295e-05, "loss": 0.0001, "step": 24368 }, { "epoch": 0.47, "learning_rate": 4.211905362653573e-05, "loss": 0.0024, "step": 24370 }, { "epoch": 0.47, "learning_rate": 4.211840666627849e-05, "loss": 0.0004, "step": 24372 }, { "epoch": 0.47, "learning_rate": 4.211775970602126e-05, "loss": 0.0002, "step": 24374 }, { "epoch": 0.47, "learning_rate": 4.211711274576403e-05, "loss": 0.0001, "step": 24376 }, { "epoch": 0.47, "learning_rate": 4.21164657855068e-05, "loss": 0.0007, "step": 24378 }, { "epoch": 0.47, "learning_rate": 4.211581882524957e-05, "loss": 0.0013, "step": 24380 }, { "epoch": 0.47, "learning_rate": 4.2115171864992336e-05, "loss": 0.006, "step": 24382 }, { "epoch": 0.47, "learning_rate": 4.2114524904735105e-05, "loss": 0.0001, "step": 24384 }, { "epoch": 0.47, "learning_rate": 4.2113877944477875e-05, "loss": 0.0001, "step": 24386 }, { "epoch": 0.47, "learning_rate": 4.2113230984220644e-05, "loss": 0.0001, "step": 24388 }, { "epoch": 0.47, "learning_rate": 4.2112584023963406e-05, "loss": 0.0002, "step": 24390 }, { "epoch": 0.47, "learning_rate": 4.211193706370618e-05, "loss": 0.0097, "step": 24392 }, { "epoch": 0.47, "learning_rate": 4.2111290103448944e-05, "loss": 0.013, "step": 24394 }, { "epoch": 0.47, "learning_rate": 4.211064314319171e-05, "loss": 0.0001, "step": 24396 }, { "epoch": 0.47, "learning_rate": 4.210999618293449e-05, "loss": 0.0001, "step": 24398 }, { "epoch": 0.47, "learning_rate": 4.210934922267725e-05, "loss": 0.0034, "step": 24400 }, { "epoch": 0.47, "learning_rate": 4.210870226242002e-05, "loss": 0.0106, "step": 24402 }, { "epoch": 0.47, "learning_rate": 4.210805530216279e-05, "loss": 0.0019, "step": 24404 }, { "epoch": 0.47, "learning_rate": 4.210740834190556e-05, "loss": 0.0001, "step": 24406 }, { "epoch": 0.47, "learning_rate": 4.210676138164833e-05, "loss": 0.007, "step": 24408 }, { "epoch": 0.47, "learning_rate": 4.21061144213911e-05, "loss": 0.0095, "step": 24410 }, { "epoch": 0.47, "learning_rate": 4.2105467461133866e-05, "loss": 0.0005, "step": 24412 }, { "epoch": 0.47, "learning_rate": 4.2104820500876635e-05, "loss": 0.0001, "step": 24414 }, { "epoch": 0.47, "learning_rate": 4.2104173540619404e-05, "loss": 0.0008, "step": 24416 }, { "epoch": 0.47, "learning_rate": 4.2103526580362166e-05, "loss": 0.0027, "step": 24418 }, { "epoch": 0.47, "learning_rate": 4.210287962010494e-05, "loss": 0.0003, "step": 24420 }, { "epoch": 0.47, "learning_rate": 4.2102232659847704e-05, "loss": 0.0001, "step": 24422 }, { "epoch": 0.47, "learning_rate": 4.210158569959048e-05, "loss": 0.0, "step": 24424 }, { "epoch": 0.47, "learning_rate": 4.210093873933324e-05, "loss": 0.0008, "step": 24426 }, { "epoch": 0.47, "learning_rate": 4.210029177907601e-05, "loss": 0.0, "step": 24428 }, { "epoch": 0.47, "learning_rate": 4.209964481881878e-05, "loss": 0.0001, "step": 24430 }, { "epoch": 0.47, "learning_rate": 4.209899785856155e-05, "loss": 0.0001, "step": 24432 }, { "epoch": 0.47, "learning_rate": 4.209835089830432e-05, "loss": 0.0023, "step": 24434 }, { "epoch": 0.47, "learning_rate": 4.209770393804709e-05, "loss": 0.0007, "step": 24436 }, { "epoch": 0.47, "learning_rate": 4.209705697778986e-05, "loss": 0.0004, "step": 24438 }, { "epoch": 0.47, "learning_rate": 4.209641001753262e-05, "loss": 0.0027, "step": 24440 }, { "epoch": 0.47, "learning_rate": 4.2095763057275395e-05, "loss": 0.0, "step": 24442 }, { "epoch": 0.47, "learning_rate": 4.2095116097018164e-05, "loss": 0.0001, "step": 24444 }, { "epoch": 0.47, "learning_rate": 4.2094469136760933e-05, "loss": 0.0, "step": 24446 }, { "epoch": 0.47, "learning_rate": 4.20938221765037e-05, "loss": 0.0002, "step": 24448 }, { "epoch": 0.47, "learning_rate": 4.2093175216246465e-05, "loss": 0.0034, "step": 24450 }, { "epoch": 0.47, "learning_rate": 4.209252825598924e-05, "loss": 0.0011, "step": 24452 }, { "epoch": 0.47, "learning_rate": 4.2091881295732e-05, "loss": 0.0001, "step": 24454 }, { "epoch": 0.47, "learning_rate": 4.209123433547477e-05, "loss": 0.0009, "step": 24456 }, { "epoch": 0.47, "learning_rate": 4.209058737521754e-05, "loss": 0.0064, "step": 24458 }, { "epoch": 0.47, "learning_rate": 4.208994041496031e-05, "loss": 0.0002, "step": 24460 }, { "epoch": 0.47, "learning_rate": 4.208929345470308e-05, "loss": 0.0081, "step": 24462 }, { "epoch": 0.47, "learning_rate": 4.208864649444585e-05, "loss": 0.0032, "step": 24464 }, { "epoch": 0.47, "learning_rate": 4.208799953418862e-05, "loss": 0.0006, "step": 24466 }, { "epoch": 0.47, "learning_rate": 4.208735257393139e-05, "loss": 0.0123, "step": 24468 }, { "epoch": 0.47, "learning_rate": 4.2086705613674156e-05, "loss": 0.0091, "step": 24470 }, { "epoch": 0.47, "learning_rate": 4.208605865341692e-05, "loss": 0.0, "step": 24472 }, { "epoch": 0.48, "learning_rate": 4.2085411693159694e-05, "loss": 0.0001, "step": 24474 }, { "epoch": 0.48, "learning_rate": 4.208476473290246e-05, "loss": 0.0001, "step": 24476 }, { "epoch": 0.48, "learning_rate": 4.2084117772645225e-05, "loss": 0.0008, "step": 24478 }, { "epoch": 0.48, "learning_rate": 4.2083470812388e-05, "loss": 0.0016, "step": 24480 }, { "epoch": 0.48, "learning_rate": 4.2082823852130763e-05, "loss": 0.0157, "step": 24482 }, { "epoch": 0.48, "learning_rate": 4.208217689187353e-05, "loss": 0.0001, "step": 24484 }, { "epoch": 0.48, "learning_rate": 4.20815299316163e-05, "loss": 0.0, "step": 24486 }, { "epoch": 0.48, "learning_rate": 4.208088297135907e-05, "loss": 0.0001, "step": 24488 }, { "epoch": 0.48, "learning_rate": 4.208023601110184e-05, "loss": 0.0007, "step": 24490 }, { "epoch": 0.48, "learning_rate": 4.207958905084461e-05, "loss": 0.0088, "step": 24492 }, { "epoch": 0.48, "learning_rate": 4.207894209058738e-05, "loss": 0.0, "step": 24494 }, { "epoch": 0.48, "learning_rate": 4.207829513033015e-05, "loss": 0.0006, "step": 24496 }, { "epoch": 0.48, "learning_rate": 4.2077648170072916e-05, "loss": 0.0068, "step": 24498 }, { "epoch": 0.48, "learning_rate": 4.207700120981568e-05, "loss": 0.0007, "step": 24500 }, { "epoch": 0.48, "learning_rate": 4.2076354249558454e-05, "loss": 0.0003, "step": 24502 }, { "epoch": 0.48, "learning_rate": 4.207570728930122e-05, "loss": 0.0, "step": 24504 }, { "epoch": 0.48, "learning_rate": 4.207506032904399e-05, "loss": 0.0002, "step": 24506 }, { "epoch": 0.48, "learning_rate": 4.2074413368786755e-05, "loss": 0.0028, "step": 24508 }, { "epoch": 0.48, "learning_rate": 4.2073766408529524e-05, "loss": 0.0003, "step": 24510 }, { "epoch": 0.48, "learning_rate": 4.20731194482723e-05, "loss": 0.0031, "step": 24512 }, { "epoch": 0.48, "learning_rate": 4.207247248801506e-05, "loss": 0.0001, "step": 24514 }, { "epoch": 0.48, "learning_rate": 4.207182552775783e-05, "loss": 0.0028, "step": 24516 }, { "epoch": 0.48, "learning_rate": 4.20711785675006e-05, "loss": 0.0003, "step": 24518 }, { "epoch": 0.48, "learning_rate": 4.207053160724337e-05, "loss": 0.0001, "step": 24520 }, { "epoch": 0.48, "learning_rate": 4.206988464698614e-05, "loss": 0.0045, "step": 24522 }, { "epoch": 0.48, "learning_rate": 4.206923768672891e-05, "loss": 0.0031, "step": 24524 }, { "epoch": 0.48, "learning_rate": 4.2068590726471677e-05, "loss": 0.0002, "step": 24526 }, { "epoch": 0.48, "learning_rate": 4.2067943766214446e-05, "loss": 0.0, "step": 24528 }, { "epoch": 0.48, "learning_rate": 4.2067296805957215e-05, "loss": 0.0001, "step": 24530 }, { "epoch": 0.48, "learning_rate": 4.206664984569998e-05, "loss": 0.0, "step": 24532 }, { "epoch": 0.48, "learning_rate": 4.206600288544275e-05, "loss": 0.0, "step": 24534 }, { "epoch": 0.48, "learning_rate": 4.2065355925185515e-05, "loss": 0.0001, "step": 24536 }, { "epoch": 0.48, "learning_rate": 4.2064708964928284e-05, "loss": 0.0, "step": 24538 }, { "epoch": 0.48, "learning_rate": 4.206406200467105e-05, "loss": 0.0012, "step": 24540 }, { "epoch": 0.48, "learning_rate": 4.206341504441382e-05, "loss": 0.0011, "step": 24542 }, { "epoch": 0.48, "learning_rate": 4.206276808415659e-05, "loss": 0.0001, "step": 24544 }, { "epoch": 0.48, "learning_rate": 4.206212112389936e-05, "loss": 0.0003, "step": 24546 }, { "epoch": 0.48, "learning_rate": 4.206147416364213e-05, "loss": 0.0001, "step": 24548 }, { "epoch": 0.48, "learning_rate": 4.20608272033849e-05, "loss": 0.0077, "step": 24550 }, { "epoch": 0.48, "learning_rate": 4.206018024312767e-05, "loss": 0.0007, "step": 24552 }, { "epoch": 0.48, "learning_rate": 4.205953328287044e-05, "loss": 0.0001, "step": 24554 }, { "epoch": 0.48, "learning_rate": 4.2058886322613206e-05, "loss": 0.0001, "step": 24556 }, { "epoch": 0.48, "learning_rate": 4.2058239362355975e-05, "loss": 0.0, "step": 24558 }, { "epoch": 0.48, "learning_rate": 4.205759240209874e-05, "loss": 0.0001, "step": 24560 }, { "epoch": 0.48, "learning_rate": 4.205694544184151e-05, "loss": 0.0039, "step": 24562 }, { "epoch": 0.48, "learning_rate": 4.2056298481584276e-05, "loss": 0.0009, "step": 24564 }, { "epoch": 0.48, "learning_rate": 4.205565152132705e-05, "loss": 0.0182, "step": 24566 }, { "epoch": 0.48, "learning_rate": 4.2055004561069814e-05, "loss": 0.0001, "step": 24568 }, { "epoch": 0.48, "learning_rate": 4.205435760081258e-05, "loss": 0.0095, "step": 24570 }, { "epoch": 0.48, "learning_rate": 4.205371064055535e-05, "loss": 0.0018, "step": 24572 }, { "epoch": 0.48, "learning_rate": 4.205306368029812e-05, "loss": 0.0, "step": 24574 }, { "epoch": 0.48, "learning_rate": 4.205241672004089e-05, "loss": 0.0114, "step": 24576 }, { "epoch": 0.48, "learning_rate": 4.205176975978366e-05, "loss": 0.0048, "step": 24578 }, { "epoch": 0.48, "learning_rate": 4.205112279952643e-05, "loss": 0.0001, "step": 24580 }, { "epoch": 0.48, "learning_rate": 4.205047583926919e-05, "loss": 0.0083, "step": 24582 }, { "epoch": 0.48, "learning_rate": 4.2049828879011966e-05, "loss": 0.0001, "step": 24584 }, { "epoch": 0.48, "learning_rate": 4.204918191875473e-05, "loss": 0.0022, "step": 24586 }, { "epoch": 0.48, "learning_rate": 4.2048534958497505e-05, "loss": 0.0002, "step": 24588 }, { "epoch": 0.48, "learning_rate": 4.2047887998240274e-05, "loss": 0.0, "step": 24590 }, { "epoch": 0.48, "learning_rate": 4.2047241037983036e-05, "loss": 0.0019, "step": 24592 }, { "epoch": 0.48, "learning_rate": 4.204659407772581e-05, "loss": 0.0131, "step": 24594 }, { "epoch": 0.48, "learning_rate": 4.2045947117468574e-05, "loss": 0.0036, "step": 24596 }, { "epoch": 0.48, "learning_rate": 4.204530015721134e-05, "loss": 0.0123, "step": 24598 }, { "epoch": 0.48, "learning_rate": 4.204465319695411e-05, "loss": 0.0074, "step": 24600 }, { "epoch": 0.48, "learning_rate": 4.204400623669688e-05, "loss": 0.0002, "step": 24602 }, { "epoch": 0.48, "learning_rate": 4.204335927643965e-05, "loss": 0.0088, "step": 24604 }, { "epoch": 0.48, "learning_rate": 4.204271231618242e-05, "loss": 0.0001, "step": 24606 }, { "epoch": 0.48, "learning_rate": 4.204206535592519e-05, "loss": 0.0013, "step": 24608 }, { "epoch": 0.48, "learning_rate": 4.204141839566796e-05, "loss": 0.0, "step": 24610 }, { "epoch": 0.48, "learning_rate": 4.204077143541073e-05, "loss": 0.0004, "step": 24612 }, { "epoch": 0.48, "learning_rate": 4.204012447515349e-05, "loss": 0.0026, "step": 24614 }, { "epoch": 0.48, "learning_rate": 4.2039477514896265e-05, "loss": 0.0001, "step": 24616 }, { "epoch": 0.48, "learning_rate": 4.203883055463903e-05, "loss": 0.0058, "step": 24618 }, { "epoch": 0.48, "learning_rate": 4.2038183594381796e-05, "loss": 0.0069, "step": 24620 }, { "epoch": 0.48, "learning_rate": 4.203753663412457e-05, "loss": 0.0, "step": 24622 }, { "epoch": 0.48, "learning_rate": 4.2036889673867335e-05, "loss": 0.0004, "step": 24624 }, { "epoch": 0.48, "learning_rate": 4.2036242713610104e-05, "loss": 0.0001, "step": 24626 }, { "epoch": 0.48, "learning_rate": 4.203559575335287e-05, "loss": 0.0034, "step": 24628 }, { "epoch": 0.48, "learning_rate": 4.203494879309564e-05, "loss": 0.0035, "step": 24630 }, { "epoch": 0.48, "learning_rate": 4.203430183283841e-05, "loss": 0.0293, "step": 24632 }, { "epoch": 0.48, "learning_rate": 4.203365487258118e-05, "loss": 0.0004, "step": 24634 }, { "epoch": 0.48, "learning_rate": 4.203300791232395e-05, "loss": 0.0, "step": 24636 }, { "epoch": 0.48, "learning_rate": 4.203236095206672e-05, "loss": 0.0013, "step": 24638 }, { "epoch": 0.48, "learning_rate": 4.203171399180949e-05, "loss": 0.0001, "step": 24640 }, { "epoch": 0.48, "learning_rate": 4.203106703155225e-05, "loss": 0.0002, "step": 24642 }, { "epoch": 0.48, "learning_rate": 4.2030420071295025e-05, "loss": 0.0, "step": 24644 }, { "epoch": 0.48, "learning_rate": 4.202977311103779e-05, "loss": 0.0002, "step": 24646 }, { "epoch": 0.48, "learning_rate": 4.2029126150780564e-05, "loss": 0.0, "step": 24648 }, { "epoch": 0.48, "learning_rate": 4.2028479190523326e-05, "loss": 0.0007, "step": 24650 }, { "epoch": 0.48, "learning_rate": 4.2027832230266095e-05, "loss": 0.008, "step": 24652 }, { "epoch": 0.48, "learning_rate": 4.202718527000887e-05, "loss": 0.0002, "step": 24654 }, { "epoch": 0.48, "learning_rate": 4.202653830975163e-05, "loss": 0.0004, "step": 24656 }, { "epoch": 0.48, "learning_rate": 4.20258913494944e-05, "loss": 0.0004, "step": 24658 }, { "epoch": 0.48, "learning_rate": 4.202524438923717e-05, "loss": 0.0001, "step": 24660 }, { "epoch": 0.48, "learning_rate": 4.202459742897994e-05, "loss": 0.0005, "step": 24662 }, { "epoch": 0.48, "learning_rate": 4.20239504687227e-05, "loss": 0.0001, "step": 24664 }, { "epoch": 0.48, "learning_rate": 4.202330350846548e-05, "loss": 0.0028, "step": 24666 }, { "epoch": 0.48, "learning_rate": 4.202265654820825e-05, "loss": 0.0001, "step": 24668 }, { "epoch": 0.48, "learning_rate": 4.202200958795102e-05, "loss": 0.0006, "step": 24670 }, { "epoch": 0.48, "learning_rate": 4.2021362627693786e-05, "loss": 0.0, "step": 24672 }, { "epoch": 0.48, "learning_rate": 4.202071566743655e-05, "loss": 0.0011, "step": 24674 }, { "epoch": 0.48, "learning_rate": 4.2020068707179324e-05, "loss": 0.0001, "step": 24676 }, { "epoch": 0.48, "learning_rate": 4.2019421746922086e-05, "loss": 0.0131, "step": 24678 }, { "epoch": 0.48, "learning_rate": 4.2018774786664855e-05, "loss": 0.0001, "step": 24680 }, { "epoch": 0.48, "learning_rate": 4.2018127826407624e-05, "loss": 0.0, "step": 24682 }, { "epoch": 0.48, "learning_rate": 4.2017480866150394e-05, "loss": 0.002, "step": 24684 }, { "epoch": 0.48, "learning_rate": 4.201683390589316e-05, "loss": 0.0103, "step": 24686 }, { "epoch": 0.48, "learning_rate": 4.201618694563593e-05, "loss": 0.033, "step": 24688 }, { "epoch": 0.48, "learning_rate": 4.20155399853787e-05, "loss": 0.0002, "step": 24690 }, { "epoch": 0.48, "learning_rate": 4.201489302512147e-05, "loss": 0.0032, "step": 24692 }, { "epoch": 0.48, "learning_rate": 4.201424606486424e-05, "loss": 0.0054, "step": 24694 }, { "epoch": 0.48, "learning_rate": 4.2013599104607e-05, "loss": 0.0004, "step": 24696 }, { "epoch": 0.48, "learning_rate": 4.201295214434978e-05, "loss": 0.0115, "step": 24698 }, { "epoch": 0.48, "learning_rate": 4.2012305184092546e-05, "loss": 0.0001, "step": 24700 }, { "epoch": 0.48, "learning_rate": 4.201165822383531e-05, "loss": 0.002, "step": 24702 }, { "epoch": 0.48, "learning_rate": 4.2011011263578084e-05, "loss": 0.0039, "step": 24704 }, { "epoch": 0.48, "learning_rate": 4.201036430332085e-05, "loss": 0.0015, "step": 24706 }, { "epoch": 0.48, "learning_rate": 4.2009717343063616e-05, "loss": 0.0, "step": 24708 }, { "epoch": 0.48, "learning_rate": 4.2009070382806385e-05, "loss": 0.0001, "step": 24710 }, { "epoch": 0.48, "learning_rate": 4.2008423422549154e-05, "loss": 0.0012, "step": 24712 }, { "epoch": 0.48, "learning_rate": 4.200777646229192e-05, "loss": 0.0017, "step": 24714 }, { "epoch": 0.48, "learning_rate": 4.200712950203469e-05, "loss": 0.008, "step": 24716 }, { "epoch": 0.48, "learning_rate": 4.200648254177746e-05, "loss": 0.0001, "step": 24718 }, { "epoch": 0.48, "learning_rate": 4.200583558152023e-05, "loss": 0.0518, "step": 24720 }, { "epoch": 0.48, "learning_rate": 4.2005188621263e-05, "loss": 0.0001, "step": 24722 }, { "epoch": 0.48, "learning_rate": 4.200454166100576e-05, "loss": 0.0045, "step": 24724 }, { "epoch": 0.48, "learning_rate": 4.200389470074854e-05, "loss": 0.0, "step": 24726 }, { "epoch": 0.48, "learning_rate": 4.20032477404913e-05, "loss": 0.0107, "step": 24728 }, { "epoch": 0.48, "learning_rate": 4.2002600780234076e-05, "loss": 0.0007, "step": 24730 }, { "epoch": 0.48, "learning_rate": 4.2001953819976845e-05, "loss": 0.0024, "step": 24732 }, { "epoch": 0.48, "learning_rate": 4.200130685971961e-05, "loss": 0.0001, "step": 24734 }, { "epoch": 0.48, "learning_rate": 4.200065989946238e-05, "loss": 0.0001, "step": 24736 }, { "epoch": 0.48, "learning_rate": 4.2000012939205145e-05, "loss": 0.0001, "step": 24738 }, { "epoch": 0.48, "learning_rate": 4.1999365978947914e-05, "loss": 0.007, "step": 24740 }, { "epoch": 0.48, "learning_rate": 4.1998719018690683e-05, "loss": 0.0163, "step": 24742 }, { "epoch": 0.48, "learning_rate": 4.199807205843345e-05, "loss": 0.0116, "step": 24744 }, { "epoch": 0.48, "learning_rate": 4.199742509817622e-05, "loss": 0.0004, "step": 24746 }, { "epoch": 0.48, "learning_rate": 4.199677813791899e-05, "loss": 0.0003, "step": 24748 }, { "epoch": 0.48, "learning_rate": 4.199613117766176e-05, "loss": 0.0133, "step": 24750 }, { "epoch": 0.48, "learning_rate": 4.199548421740453e-05, "loss": 0.0, "step": 24752 }, { "epoch": 0.48, "learning_rate": 4.19948372571473e-05, "loss": 0.0, "step": 24754 }, { "epoch": 0.48, "learning_rate": 4.199419029689006e-05, "loss": 0.0009, "step": 24756 }, { "epoch": 0.48, "learning_rate": 4.1993543336632836e-05, "loss": 0.0001, "step": 24758 }, { "epoch": 0.48, "learning_rate": 4.19928963763756e-05, "loss": 0.0003, "step": 24760 }, { "epoch": 0.48, "learning_rate": 4.199224941611837e-05, "loss": 0.0074, "step": 24762 }, { "epoch": 0.48, "learning_rate": 4.1991602455861137e-05, "loss": 0.0017, "step": 24764 }, { "epoch": 0.48, "learning_rate": 4.1990955495603906e-05, "loss": 0.0005, "step": 24766 }, { "epoch": 0.48, "learning_rate": 4.1990308535346675e-05, "loss": 0.0007, "step": 24768 }, { "epoch": 0.48, "learning_rate": 4.1989661575089444e-05, "loss": 0.0001, "step": 24770 }, { "epoch": 0.48, "learning_rate": 4.198901461483221e-05, "loss": 0.0004, "step": 24772 }, { "epoch": 0.48, "learning_rate": 4.198836765457498e-05, "loss": 0.0002, "step": 24774 }, { "epoch": 0.48, "learning_rate": 4.198772069431775e-05, "loss": 0.0005, "step": 24776 }, { "epoch": 0.48, "learning_rate": 4.198707373406052e-05, "loss": 0.0001, "step": 24778 }, { "epoch": 0.48, "learning_rate": 4.198642677380329e-05, "loss": 0.0001, "step": 24780 }, { "epoch": 0.48, "learning_rate": 4.198577981354606e-05, "loss": 0.0018, "step": 24782 }, { "epoch": 0.48, "learning_rate": 4.198513285328882e-05, "loss": 0.0003, "step": 24784 }, { "epoch": 0.48, "learning_rate": 4.1984485893031597e-05, "loss": 0.0001, "step": 24786 }, { "epoch": 0.48, "learning_rate": 4.198383893277436e-05, "loss": 0.0, "step": 24788 }, { "epoch": 0.48, "learning_rate": 4.1983191972517135e-05, "loss": 0.0045, "step": 24790 }, { "epoch": 0.48, "learning_rate": 4.19825450122599e-05, "loss": 0.0003, "step": 24792 }, { "epoch": 0.48, "learning_rate": 4.1981898052002666e-05, "loss": 0.0, "step": 24794 }, { "epoch": 0.48, "learning_rate": 4.1981251091745435e-05, "loss": 0.0, "step": 24796 }, { "epoch": 0.48, "learning_rate": 4.1980604131488204e-05, "loss": 0.0001, "step": 24798 }, { "epoch": 0.48, "learning_rate": 4.197995717123097e-05, "loss": 0.0004, "step": 24800 }, { "epoch": 0.48, "learning_rate": 4.197931021097374e-05, "loss": 0.0005, "step": 24802 }, { "epoch": 0.48, "learning_rate": 4.197866325071651e-05, "loss": 0.0002, "step": 24804 }, { "epoch": 0.48, "learning_rate": 4.1978016290459274e-05, "loss": 0.0001, "step": 24806 }, { "epoch": 0.48, "learning_rate": 4.197736933020205e-05, "loss": 0.0035, "step": 24808 }, { "epoch": 0.48, "learning_rate": 4.197672236994481e-05, "loss": 0.0003, "step": 24810 }, { "epoch": 0.48, "learning_rate": 4.197607540968759e-05, "loss": 0.0004, "step": 24812 }, { "epoch": 0.48, "learning_rate": 4.197542844943036e-05, "loss": 0.0001, "step": 24814 }, { "epoch": 0.48, "learning_rate": 4.197478148917312e-05, "loss": 0.0064, "step": 24816 }, { "epoch": 0.48, "learning_rate": 4.1974134528915895e-05, "loss": 0.0001, "step": 24818 }, { "epoch": 0.48, "learning_rate": 4.197348756865866e-05, "loss": 0.0569, "step": 24820 }, { "epoch": 0.48, "learning_rate": 4.1972840608401426e-05, "loss": 0.0043, "step": 24822 }, { "epoch": 0.48, "learning_rate": 4.1972193648144196e-05, "loss": 0.0, "step": 24824 }, { "epoch": 0.48, "learning_rate": 4.1971546687886965e-05, "loss": 0.0, "step": 24826 }, { "epoch": 0.48, "learning_rate": 4.1970899727629734e-05, "loss": 0.0104, "step": 24828 }, { "epoch": 0.48, "learning_rate": 4.19702527673725e-05, "loss": 0.0018, "step": 24830 }, { "epoch": 0.48, "learning_rate": 4.196960580711527e-05, "loss": 0.0003, "step": 24832 }, { "epoch": 0.48, "learning_rate": 4.196895884685804e-05, "loss": 0.0312, "step": 24834 }, { "epoch": 0.48, "learning_rate": 4.196831188660081e-05, "loss": 0.0006, "step": 24836 }, { "epoch": 0.48, "learning_rate": 4.196766492634357e-05, "loss": 0.0001, "step": 24838 }, { "epoch": 0.48, "learning_rate": 4.196701796608635e-05, "loss": 0.0008, "step": 24840 }, { "epoch": 0.48, "learning_rate": 4.196637100582911e-05, "loss": 0.0009, "step": 24842 }, { "epoch": 0.48, "learning_rate": 4.196572404557188e-05, "loss": 0.0103, "step": 24844 }, { "epoch": 0.48, "learning_rate": 4.1965077085314656e-05, "loss": 0.0015, "step": 24846 }, { "epoch": 0.48, "learning_rate": 4.196443012505742e-05, "loss": 0.0021, "step": 24848 }, { "epoch": 0.48, "learning_rate": 4.196378316480019e-05, "loss": 0.0002, "step": 24850 }, { "epoch": 0.48, "learning_rate": 4.1963136204542956e-05, "loss": 0.0, "step": 24852 }, { "epoch": 0.48, "learning_rate": 4.1962489244285725e-05, "loss": 0.0116, "step": 24854 }, { "epoch": 0.48, "learning_rate": 4.1961842284028494e-05, "loss": 0.0002, "step": 24856 }, { "epoch": 0.48, "learning_rate": 4.196119532377126e-05, "loss": 0.0008, "step": 24858 }, { "epoch": 0.48, "learning_rate": 4.196054836351403e-05, "loss": 0.0115, "step": 24860 }, { "epoch": 0.48, "learning_rate": 4.19599014032568e-05, "loss": 0.0022, "step": 24862 }, { "epoch": 0.48, "learning_rate": 4.195925444299957e-05, "loss": 0.0027, "step": 24864 }, { "epoch": 0.48, "learning_rate": 4.195860748274233e-05, "loss": 0.0, "step": 24866 }, { "epoch": 0.48, "learning_rate": 4.195796052248511e-05, "loss": 0.0, "step": 24868 }, { "epoch": 0.48, "learning_rate": 4.195731356222787e-05, "loss": 0.0004, "step": 24870 }, { "epoch": 0.48, "learning_rate": 4.195666660197065e-05, "loss": 0.0002, "step": 24872 }, { "epoch": 0.48, "learning_rate": 4.195601964171341e-05, "loss": 0.0005, "step": 24874 }, { "epoch": 0.48, "learning_rate": 4.195537268145618e-05, "loss": 0.0, "step": 24876 }, { "epoch": 0.48, "learning_rate": 4.1954725721198954e-05, "loss": 0.0001, "step": 24878 }, { "epoch": 0.48, "learning_rate": 4.1954078760941716e-05, "loss": 0.004, "step": 24880 }, { "epoch": 0.48, "learning_rate": 4.1953431800684485e-05, "loss": 0.0, "step": 24882 }, { "epoch": 0.48, "learning_rate": 4.1952784840427255e-05, "loss": 0.0, "step": 24884 }, { "epoch": 0.48, "learning_rate": 4.1952137880170024e-05, "loss": 0.0047, "step": 24886 }, { "epoch": 0.48, "learning_rate": 4.1951490919912786e-05, "loss": 0.003, "step": 24888 }, { "epoch": 0.48, "learning_rate": 4.195084395965556e-05, "loss": 0.0074, "step": 24890 }, { "epoch": 0.48, "learning_rate": 4.195019699939833e-05, "loss": 0.0035, "step": 24892 }, { "epoch": 0.48, "learning_rate": 4.19495500391411e-05, "loss": 0.004, "step": 24894 }, { "epoch": 0.48, "learning_rate": 4.194890307888387e-05, "loss": 0.0025, "step": 24896 }, { "epoch": 0.48, "learning_rate": 4.194825611862663e-05, "loss": 0.0004, "step": 24898 }, { "epoch": 0.48, "learning_rate": 4.194760915836941e-05, "loss": 0.0002, "step": 24900 }, { "epoch": 0.48, "learning_rate": 4.194696219811217e-05, "loss": 0.0007, "step": 24902 }, { "epoch": 0.48, "learning_rate": 4.194631523785494e-05, "loss": 0.0083, "step": 24904 }, { "epoch": 0.48, "learning_rate": 4.194566827759771e-05, "loss": 0.0, "step": 24906 }, { "epoch": 0.48, "learning_rate": 4.194502131734048e-05, "loss": 0.0001, "step": 24908 }, { "epoch": 0.48, "learning_rate": 4.1944374357083246e-05, "loss": 0.0076, "step": 24910 }, { "epoch": 0.48, "learning_rate": 4.1943727396826015e-05, "loss": 0.0002, "step": 24912 }, { "epoch": 0.48, "learning_rate": 4.1943080436568784e-05, "loss": 0.0001, "step": 24914 }, { "epoch": 0.48, "learning_rate": 4.194243347631155e-05, "loss": 0.0, "step": 24916 }, { "epoch": 0.48, "learning_rate": 4.194178651605432e-05, "loss": 0.0001, "step": 24918 }, { "epoch": 0.48, "learning_rate": 4.1941139555797085e-05, "loss": 0.0187, "step": 24920 }, { "epoch": 0.48, "learning_rate": 4.194049259553986e-05, "loss": 0.0084, "step": 24922 }, { "epoch": 0.48, "learning_rate": 4.193984563528263e-05, "loss": 0.0004, "step": 24924 }, { "epoch": 0.48, "learning_rate": 4.193919867502539e-05, "loss": 0.0001, "step": 24926 }, { "epoch": 0.48, "learning_rate": 4.193855171476817e-05, "loss": 0.0001, "step": 24928 }, { "epoch": 0.48, "learning_rate": 4.193790475451093e-05, "loss": 0.0003, "step": 24930 }, { "epoch": 0.48, "learning_rate": 4.1937257794253706e-05, "loss": 0.0003, "step": 24932 }, { "epoch": 0.48, "learning_rate": 4.193661083399647e-05, "loss": 0.0003, "step": 24934 }, { "epoch": 0.48, "learning_rate": 4.193596387373924e-05, "loss": 0.0329, "step": 24936 }, { "epoch": 0.48, "learning_rate": 4.1935316913482006e-05, "loss": 0.0064, "step": 24938 }, { "epoch": 0.48, "learning_rate": 4.1934669953224775e-05, "loss": 0.0016, "step": 24940 }, { "epoch": 0.48, "learning_rate": 4.1934022992967544e-05, "loss": 0.0044, "step": 24942 }, { "epoch": 0.48, "learning_rate": 4.1933376032710314e-05, "loss": 0.0, "step": 24944 }, { "epoch": 0.48, "learning_rate": 4.193272907245308e-05, "loss": 0.0059, "step": 24946 }, { "epoch": 0.48, "learning_rate": 4.1932082112195845e-05, "loss": 0.0071, "step": 24948 }, { "epoch": 0.48, "learning_rate": 4.193143515193862e-05, "loss": 0.0006, "step": 24950 }, { "epoch": 0.48, "learning_rate": 4.193078819168138e-05, "loss": 0.0, "step": 24952 }, { "epoch": 0.48, "learning_rate": 4.193014123142416e-05, "loss": 0.0002, "step": 24954 }, { "epoch": 0.48, "learning_rate": 4.192949427116693e-05, "loss": 0.005, "step": 24956 }, { "epoch": 0.48, "learning_rate": 4.192884731090969e-05, "loss": 0.0073, "step": 24958 }, { "epoch": 0.48, "learning_rate": 4.1928200350652466e-05, "loss": 0.0056, "step": 24960 }, { "epoch": 0.48, "learning_rate": 4.192755339039523e-05, "loss": 0.0062, "step": 24962 }, { "epoch": 0.48, "learning_rate": 4.1926906430138e-05, "loss": 0.0034, "step": 24964 }, { "epoch": 0.48, "learning_rate": 4.192625946988077e-05, "loss": 0.0171, "step": 24966 }, { "epoch": 0.48, "learning_rate": 4.1925612509623536e-05, "loss": 0.0007, "step": 24968 }, { "epoch": 0.48, "learning_rate": 4.1924965549366305e-05, "loss": 0.0, "step": 24970 }, { "epoch": 0.48, "learning_rate": 4.1924318589109074e-05, "loss": 0.0041, "step": 24972 }, { "epoch": 0.48, "learning_rate": 4.192367162885184e-05, "loss": 0.0002, "step": 24974 }, { "epoch": 0.48, "learning_rate": 4.192302466859461e-05, "loss": 0.0001, "step": 24976 }, { "epoch": 0.48, "learning_rate": 4.192237770833738e-05, "loss": 0.0003, "step": 24978 }, { "epoch": 0.48, "learning_rate": 4.1921730748080143e-05, "loss": 0.0001, "step": 24980 }, { "epoch": 0.48, "learning_rate": 4.192108378782292e-05, "loss": 0.0003, "step": 24982 }, { "epoch": 0.48, "learning_rate": 4.192043682756568e-05, "loss": 0.0006, "step": 24984 }, { "epoch": 0.48, "learning_rate": 4.191978986730845e-05, "loss": 0.0018, "step": 24986 }, { "epoch": 0.48, "learning_rate": 4.191914290705122e-05, "loss": 0.0011, "step": 24988 }, { "epoch": 0.49, "learning_rate": 4.191849594679399e-05, "loss": 0.0025, "step": 24990 }, { "epoch": 0.49, "learning_rate": 4.191784898653676e-05, "loss": 0.0005, "step": 24992 }, { "epoch": 0.49, "learning_rate": 4.191720202627953e-05, "loss": 0.0144, "step": 24994 }, { "epoch": 0.49, "learning_rate": 4.1916555066022296e-05, "loss": 0.0126, "step": 24996 }, { "epoch": 0.49, "learning_rate": 4.1915908105765065e-05, "loss": 0.0006, "step": 24998 }, { "epoch": 0.49, "learning_rate": 4.1915261145507834e-05, "loss": 0.0002, "step": 25000 }, { "epoch": 0.49, "learning_rate": 4.1914614185250603e-05, "loss": 0.0197, "step": 25002 }, { "epoch": 0.49, "learning_rate": 4.191396722499337e-05, "loss": 0.0, "step": 25004 }, { "epoch": 0.49, "learning_rate": 4.191332026473614e-05, "loss": 0.0025, "step": 25006 }, { "epoch": 0.49, "learning_rate": 4.1912673304478904e-05, "loss": 0.0005, "step": 25008 }, { "epoch": 0.49, "learning_rate": 4.191202634422168e-05, "loss": 0.0044, "step": 25010 }, { "epoch": 0.49, "learning_rate": 4.191137938396444e-05, "loss": 0.0373, "step": 25012 }, { "epoch": 0.49, "learning_rate": 4.191073242370722e-05, "loss": 0.0085, "step": 25014 }, { "epoch": 0.49, "learning_rate": 4.191008546344998e-05, "loss": 0.0001, "step": 25016 }, { "epoch": 0.49, "learning_rate": 4.190943850319275e-05, "loss": 0.0036, "step": 25018 }, { "epoch": 0.49, "learning_rate": 4.190879154293552e-05, "loss": 0.0073, "step": 25020 }, { "epoch": 0.49, "learning_rate": 4.190814458267829e-05, "loss": 0.0001, "step": 25022 }, { "epoch": 0.49, "learning_rate": 4.1907497622421057e-05, "loss": 0.0004, "step": 25024 }, { "epoch": 0.49, "learning_rate": 4.1906850662163826e-05, "loss": 0.0001, "step": 25026 }, { "epoch": 0.49, "learning_rate": 4.1906203701906595e-05, "loss": 0.003, "step": 25028 }, { "epoch": 0.49, "learning_rate": 4.190555674164936e-05, "loss": 0.0019, "step": 25030 }, { "epoch": 0.49, "learning_rate": 4.190490978139213e-05, "loss": 0.0009, "step": 25032 }, { "epoch": 0.49, "learning_rate": 4.19042628211349e-05, "loss": 0.0001, "step": 25034 }, { "epoch": 0.49, "learning_rate": 4.190361586087767e-05, "loss": 0.0182, "step": 25036 }, { "epoch": 0.49, "learning_rate": 4.190296890062044e-05, "loss": 0.0004, "step": 25038 }, { "epoch": 0.49, "learning_rate": 4.19023219403632e-05, "loss": 0.0003, "step": 25040 }, { "epoch": 0.49, "learning_rate": 4.190167498010598e-05, "loss": 0.0029, "step": 25042 }, { "epoch": 0.49, "learning_rate": 4.190102801984874e-05, "loss": 0.0001, "step": 25044 }, { "epoch": 0.49, "learning_rate": 4.190038105959151e-05, "loss": 0.0002, "step": 25046 }, { "epoch": 0.49, "learning_rate": 4.189973409933428e-05, "loss": 0.0125, "step": 25048 }, { "epoch": 0.49, "learning_rate": 4.189908713907705e-05, "loss": 0.0001, "step": 25050 }, { "epoch": 0.49, "learning_rate": 4.189844017881982e-05, "loss": 0.0003, "step": 25052 }, { "epoch": 0.49, "learning_rate": 4.1897793218562586e-05, "loss": 0.0002, "step": 25054 }, { "epoch": 0.49, "learning_rate": 4.1897146258305355e-05, "loss": 0.0364, "step": 25056 }, { "epoch": 0.49, "learning_rate": 4.1896499298048124e-05, "loss": 0.0001, "step": 25058 }, { "epoch": 0.49, "learning_rate": 4.189585233779089e-05, "loss": 0.0149, "step": 25060 }, { "epoch": 0.49, "learning_rate": 4.1895205377533656e-05, "loss": 0.0006, "step": 25062 }, { "epoch": 0.49, "learning_rate": 4.189455841727643e-05, "loss": 0.0016, "step": 25064 }, { "epoch": 0.49, "learning_rate": 4.1893911457019194e-05, "loss": 0.0002, "step": 25066 }, { "epoch": 0.49, "learning_rate": 4.189326449676196e-05, "loss": 0.0142, "step": 25068 }, { "epoch": 0.49, "learning_rate": 4.189261753650474e-05, "loss": 0.0005, "step": 25070 }, { "epoch": 0.49, "learning_rate": 4.18919705762475e-05, "loss": 0.0004, "step": 25072 }, { "epoch": 0.49, "learning_rate": 4.189132361599028e-05, "loss": 0.0, "step": 25074 }, { "epoch": 0.49, "learning_rate": 4.189067665573304e-05, "loss": 0.01, "step": 25076 }, { "epoch": 0.49, "learning_rate": 4.189002969547581e-05, "loss": 0.0001, "step": 25078 }, { "epoch": 0.49, "learning_rate": 4.188938273521858e-05, "loss": 0.0002, "step": 25080 }, { "epoch": 0.49, "learning_rate": 4.1888735774961346e-05, "loss": 0.0009, "step": 25082 }, { "epoch": 0.49, "learning_rate": 4.1888088814704116e-05, "loss": 0.0, "step": 25084 }, { "epoch": 0.49, "learning_rate": 4.1887441854446885e-05, "loss": 0.0, "step": 25086 }, { "epoch": 0.49, "learning_rate": 4.1886794894189654e-05, "loss": 0.0028, "step": 25088 }, { "epoch": 0.49, "learning_rate": 4.1886147933932416e-05, "loss": 0.0003, "step": 25090 }, { "epoch": 0.49, "learning_rate": 4.188550097367519e-05, "loss": 0.0001, "step": 25092 }, { "epoch": 0.49, "learning_rate": 4.1884854013417954e-05, "loss": 0.0104, "step": 25094 }, { "epoch": 0.49, "learning_rate": 4.188420705316073e-05, "loss": 0.0101, "step": 25096 }, { "epoch": 0.49, "learning_rate": 4.188356009290349e-05, "loss": 0.0002, "step": 25098 }, { "epoch": 0.49, "learning_rate": 4.188291313264626e-05, "loss": 0.0002, "step": 25100 }, { "epoch": 0.49, "learning_rate": 4.188226617238904e-05, "loss": 0.0091, "step": 25102 }, { "epoch": 0.49, "learning_rate": 4.18816192121318e-05, "loss": 0.0, "step": 25104 }, { "epoch": 0.49, "learning_rate": 4.188097225187457e-05, "loss": 0.0006, "step": 25106 }, { "epoch": 0.49, "learning_rate": 4.188032529161734e-05, "loss": 0.0091, "step": 25108 }, { "epoch": 0.49, "learning_rate": 4.187967833136011e-05, "loss": 0.0002, "step": 25110 }, { "epoch": 0.49, "learning_rate": 4.187903137110287e-05, "loss": 0.0001, "step": 25112 }, { "epoch": 0.49, "learning_rate": 4.1878384410845645e-05, "loss": 0.0096, "step": 25114 }, { "epoch": 0.49, "learning_rate": 4.1877737450588414e-05, "loss": 0.0522, "step": 25116 }, { "epoch": 0.49, "learning_rate": 4.187709049033118e-05, "loss": 0.0068, "step": 25118 }, { "epoch": 0.49, "learning_rate": 4.187644353007395e-05, "loss": 0.0001, "step": 25120 }, { "epoch": 0.49, "learning_rate": 4.1875796569816715e-05, "loss": 0.0002, "step": 25122 }, { "epoch": 0.49, "learning_rate": 4.187514960955949e-05, "loss": 0.0084, "step": 25124 }, { "epoch": 0.49, "learning_rate": 4.187450264930225e-05, "loss": 0.0, "step": 25126 }, { "epoch": 0.49, "learning_rate": 4.187385568904502e-05, "loss": 0.0282, "step": 25128 }, { "epoch": 0.49, "learning_rate": 4.187320872878779e-05, "loss": 0.0085, "step": 25130 }, { "epoch": 0.49, "learning_rate": 4.187256176853056e-05, "loss": 0.0006, "step": 25132 }, { "epoch": 0.49, "learning_rate": 4.187191480827333e-05, "loss": 0.0015, "step": 25134 }, { "epoch": 0.49, "learning_rate": 4.18712678480161e-05, "loss": 0.0005, "step": 25136 }, { "epoch": 0.49, "learning_rate": 4.187062088775887e-05, "loss": 0.0001, "step": 25138 }, { "epoch": 0.49, "learning_rate": 4.1869973927501636e-05, "loss": 0.0028, "step": 25140 }, { "epoch": 0.49, "learning_rate": 4.1869326967244405e-05, "loss": 0.0003, "step": 25142 }, { "epoch": 0.49, "learning_rate": 4.186868000698717e-05, "loss": 0.0001, "step": 25144 }, { "epoch": 0.49, "learning_rate": 4.1868033046729944e-05, "loss": 0.0004, "step": 25146 }, { "epoch": 0.49, "learning_rate": 4.186738608647271e-05, "loss": 0.0, "step": 25148 }, { "epoch": 0.49, "learning_rate": 4.1866739126215475e-05, "loss": 0.0112, "step": 25150 }, { "epoch": 0.49, "learning_rate": 4.186609216595825e-05, "loss": 0.0002, "step": 25152 }, { "epoch": 0.49, "learning_rate": 4.186544520570101e-05, "loss": 0.0022, "step": 25154 }, { "epoch": 0.49, "learning_rate": 4.186479824544379e-05, "loss": 0.0005, "step": 25156 }, { "epoch": 0.49, "learning_rate": 4.186415128518655e-05, "loss": 0.0079, "step": 25158 }, { "epoch": 0.49, "learning_rate": 4.186350432492932e-05, "loss": 0.0007, "step": 25160 }, { "epoch": 0.49, "learning_rate": 4.186285736467209e-05, "loss": 0.0061, "step": 25162 }, { "epoch": 0.49, "learning_rate": 4.186221040441486e-05, "loss": 0.0066, "step": 25164 }, { "epoch": 0.49, "learning_rate": 4.186156344415763e-05, "loss": 0.0002, "step": 25166 }, { "epoch": 0.49, "learning_rate": 4.18609164839004e-05, "loss": 0.0, "step": 25168 }, { "epoch": 0.49, "learning_rate": 4.1860269523643166e-05, "loss": 0.0002, "step": 25170 }, { "epoch": 0.49, "learning_rate": 4.185962256338593e-05, "loss": 0.0, "step": 25172 }, { "epoch": 0.49, "learning_rate": 4.1858975603128704e-05, "loss": 0.0001, "step": 25174 }, { "epoch": 0.49, "learning_rate": 4.1858328642871466e-05, "loss": 0.0002, "step": 25176 }, { "epoch": 0.49, "learning_rate": 4.185768168261424e-05, "loss": 0.0, "step": 25178 }, { "epoch": 0.49, "learning_rate": 4.185703472235701e-05, "loss": 0.001, "step": 25180 }, { "epoch": 0.49, "learning_rate": 4.1856387762099774e-05, "loss": 0.0116, "step": 25182 }, { "epoch": 0.49, "learning_rate": 4.185574080184255e-05, "loss": 0.0, "step": 25184 }, { "epoch": 0.49, "learning_rate": 4.185509384158531e-05, "loss": 0.0002, "step": 25186 }, { "epoch": 0.49, "learning_rate": 4.185444688132808e-05, "loss": 0.0003, "step": 25188 }, { "epoch": 0.49, "learning_rate": 4.185379992107085e-05, "loss": 0.0002, "step": 25190 }, { "epoch": 0.49, "learning_rate": 4.185315296081362e-05, "loss": 0.001, "step": 25192 }, { "epoch": 0.49, "learning_rate": 4.185250600055639e-05, "loss": 0.0003, "step": 25194 }, { "epoch": 0.49, "learning_rate": 4.185185904029916e-05, "loss": 0.0103, "step": 25196 }, { "epoch": 0.49, "learning_rate": 4.1851212080041926e-05, "loss": 0.0001, "step": 25198 }, { "epoch": 0.49, "learning_rate": 4.1850565119784695e-05, "loss": 0.0019, "step": 25200 }, { "epoch": 0.49, "learning_rate": 4.1849918159527464e-05, "loss": 0.0006, "step": 25202 }, { "epoch": 0.49, "learning_rate": 4.184927119927023e-05, "loss": 0.003, "step": 25204 }, { "epoch": 0.49, "learning_rate": 4.1848624239013e-05, "loss": 0.0005, "step": 25206 }, { "epoch": 0.49, "learning_rate": 4.1847977278755765e-05, "loss": 0.0001, "step": 25208 }, { "epoch": 0.49, "learning_rate": 4.1847330318498534e-05, "loss": 0.0001, "step": 25210 }, { "epoch": 0.49, "learning_rate": 4.184668335824131e-05, "loss": 0.0042, "step": 25212 }, { "epoch": 0.49, "learning_rate": 4.184603639798407e-05, "loss": 0.0006, "step": 25214 }, { "epoch": 0.49, "learning_rate": 4.184538943772685e-05, "loss": 0.0, "step": 25216 }, { "epoch": 0.49, "learning_rate": 4.184474247746961e-05, "loss": 0.0007, "step": 25218 }, { "epoch": 0.49, "learning_rate": 4.184409551721238e-05, "loss": 0.0006, "step": 25220 }, { "epoch": 0.49, "learning_rate": 4.184344855695515e-05, "loss": 0.0001, "step": 25222 }, { "epoch": 0.49, "learning_rate": 4.184280159669792e-05, "loss": 0.0002, "step": 25224 }, { "epoch": 0.49, "learning_rate": 4.184215463644069e-05, "loss": 0.0037, "step": 25226 }, { "epoch": 0.49, "learning_rate": 4.1841507676183456e-05, "loss": 0.0004, "step": 25228 }, { "epoch": 0.49, "learning_rate": 4.1840860715926225e-05, "loss": 0.0046, "step": 25230 }, { "epoch": 0.49, "learning_rate": 4.184021375566899e-05, "loss": 0.0004, "step": 25232 }, { "epoch": 0.49, "learning_rate": 4.183956679541176e-05, "loss": 0.0002, "step": 25234 }, { "epoch": 0.49, "learning_rate": 4.1838919835154525e-05, "loss": 0.0002, "step": 25236 }, { "epoch": 0.49, "learning_rate": 4.18382728748973e-05, "loss": 0.0, "step": 25238 }, { "epoch": 0.49, "learning_rate": 4.1837625914640063e-05, "loss": 0.0001, "step": 25240 }, { "epoch": 0.49, "learning_rate": 4.183697895438283e-05, "loss": 0.0003, "step": 25242 }, { "epoch": 0.49, "learning_rate": 4.18363319941256e-05, "loss": 0.0026, "step": 25244 }, { "epoch": 0.49, "learning_rate": 4.183568503386837e-05, "loss": 0.0004, "step": 25246 }, { "epoch": 0.49, "learning_rate": 4.183503807361114e-05, "loss": 0.0161, "step": 25248 }, { "epoch": 0.49, "learning_rate": 4.183439111335391e-05, "loss": 0.0054, "step": 25250 }, { "epoch": 0.49, "learning_rate": 4.183374415309668e-05, "loss": 0.0059, "step": 25252 }, { "epoch": 0.49, "learning_rate": 4.183309719283944e-05, "loss": 0.0045, "step": 25254 }, { "epoch": 0.49, "learning_rate": 4.1832450232582216e-05, "loss": 0.0001, "step": 25256 }, { "epoch": 0.49, "learning_rate": 4.1831803272324985e-05, "loss": 0.0001, "step": 25258 }, { "epoch": 0.49, "learning_rate": 4.1831156312067754e-05, "loss": 0.0002, "step": 25260 }, { "epoch": 0.49, "learning_rate": 4.1830509351810523e-05, "loss": 0.0001, "step": 25262 }, { "epoch": 0.49, "learning_rate": 4.1829862391553286e-05, "loss": 0.0001, "step": 25264 }, { "epoch": 0.49, "learning_rate": 4.182921543129606e-05, "loss": 0.0005, "step": 25266 }, { "epoch": 0.49, "learning_rate": 4.1828568471038824e-05, "loss": 0.0001, "step": 25268 }, { "epoch": 0.49, "learning_rate": 4.182792151078159e-05, "loss": 0.0027, "step": 25270 }, { "epoch": 0.49, "learning_rate": 4.182727455052436e-05, "loss": 0.0, "step": 25272 }, { "epoch": 0.49, "learning_rate": 4.182662759026713e-05, "loss": 0.0001, "step": 25274 }, { "epoch": 0.49, "learning_rate": 4.18259806300099e-05, "loss": 0.0032, "step": 25276 }, { "epoch": 0.49, "learning_rate": 4.182533366975267e-05, "loss": 0.0008, "step": 25278 }, { "epoch": 0.49, "learning_rate": 4.182468670949544e-05, "loss": 0.0001, "step": 25280 }, { "epoch": 0.49, "learning_rate": 4.182403974923821e-05, "loss": 0.0071, "step": 25282 }, { "epoch": 0.49, "learning_rate": 4.182371626910959e-05, "loss": 0.0499, "step": 25284 }, { "epoch": 0.49, "learning_rate": 4.182306930885236e-05, "loss": 0.0034, "step": 25286 }, { "epoch": 0.49, "learning_rate": 4.182242234859513e-05, "loss": 0.0018, "step": 25288 }, { "epoch": 0.49, "learning_rate": 4.18217753883379e-05, "loss": 0.0349, "step": 25290 }, { "epoch": 0.49, "learning_rate": 4.182112842808066e-05, "loss": 0.0001, "step": 25292 }, { "epoch": 0.49, "learning_rate": 4.182048146782344e-05, "loss": 0.0097, "step": 25294 }, { "epoch": 0.49, "learning_rate": 4.18198345075662e-05, "loss": 0.0017, "step": 25296 }, { "epoch": 0.49, "learning_rate": 4.181918754730897e-05, "loss": 0.0068, "step": 25298 }, { "epoch": 0.49, "learning_rate": 4.181854058705174e-05, "loss": 0.0009, "step": 25300 }, { "epoch": 0.49, "learning_rate": 4.181789362679451e-05, "loss": 0.0003, "step": 25302 }, { "epoch": 0.49, "learning_rate": 4.1817246666537276e-05, "loss": 0.0001, "step": 25304 }, { "epoch": 0.49, "learning_rate": 4.1816599706280045e-05, "loss": 0.0001, "step": 25306 }, { "epoch": 0.49, "learning_rate": 4.1815952746022814e-05, "loss": 0.0002, "step": 25308 }, { "epoch": 0.49, "learning_rate": 4.181530578576558e-05, "loss": 0.0026, "step": 25310 }, { "epoch": 0.49, "learning_rate": 4.181465882550835e-05, "loss": 0.0012, "step": 25312 }, { "epoch": 0.49, "learning_rate": 4.181401186525112e-05, "loss": 0.0017, "step": 25314 }, { "epoch": 0.49, "learning_rate": 4.181336490499389e-05, "loss": 0.0002, "step": 25316 }, { "epoch": 0.49, "learning_rate": 4.181271794473666e-05, "loss": 0.0007, "step": 25318 }, { "epoch": 0.49, "learning_rate": 4.181207098447942e-05, "loss": 0.0001, "step": 25320 }, { "epoch": 0.49, "learning_rate": 4.18114240242222e-05, "loss": 0.0007, "step": 25322 }, { "epoch": 0.49, "learning_rate": 4.181077706396496e-05, "loss": 0.0011, "step": 25324 }, { "epoch": 0.49, "learning_rate": 4.181013010370773e-05, "loss": 0.0011, "step": 25326 }, { "epoch": 0.49, "learning_rate": 4.18094831434505e-05, "loss": 0.0017, "step": 25328 }, { "epoch": 0.49, "learning_rate": 4.180883618319327e-05, "loss": 0.0002, "step": 25330 }, { "epoch": 0.49, "learning_rate": 4.1808189222936036e-05, "loss": 0.0035, "step": 25332 }, { "epoch": 0.49, "learning_rate": 4.1807542262678806e-05, "loss": 0.0001, "step": 25334 }, { "epoch": 0.49, "learning_rate": 4.1806895302421575e-05, "loss": 0.0003, "step": 25336 }, { "epoch": 0.49, "learning_rate": 4.1806248342164344e-05, "loss": 0.0, "step": 25338 }, { "epoch": 0.49, "learning_rate": 4.180560138190711e-05, "loss": 0.0002, "step": 25340 }, { "epoch": 0.49, "learning_rate": 4.1804954421649875e-05, "loss": 0.0092, "step": 25342 }, { "epoch": 0.49, "learning_rate": 4.180430746139265e-05, "loss": 0.0, "step": 25344 }, { "epoch": 0.49, "learning_rate": 4.180366050113541e-05, "loss": 0.0011, "step": 25346 }, { "epoch": 0.49, "learning_rate": 4.180301354087818e-05, "loss": 0.0031, "step": 25348 }, { "epoch": 0.49, "learning_rate": 4.180236658062096e-05, "loss": 0.0014, "step": 25350 }, { "epoch": 0.49, "learning_rate": 4.180171962036372e-05, "loss": 0.0001, "step": 25352 }, { "epoch": 0.49, "learning_rate": 4.1801072660106496e-05, "loss": 0.0001, "step": 25354 }, { "epoch": 0.49, "learning_rate": 4.180042569984926e-05, "loss": 0.0002, "step": 25356 }, { "epoch": 0.49, "learning_rate": 4.179977873959203e-05, "loss": 0.0091, "step": 25358 }, { "epoch": 0.49, "learning_rate": 4.17991317793348e-05, "loss": 0.005, "step": 25360 }, { "epoch": 0.49, "learning_rate": 4.1798484819077566e-05, "loss": 0.0499, "step": 25362 }, { "epoch": 0.49, "learning_rate": 4.1797837858820335e-05, "loss": 0.0046, "step": 25364 }, { "epoch": 0.49, "learning_rate": 4.1797190898563104e-05, "loss": 0.0005, "step": 25366 }, { "epoch": 0.49, "learning_rate": 4.179654393830587e-05, "loss": 0.0001, "step": 25368 }, { "epoch": 0.49, "learning_rate": 4.179589697804864e-05, "loss": 0.0, "step": 25370 }, { "epoch": 0.49, "learning_rate": 4.179525001779141e-05, "loss": 0.0097, "step": 25372 }, { "epoch": 0.49, "learning_rate": 4.1794603057534174e-05, "loss": 0.0002, "step": 25374 }, { "epoch": 0.49, "learning_rate": 4.179395609727695e-05, "loss": 0.0051, "step": 25376 }, { "epoch": 0.49, "learning_rate": 4.179330913701971e-05, "loss": 0.0024, "step": 25378 }, { "epoch": 0.49, "learning_rate": 4.179266217676248e-05, "loss": 0.0032, "step": 25380 }, { "epoch": 0.49, "learning_rate": 4.179201521650526e-05, "loss": 0.0051, "step": 25382 }, { "epoch": 0.49, "learning_rate": 4.179136825624802e-05, "loss": 0.0001, "step": 25384 }, { "epoch": 0.49, "learning_rate": 4.179072129599079e-05, "loss": 0.0004, "step": 25386 }, { "epoch": 0.49, "learning_rate": 4.179007433573356e-05, "loss": 0.0077, "step": 25388 }, { "epoch": 0.49, "learning_rate": 4.1789427375476326e-05, "loss": 0.0165, "step": 25390 }, { "epoch": 0.49, "learning_rate": 4.1788780415219095e-05, "loss": 0.0001, "step": 25392 }, { "epoch": 0.49, "learning_rate": 4.1788133454961865e-05, "loss": 0.0115, "step": 25394 }, { "epoch": 0.49, "learning_rate": 4.1787486494704634e-05, "loss": 0.0016, "step": 25396 }, { "epoch": 0.49, "learning_rate": 4.17868395344474e-05, "loss": 0.0003, "step": 25398 }, { "epoch": 0.49, "learning_rate": 4.178619257419017e-05, "loss": 0.0002, "step": 25400 }, { "epoch": 0.49, "learning_rate": 4.1785545613932934e-05, "loss": 0.0427, "step": 25402 }, { "epoch": 0.49, "learning_rate": 4.178489865367571e-05, "loss": 0.0008, "step": 25404 }, { "epoch": 0.49, "learning_rate": 4.178425169341847e-05, "loss": 0.0323, "step": 25406 }, { "epoch": 0.49, "learning_rate": 4.178360473316124e-05, "loss": 0.0005, "step": 25408 }, { "epoch": 0.49, "learning_rate": 4.178295777290401e-05, "loss": 0.0034, "step": 25410 }, { "epoch": 0.49, "learning_rate": 4.178231081264678e-05, "loss": 0.0066, "step": 25412 }, { "epoch": 0.49, "learning_rate": 4.1781663852389555e-05, "loss": 0.0009, "step": 25414 }, { "epoch": 0.49, "learning_rate": 4.178101689213232e-05, "loss": 0.0001, "step": 25416 }, { "epoch": 0.49, "learning_rate": 4.178036993187509e-05, "loss": 0.0028, "step": 25418 }, { "epoch": 0.49, "learning_rate": 4.1779722971617856e-05, "loss": 0.0039, "step": 25420 }, { "epoch": 0.49, "learning_rate": 4.1779076011360625e-05, "loss": 0.0001, "step": 25422 }, { "epoch": 0.49, "learning_rate": 4.177842905110339e-05, "loss": 0.0002, "step": 25424 }, { "epoch": 0.49, "learning_rate": 4.177778209084616e-05, "loss": 0.0144, "step": 25426 }, { "epoch": 0.49, "learning_rate": 4.177713513058893e-05, "loss": 0.0004, "step": 25428 }, { "epoch": 0.49, "learning_rate": 4.17764881703317e-05, "loss": 0.0038, "step": 25430 }, { "epoch": 0.49, "learning_rate": 4.177584121007447e-05, "loss": 0.0003, "step": 25432 }, { "epoch": 0.49, "learning_rate": 4.177519424981723e-05, "loss": 0.0012, "step": 25434 }, { "epoch": 0.49, "learning_rate": 4.177454728956001e-05, "loss": 0.0001, "step": 25436 }, { "epoch": 0.49, "learning_rate": 4.177390032930277e-05, "loss": 0.0001, "step": 25438 }, { "epoch": 0.49, "learning_rate": 4.177325336904554e-05, "loss": 0.0001, "step": 25440 }, { "epoch": 0.49, "learning_rate": 4.177260640878831e-05, "loss": 0.0036, "step": 25442 }, { "epoch": 0.49, "learning_rate": 4.177195944853108e-05, "loss": 0.0037, "step": 25444 }, { "epoch": 0.49, "learning_rate": 4.177131248827385e-05, "loss": 0.0002, "step": 25446 }, { "epoch": 0.49, "learning_rate": 4.1770665528016616e-05, "loss": 0.0, "step": 25448 }, { "epoch": 0.49, "learning_rate": 4.1770018567759385e-05, "loss": 0.0008, "step": 25450 }, { "epoch": 0.49, "learning_rate": 4.1769371607502154e-05, "loss": 0.0001, "step": 25452 }, { "epoch": 0.49, "learning_rate": 4.1768724647244924e-05, "loss": 0.0141, "step": 25454 }, { "epoch": 0.49, "learning_rate": 4.1768077686987686e-05, "loss": 0.0096, "step": 25456 }, { "epoch": 0.49, "learning_rate": 4.176743072673046e-05, "loss": 0.0015, "step": 25458 }, { "epoch": 0.49, "learning_rate": 4.176678376647323e-05, "loss": 0.0014, "step": 25460 }, { "epoch": 0.49, "learning_rate": 4.176613680621599e-05, "loss": 0.0001, "step": 25462 }, { "epoch": 0.49, "learning_rate": 4.176548984595877e-05, "loss": 0.0005, "step": 25464 }, { "epoch": 0.49, "learning_rate": 4.176484288570153e-05, "loss": 0.0255, "step": 25466 }, { "epoch": 0.49, "learning_rate": 4.17641959254443e-05, "loss": 0.0059, "step": 25468 }, { "epoch": 0.49, "learning_rate": 4.176354896518707e-05, "loss": 0.0002, "step": 25470 }, { "epoch": 0.49, "learning_rate": 4.176290200492984e-05, "loss": 0.0002, "step": 25472 }, { "epoch": 0.49, "learning_rate": 4.176225504467261e-05, "loss": 0.0054, "step": 25474 }, { "epoch": 0.49, "learning_rate": 4.176160808441538e-05, "loss": 0.001, "step": 25476 }, { "epoch": 0.49, "learning_rate": 4.1760961124158146e-05, "loss": 0.0002, "step": 25478 }, { "epoch": 0.49, "learning_rate": 4.1760314163900915e-05, "loss": 0.0001, "step": 25480 }, { "epoch": 0.49, "learning_rate": 4.1759667203643684e-05, "loss": 0.0001, "step": 25482 }, { "epoch": 0.49, "learning_rate": 4.1759020243386446e-05, "loss": 0.0161, "step": 25484 }, { "epoch": 0.49, "learning_rate": 4.175837328312922e-05, "loss": 0.0004, "step": 25486 }, { "epoch": 0.49, "learning_rate": 4.1757726322871984e-05, "loss": 0.0016, "step": 25488 }, { "epoch": 0.49, "learning_rate": 4.1757079362614753e-05, "loss": 0.0001, "step": 25490 }, { "epoch": 0.49, "learning_rate": 4.175643240235752e-05, "loss": 0.0032, "step": 25492 }, { "epoch": 0.49, "learning_rate": 4.175578544210029e-05, "loss": 0.0003, "step": 25494 }, { "epoch": 0.49, "learning_rate": 4.175513848184307e-05, "loss": 0.0024, "step": 25496 }, { "epoch": 0.49, "learning_rate": 4.175449152158583e-05, "loss": 0.0096, "step": 25498 }, { "epoch": 0.49, "learning_rate": 4.17538445613286e-05, "loss": 0.0001, "step": 25500 }, { "epoch": 0.49, "learning_rate": 4.175319760107137e-05, "loss": 0.0009, "step": 25502 }, { "epoch": 0.5, "learning_rate": 4.175255064081414e-05, "loss": 0.0095, "step": 25504 }, { "epoch": 0.5, "learning_rate": 4.1751903680556906e-05, "loss": 0.0048, "step": 25506 }, { "epoch": 0.5, "learning_rate": 4.1751256720299675e-05, "loss": 0.0009, "step": 25508 }, { "epoch": 0.5, "learning_rate": 4.1750609760042444e-05, "loss": 0.0001, "step": 25510 }, { "epoch": 0.5, "learning_rate": 4.1749962799785213e-05, "loss": 0.0003, "step": 25512 }, { "epoch": 0.5, "learning_rate": 4.174931583952798e-05, "loss": 0.0002, "step": 25514 }, { "epoch": 0.5, "learning_rate": 4.1748668879270745e-05, "loss": 0.0002, "step": 25516 }, { "epoch": 0.5, "learning_rate": 4.174802191901352e-05, "loss": 0.0003, "step": 25518 }, { "epoch": 0.5, "learning_rate": 4.174737495875628e-05, "loss": 0.0002, "step": 25520 }, { "epoch": 0.5, "learning_rate": 4.174672799849905e-05, "loss": 0.0001, "step": 25522 }, { "epoch": 0.5, "learning_rate": 4.174608103824182e-05, "loss": 0.0008, "step": 25524 }, { "epoch": 0.5, "learning_rate": 4.174543407798459e-05, "loss": 0.0005, "step": 25526 }, { "epoch": 0.5, "learning_rate": 4.174478711772736e-05, "loss": 0.0001, "step": 25528 }, { "epoch": 0.5, "learning_rate": 4.174414015747013e-05, "loss": 0.0001, "step": 25530 }, { "epoch": 0.5, "learning_rate": 4.17434931972129e-05, "loss": 0.002, "step": 25532 }, { "epoch": 0.5, "learning_rate": 4.1742846236955667e-05, "loss": 0.0002, "step": 25534 }, { "epoch": 0.5, "learning_rate": 4.1742199276698436e-05, "loss": 0.0004, "step": 25536 }, { "epoch": 0.5, "learning_rate": 4.1741552316441205e-05, "loss": 0.0, "step": 25538 }, { "epoch": 0.5, "learning_rate": 4.1740905356183974e-05, "loss": 0.0046, "step": 25540 }, { "epoch": 0.5, "learning_rate": 4.174025839592674e-05, "loss": 0.0014, "step": 25542 }, { "epoch": 0.5, "learning_rate": 4.1739611435669505e-05, "loss": 0.0125, "step": 25544 }, { "epoch": 0.5, "learning_rate": 4.173896447541228e-05, "loss": 0.0123, "step": 25546 }, { "epoch": 0.5, "learning_rate": 4.1738317515155043e-05, "loss": 0.0138, "step": 25548 }, { "epoch": 0.5, "learning_rate": 4.173767055489781e-05, "loss": 0.0081, "step": 25550 }, { "epoch": 0.5, "learning_rate": 4.173702359464058e-05, "loss": 0.0014, "step": 25552 }, { "epoch": 0.5, "learning_rate": 4.173637663438335e-05, "loss": 0.0, "step": 25554 }, { "epoch": 0.5, "learning_rate": 4.173572967412612e-05, "loss": 0.0, "step": 25556 }, { "epoch": 0.5, "learning_rate": 4.173508271386889e-05, "loss": 0.0, "step": 25558 }, { "epoch": 0.5, "learning_rate": 4.173443575361166e-05, "loss": 0.0001, "step": 25560 }, { "epoch": 0.5, "learning_rate": 4.173378879335443e-05, "loss": 0.0116, "step": 25562 }, { "epoch": 0.5, "learning_rate": 4.1733141833097196e-05, "loss": 0.0092, "step": 25564 }, { "epoch": 0.5, "learning_rate": 4.173249487283996e-05, "loss": 0.0008, "step": 25566 }, { "epoch": 0.5, "learning_rate": 4.1731847912582734e-05, "loss": 0.0003, "step": 25568 }, { "epoch": 0.5, "learning_rate": 4.1731200952325497e-05, "loss": 0.0001, "step": 25570 }, { "epoch": 0.5, "learning_rate": 4.173055399206827e-05, "loss": 0.0, "step": 25572 }, { "epoch": 0.5, "learning_rate": 4.172990703181104e-05, "loss": 0.0064, "step": 25574 }, { "epoch": 0.5, "learning_rate": 4.1729260071553804e-05, "loss": 0.0023, "step": 25576 }, { "epoch": 0.5, "learning_rate": 4.172861311129658e-05, "loss": 0.0005, "step": 25578 }, { "epoch": 0.5, "learning_rate": 4.172796615103934e-05, "loss": 0.0024, "step": 25580 }, { "epoch": 0.5, "learning_rate": 4.172731919078211e-05, "loss": 0.0003, "step": 25582 }, { "epoch": 0.5, "learning_rate": 4.172667223052488e-05, "loss": 0.0074, "step": 25584 }, { "epoch": 0.5, "learning_rate": 4.172602527026765e-05, "loss": 0.0, "step": 25586 }, { "epoch": 0.5, "learning_rate": 4.172537831001042e-05, "loss": 0.0032, "step": 25588 }, { "epoch": 0.5, "learning_rate": 4.172473134975319e-05, "loss": 0.0001, "step": 25590 }, { "epoch": 0.5, "learning_rate": 4.1724084389495956e-05, "loss": 0.0, "step": 25592 }, { "epoch": 0.5, "learning_rate": 4.1723437429238726e-05, "loss": 0.0001, "step": 25594 }, { "epoch": 0.5, "learning_rate": 4.1722790468981495e-05, "loss": 0.0006, "step": 25596 }, { "epoch": 0.5, "learning_rate": 4.172214350872426e-05, "loss": 0.0, "step": 25598 }, { "epoch": 0.5, "learning_rate": 4.172149654846703e-05, "loss": 0.0075, "step": 25600 }, { "epoch": 0.5, "learning_rate": 4.1720849588209795e-05, "loss": 0.0076, "step": 25602 }, { "epoch": 0.5, "learning_rate": 4.1720202627952564e-05, "loss": 0.0047, "step": 25604 }, { "epoch": 0.5, "learning_rate": 4.171955566769534e-05, "loss": 0.0009, "step": 25606 }, { "epoch": 0.5, "learning_rate": 4.17189087074381e-05, "loss": 0.0185, "step": 25608 }, { "epoch": 0.5, "learning_rate": 4.171826174718087e-05, "loss": 0.0074, "step": 25610 }, { "epoch": 0.5, "learning_rate": 4.171761478692364e-05, "loss": 0.0018, "step": 25612 }, { "epoch": 0.5, "learning_rate": 4.171696782666641e-05, "loss": 0.0094, "step": 25614 }, { "epoch": 0.5, "learning_rate": 4.171632086640918e-05, "loss": 0.0, "step": 25616 }, { "epoch": 0.5, "learning_rate": 4.171567390615195e-05, "loss": 0.0297, "step": 25618 }, { "epoch": 0.5, "learning_rate": 4.171502694589472e-05, "loss": 0.0016, "step": 25620 }, { "epoch": 0.5, "learning_rate": 4.1714379985637486e-05, "loss": 0.0001, "step": 25622 }, { "epoch": 0.5, "learning_rate": 4.1713733025380255e-05, "loss": 0.0099, "step": 25624 }, { "epoch": 0.5, "learning_rate": 4.171308606512302e-05, "loss": 0.0094, "step": 25626 }, { "epoch": 0.5, "learning_rate": 4.171243910486579e-05, "loss": 0.0001, "step": 25628 }, { "epoch": 0.5, "learning_rate": 4.1711792144608556e-05, "loss": 0.0059, "step": 25630 }, { "epoch": 0.5, "learning_rate": 4.1711145184351325e-05, "loss": 0.0007, "step": 25632 }, { "epoch": 0.5, "learning_rate": 4.1710498224094094e-05, "loss": 0.0003, "step": 25634 }, { "epoch": 0.5, "learning_rate": 4.170985126383686e-05, "loss": 0.0001, "step": 25636 }, { "epoch": 0.5, "learning_rate": 4.170920430357964e-05, "loss": 0.0022, "step": 25638 }, { "epoch": 0.5, "learning_rate": 4.17085573433224e-05, "loss": 0.0, "step": 25640 }, { "epoch": 0.5, "learning_rate": 4.170791038306517e-05, "loss": 0.0014, "step": 25642 }, { "epoch": 0.5, "learning_rate": 4.170726342280794e-05, "loss": 0.0001, "step": 25644 }, { "epoch": 0.5, "learning_rate": 4.170661646255071e-05, "loss": 0.0012, "step": 25646 }, { "epoch": 0.5, "learning_rate": 4.170596950229347e-05, "loss": 0.01, "step": 25648 }, { "epoch": 0.5, "learning_rate": 4.1705322542036246e-05, "loss": 0.0096, "step": 25650 }, { "epoch": 0.5, "learning_rate": 4.1704675581779015e-05, "loss": 0.0044, "step": 25652 }, { "epoch": 0.5, "learning_rate": 4.1704028621521785e-05, "loss": 0.0038, "step": 25654 }, { "epoch": 0.5, "learning_rate": 4.1703381661264554e-05, "loss": 0.0004, "step": 25656 }, { "epoch": 0.5, "learning_rate": 4.1702734701007316e-05, "loss": 0.0001, "step": 25658 }, { "epoch": 0.5, "learning_rate": 4.170208774075009e-05, "loss": 0.0006, "step": 25660 }, { "epoch": 0.5, "learning_rate": 4.1701440780492854e-05, "loss": 0.0013, "step": 25662 }, { "epoch": 0.5, "learning_rate": 4.170079382023562e-05, "loss": 0.0029, "step": 25664 }, { "epoch": 0.5, "learning_rate": 4.170014685997839e-05, "loss": 0.0031, "step": 25666 }, { "epoch": 0.5, "learning_rate": 4.169949989972116e-05, "loss": 0.0, "step": 25668 }, { "epoch": 0.5, "learning_rate": 4.169885293946393e-05, "loss": 0.0003, "step": 25670 }, { "epoch": 0.5, "learning_rate": 4.16982059792067e-05, "loss": 0.0, "step": 25672 }, { "epoch": 0.5, "learning_rate": 4.169755901894947e-05, "loss": 0.0034, "step": 25674 }, { "epoch": 0.5, "learning_rate": 4.169691205869224e-05, "loss": 0.0003, "step": 25676 }, { "epoch": 0.5, "learning_rate": 4.169626509843501e-05, "loss": 0.0138, "step": 25678 }, { "epoch": 0.5, "learning_rate": 4.169561813817777e-05, "loss": 0.0, "step": 25680 }, { "epoch": 0.5, "learning_rate": 4.1694971177920545e-05, "loss": 0.0038, "step": 25682 }, { "epoch": 0.5, "learning_rate": 4.1694324217663314e-05, "loss": 0.0012, "step": 25684 }, { "epoch": 0.5, "learning_rate": 4.1693677257406076e-05, "loss": 0.0155, "step": 25686 }, { "epoch": 0.5, "learning_rate": 4.169303029714885e-05, "loss": 0.0014, "step": 25688 }, { "epoch": 0.5, "learning_rate": 4.1692383336891615e-05, "loss": 0.0001, "step": 25690 }, { "epoch": 0.5, "learning_rate": 4.1691736376634384e-05, "loss": 0.0004, "step": 25692 }, { "epoch": 0.5, "learning_rate": 4.169108941637715e-05, "loss": 0.0001, "step": 25694 }, { "epoch": 0.5, "learning_rate": 4.169044245611992e-05, "loss": 0.0009, "step": 25696 }, { "epoch": 0.5, "learning_rate": 4.168979549586269e-05, "loss": 0.0001, "step": 25698 }, { "epoch": 0.5, "learning_rate": 4.168914853560546e-05, "loss": 0.0021, "step": 25700 }, { "epoch": 0.5, "learning_rate": 4.168850157534823e-05, "loss": 0.0073, "step": 25702 }, { "epoch": 0.5, "learning_rate": 4.1687854615091e-05, "loss": 0.0001, "step": 25704 }, { "epoch": 0.5, "learning_rate": 4.168720765483377e-05, "loss": 0.0001, "step": 25706 }, { "epoch": 0.5, "learning_rate": 4.168656069457653e-05, "loss": 0.0001, "step": 25708 }, { "epoch": 0.5, "learning_rate": 4.1685913734319305e-05, "loss": 0.004, "step": 25710 }, { "epoch": 0.5, "learning_rate": 4.168526677406207e-05, "loss": 0.0003, "step": 25712 }, { "epoch": 0.5, "learning_rate": 4.1684619813804844e-05, "loss": 0.0, "step": 25714 }, { "epoch": 0.5, "learning_rate": 4.168397285354761e-05, "loss": 0.0, "step": 25716 }, { "epoch": 0.5, "learning_rate": 4.1683325893290375e-05, "loss": 0.0011, "step": 25718 }, { "epoch": 0.5, "learning_rate": 4.168267893303315e-05, "loss": 0.0005, "step": 25720 }, { "epoch": 0.5, "learning_rate": 4.168203197277591e-05, "loss": 0.0014, "step": 25722 }, { "epoch": 0.5, "learning_rate": 4.168138501251868e-05, "loss": 0.0016, "step": 25724 }, { "epoch": 0.5, "learning_rate": 4.168073805226145e-05, "loss": 0.0027, "step": 25726 }, { "epoch": 0.5, "learning_rate": 4.168009109200422e-05, "loss": 0.0006, "step": 25728 }, { "epoch": 0.5, "learning_rate": 4.167944413174699e-05, "loss": 0.0, "step": 25730 }, { "epoch": 0.5, "learning_rate": 4.167879717148976e-05, "loss": 0.003, "step": 25732 }, { "epoch": 0.5, "learning_rate": 4.167815021123253e-05, "loss": 0.0001, "step": 25734 }, { "epoch": 0.5, "learning_rate": 4.16775032509753e-05, "loss": 0.0003, "step": 25736 }, { "epoch": 0.5, "learning_rate": 4.1676856290718066e-05, "loss": 0.0016, "step": 25738 }, { "epoch": 0.5, "learning_rate": 4.167620933046083e-05, "loss": 0.0011, "step": 25740 }, { "epoch": 0.5, "learning_rate": 4.1675562370203604e-05, "loss": 0.0, "step": 25742 }, { "epoch": 0.5, "learning_rate": 4.1674915409946366e-05, "loss": 0.0002, "step": 25744 }, { "epoch": 0.5, "learning_rate": 4.1674268449689135e-05, "loss": 0.0001, "step": 25746 }, { "epoch": 0.5, "learning_rate": 4.1673621489431904e-05, "loss": 0.0007, "step": 25748 }, { "epoch": 0.5, "learning_rate": 4.1672974529174673e-05, "loss": 0.0001, "step": 25750 }, { "epoch": 0.5, "learning_rate": 4.167232756891744e-05, "loss": 0.0001, "step": 25752 }, { "epoch": 0.5, "learning_rate": 4.167168060866021e-05, "loss": 0.0001, "step": 25754 }, { "epoch": 0.5, "learning_rate": 4.167103364840298e-05, "loss": 0.0001, "step": 25756 }, { "epoch": 0.5, "learning_rate": 4.167038668814575e-05, "loss": 0.0, "step": 25758 }, { "epoch": 0.5, "learning_rate": 4.166973972788852e-05, "loss": 0.0, "step": 25760 }, { "epoch": 0.5, "learning_rate": 4.166909276763129e-05, "loss": 0.0081, "step": 25762 }, { "epoch": 0.5, "learning_rate": 4.166844580737406e-05, "loss": 0.0174, "step": 25764 }, { "epoch": 0.5, "learning_rate": 4.1667798847116826e-05, "loss": 0.0001, "step": 25766 }, { "epoch": 0.5, "learning_rate": 4.166715188685959e-05, "loss": 0.0, "step": 25768 }, { "epoch": 0.5, "learning_rate": 4.1666504926602364e-05, "loss": 0.0003, "step": 25770 }, { "epoch": 0.5, "learning_rate": 4.166585796634513e-05, "loss": 0.0004, "step": 25772 }, { "epoch": 0.5, "learning_rate": 4.1665211006087896e-05, "loss": 0.0025, "step": 25774 }, { "epoch": 0.5, "learning_rate": 4.1664564045830665e-05, "loss": 0.0012, "step": 25776 }, { "epoch": 0.5, "learning_rate": 4.1663917085573434e-05, "loss": 0.0001, "step": 25778 }, { "epoch": 0.5, "learning_rate": 4.16632701253162e-05, "loss": 0.0175, "step": 25780 }, { "epoch": 0.5, "learning_rate": 4.166262316505897e-05, "loss": 0.0169, "step": 25782 }, { "epoch": 0.5, "learning_rate": 4.166197620480174e-05, "loss": 0.0005, "step": 25784 }, { "epoch": 0.5, "learning_rate": 4.166132924454451e-05, "loss": 0.0005, "step": 25786 }, { "epoch": 0.5, "learning_rate": 4.166068228428728e-05, "loss": 0.0004, "step": 25788 }, { "epoch": 0.5, "learning_rate": 4.166003532403004e-05, "loss": 0.0105, "step": 25790 }, { "epoch": 0.5, "learning_rate": 4.165938836377282e-05, "loss": 0.0014, "step": 25792 }, { "epoch": 0.5, "learning_rate": 4.1658741403515587e-05, "loss": 0.0006, "step": 25794 }, { "epoch": 0.5, "learning_rate": 4.1658094443258356e-05, "loss": 0.0001, "step": 25796 }, { "epoch": 0.5, "learning_rate": 4.1657447483001125e-05, "loss": 0.0009, "step": 25798 }, { "epoch": 0.5, "learning_rate": 4.165680052274389e-05, "loss": 0.0019, "step": 25800 }, { "epoch": 0.5, "learning_rate": 4.165615356248666e-05, "loss": 0.0015, "step": 25802 }, { "epoch": 0.5, "learning_rate": 4.1655506602229425e-05, "loss": 0.0, "step": 25804 }, { "epoch": 0.5, "learning_rate": 4.1654859641972194e-05, "loss": 0.0093, "step": 25806 }, { "epoch": 0.5, "learning_rate": 4.165421268171496e-05, "loss": 0.0003, "step": 25808 }, { "epoch": 0.5, "learning_rate": 4.165356572145773e-05, "loss": 0.006, "step": 25810 }, { "epoch": 0.5, "learning_rate": 4.16529187612005e-05, "loss": 0.0124, "step": 25812 }, { "epoch": 0.5, "learning_rate": 4.165227180094327e-05, "loss": 0.0002, "step": 25814 }, { "epoch": 0.5, "learning_rate": 4.165162484068604e-05, "loss": 0.1634, "step": 25816 }, { "epoch": 0.5, "learning_rate": 4.165097788042881e-05, "loss": 0.0001, "step": 25818 }, { "epoch": 0.5, "learning_rate": 4.165033092017158e-05, "loss": 0.0002, "step": 25820 }, { "epoch": 0.5, "learning_rate": 4.164968395991434e-05, "loss": 0.0315, "step": 25822 }, { "epoch": 0.5, "learning_rate": 4.1649036999657116e-05, "loss": 0.0001, "step": 25824 }, { "epoch": 0.5, "learning_rate": 4.164839003939988e-05, "loss": 0.0001, "step": 25826 }, { "epoch": 0.5, "learning_rate": 4.164774307914265e-05, "loss": 0.0003, "step": 25828 }, { "epoch": 0.5, "learning_rate": 4.164709611888542e-05, "loss": 0.011, "step": 25830 }, { "epoch": 0.5, "learning_rate": 4.1646449158628186e-05, "loss": 0.0001, "step": 25832 }, { "epoch": 0.5, "learning_rate": 4.1645802198370955e-05, "loss": 0.0002, "step": 25834 }, { "epoch": 0.5, "learning_rate": 4.1645155238113724e-05, "loss": 0.0015, "step": 25836 }, { "epoch": 0.5, "learning_rate": 4.164450827785649e-05, "loss": 0.0194, "step": 25838 }, { "epoch": 0.5, "learning_rate": 4.164386131759926e-05, "loss": 0.0015, "step": 25840 }, { "epoch": 0.5, "learning_rate": 4.164321435734203e-05, "loss": 0.0001, "step": 25842 }, { "epoch": 0.5, "learning_rate": 4.16425673970848e-05, "loss": 0.0034, "step": 25844 }, { "epoch": 0.5, "learning_rate": 4.164192043682757e-05, "loss": 0.0056, "step": 25846 }, { "epoch": 0.5, "learning_rate": 4.164127347657034e-05, "loss": 0.0004, "step": 25848 }, { "epoch": 0.5, "learning_rate": 4.16406265163131e-05, "loss": 0.0127, "step": 25850 }, { "epoch": 0.5, "learning_rate": 4.1639979556055876e-05, "loss": 0.0034, "step": 25852 }, { "epoch": 0.5, "learning_rate": 4.163933259579864e-05, "loss": 0.007, "step": 25854 }, { "epoch": 0.5, "learning_rate": 4.1638685635541415e-05, "loss": 0.0, "step": 25856 }, { "epoch": 0.5, "learning_rate": 4.163803867528418e-05, "loss": 0.0, "step": 25858 }, { "epoch": 0.5, "learning_rate": 4.1637391715026946e-05, "loss": 0.0001, "step": 25860 }, { "epoch": 0.5, "learning_rate": 4.163674475476972e-05, "loss": 0.0064, "step": 25862 }, { "epoch": 0.5, "learning_rate": 4.1636097794512484e-05, "loss": 0.0, "step": 25864 }, { "epoch": 0.5, "learning_rate": 4.163545083425525e-05, "loss": 0.0001, "step": 25866 }, { "epoch": 0.5, "learning_rate": 4.163480387399802e-05, "loss": 0.0023, "step": 25868 }, { "epoch": 0.5, "learning_rate": 4.163415691374079e-05, "loss": 0.0002, "step": 25870 }, { "epoch": 0.5, "learning_rate": 4.1633509953483554e-05, "loss": 0.007, "step": 25872 }, { "epoch": 0.5, "learning_rate": 4.163286299322633e-05, "loss": 0.0001, "step": 25874 }, { "epoch": 0.5, "learning_rate": 4.16322160329691e-05, "loss": 0.0006, "step": 25876 }, { "epoch": 0.5, "learning_rate": 4.163156907271187e-05, "loss": 0.0002, "step": 25878 }, { "epoch": 0.5, "learning_rate": 4.163092211245464e-05, "loss": 0.0003, "step": 25880 }, { "epoch": 0.5, "learning_rate": 4.16302751521974e-05, "loss": 0.0116, "step": 25882 }, { "epoch": 0.5, "learning_rate": 4.1629628191940175e-05, "loss": 0.0001, "step": 25884 }, { "epoch": 0.5, "learning_rate": 4.162898123168294e-05, "loss": 0.0001, "step": 25886 }, { "epoch": 0.5, "learning_rate": 4.1628334271425706e-05, "loss": 0.0091, "step": 25888 }, { "epoch": 0.5, "learning_rate": 4.1627687311168476e-05, "loss": 0.0, "step": 25890 }, { "epoch": 0.5, "learning_rate": 4.1627040350911245e-05, "loss": 0.0001, "step": 25892 }, { "epoch": 0.5, "learning_rate": 4.1626393390654014e-05, "loss": 0.0, "step": 25894 }, { "epoch": 0.5, "learning_rate": 4.162574643039678e-05, "loss": 0.0064, "step": 25896 }, { "epoch": 0.5, "learning_rate": 4.162509947013955e-05, "loss": 0.0001, "step": 25898 }, { "epoch": 0.5, "learning_rate": 4.162445250988232e-05, "loss": 0.0054, "step": 25900 }, { "epoch": 0.5, "learning_rate": 4.162380554962509e-05, "loss": 0.0001, "step": 25902 }, { "epoch": 0.5, "learning_rate": 4.162315858936785e-05, "loss": 0.0006, "step": 25904 }, { "epoch": 0.5, "learning_rate": 4.162251162911063e-05, "loss": 0.0001, "step": 25906 }, { "epoch": 0.5, "learning_rate": 4.16218646688534e-05, "loss": 0.0046, "step": 25908 }, { "epoch": 0.5, "learning_rate": 4.162121770859616e-05, "loss": 0.0, "step": 25910 }, { "epoch": 0.5, "learning_rate": 4.1620570748338935e-05, "loss": 0.0031, "step": 25912 }, { "epoch": 0.5, "learning_rate": 4.16199237880817e-05, "loss": 0.0189, "step": 25914 }, { "epoch": 0.5, "learning_rate": 4.161927682782447e-05, "loss": 0.0024, "step": 25916 }, { "epoch": 0.5, "learning_rate": 4.1618629867567236e-05, "loss": 0.0001, "step": 25918 }, { "epoch": 0.5, "learning_rate": 4.1617982907310005e-05, "loss": 0.0003, "step": 25920 }, { "epoch": 0.5, "learning_rate": 4.1617335947052774e-05, "loss": 0.0003, "step": 25922 }, { "epoch": 0.5, "learning_rate": 4.161668898679554e-05, "loss": 0.0001, "step": 25924 }, { "epoch": 0.5, "learning_rate": 4.161604202653831e-05, "loss": 0.0001, "step": 25926 }, { "epoch": 0.5, "learning_rate": 4.161539506628108e-05, "loss": 0.0004, "step": 25928 }, { "epoch": 0.5, "learning_rate": 4.161474810602385e-05, "loss": 0.013, "step": 25930 }, { "epoch": 0.5, "learning_rate": 4.161410114576661e-05, "loss": 0.0092, "step": 25932 }, { "epoch": 0.5, "learning_rate": 4.161345418550939e-05, "loss": 0.0, "step": 25934 }, { "epoch": 0.5, "learning_rate": 4.161280722525215e-05, "loss": 0.008, "step": 25936 }, { "epoch": 0.5, "learning_rate": 4.161216026499493e-05, "loss": 0.041, "step": 25938 }, { "epoch": 0.5, "learning_rate": 4.1611513304737696e-05, "loss": 0.0383, "step": 25940 }, { "epoch": 0.5, "learning_rate": 4.161086634448046e-05, "loss": 0.0016, "step": 25942 }, { "epoch": 0.5, "learning_rate": 4.1610219384223234e-05, "loss": 0.0028, "step": 25944 }, { "epoch": 0.5, "learning_rate": 4.1609572423965996e-05, "loss": 0.0479, "step": 25946 }, { "epoch": 0.5, "learning_rate": 4.1608925463708765e-05, "loss": 0.0057, "step": 25948 }, { "epoch": 0.5, "learning_rate": 4.1608278503451534e-05, "loss": 0.0005, "step": 25950 }, { "epoch": 0.5, "learning_rate": 4.1607631543194304e-05, "loss": 0.0052, "step": 25952 }, { "epoch": 0.5, "learning_rate": 4.160698458293707e-05, "loss": 0.0015, "step": 25954 }, { "epoch": 0.5, "learning_rate": 4.160633762267984e-05, "loss": 0.0029, "step": 25956 }, { "epoch": 0.5, "learning_rate": 4.160569066242261e-05, "loss": 0.0004, "step": 25958 }, { "epoch": 0.5, "learning_rate": 4.160504370216538e-05, "loss": 0.0001, "step": 25960 }, { "epoch": 0.5, "learning_rate": 4.160439674190815e-05, "loss": 0.0018, "step": 25962 }, { "epoch": 0.5, "learning_rate": 4.160374978165091e-05, "loss": 0.0006, "step": 25964 }, { "epoch": 0.5, "learning_rate": 4.160310282139369e-05, "loss": 0.0001, "step": 25966 }, { "epoch": 0.5, "learning_rate": 4.160245586113645e-05, "loss": 0.0053, "step": 25968 }, { "epoch": 0.5, "learning_rate": 4.160180890087922e-05, "loss": 0.0003, "step": 25970 }, { "epoch": 0.5, "learning_rate": 4.160116194062199e-05, "loss": 0.0088, "step": 25972 }, { "epoch": 0.5, "learning_rate": 4.160051498036476e-05, "loss": 0.0077, "step": 25974 }, { "epoch": 0.5, "learning_rate": 4.1599868020107526e-05, "loss": 0.0035, "step": 25976 }, { "epoch": 0.5, "learning_rate": 4.1599221059850295e-05, "loss": 0.0, "step": 25978 }, { "epoch": 0.5, "learning_rate": 4.1598574099593064e-05, "loss": 0.0004, "step": 25980 }, { "epoch": 0.5, "learning_rate": 4.159792713933583e-05, "loss": 0.0045, "step": 25982 }, { "epoch": 0.5, "learning_rate": 4.15972801790786e-05, "loss": 0.0, "step": 25984 }, { "epoch": 0.5, "learning_rate": 4.159663321882137e-05, "loss": 0.0001, "step": 25986 }, { "epoch": 0.5, "learning_rate": 4.159598625856414e-05, "loss": 0.0131, "step": 25988 }, { "epoch": 0.5, "learning_rate": 4.159533929830691e-05, "loss": 0.0063, "step": 25990 }, { "epoch": 0.5, "learning_rate": 4.159469233804967e-05, "loss": 0.0097, "step": 25992 }, { "epoch": 0.5, "learning_rate": 4.159404537779245e-05, "loss": 0.0002, "step": 25994 }, { "epoch": 0.5, "learning_rate": 4.159339841753521e-05, "loss": 0.0019, "step": 25996 }, { "epoch": 0.5, "learning_rate": 4.1592751457277986e-05, "loss": 0.0001, "step": 25998 }, { "epoch": 0.5, "learning_rate": 4.159210449702075e-05, "loss": 0.0002, "step": 26000 }, { "epoch": 0.5, "learning_rate": 4.159145753676352e-05, "loss": 0.0001, "step": 26002 }, { "epoch": 0.5, "learning_rate": 4.1590810576506286e-05, "loss": 0.0019, "step": 26004 }, { "epoch": 0.5, "learning_rate": 4.1590163616249055e-05, "loss": 0.0001, "step": 26006 }, { "epoch": 0.5, "learning_rate": 4.1589516655991824e-05, "loss": 0.0001, "step": 26008 }, { "epoch": 0.5, "learning_rate": 4.1588869695734593e-05, "loss": 0.0001, "step": 26010 }, { "epoch": 0.5, "learning_rate": 4.158822273547736e-05, "loss": 0.0031, "step": 26012 }, { "epoch": 0.5, "learning_rate": 4.1587575775220125e-05, "loss": 0.0068, "step": 26014 }, { "epoch": 0.5, "learning_rate": 4.15869288149629e-05, "loss": 0.0199, "step": 26016 }, { "epoch": 0.5, "learning_rate": 4.158628185470567e-05, "loss": 0.0005, "step": 26018 }, { "epoch": 0.51, "learning_rate": 4.158563489444844e-05, "loss": 0.002, "step": 26020 }, { "epoch": 0.51, "learning_rate": 4.158498793419121e-05, "loss": 0.0, "step": 26022 }, { "epoch": 0.51, "learning_rate": 4.158434097393397e-05, "loss": 0.0002, "step": 26024 }, { "epoch": 0.51, "learning_rate": 4.1583694013676746e-05, "loss": 0.0003, "step": 26026 }, { "epoch": 0.51, "learning_rate": 4.158304705341951e-05, "loss": 0.0053, "step": 26028 }, { "epoch": 0.51, "learning_rate": 4.158240009316228e-05, "loss": 0.0, "step": 26030 }, { "epoch": 0.51, "learning_rate": 4.158175313290505e-05, "loss": 0.0012, "step": 26032 }, { "epoch": 0.51, "learning_rate": 4.1581106172647816e-05, "loss": 0.0046, "step": 26034 }, { "epoch": 0.51, "learning_rate": 4.1580459212390585e-05, "loss": 0.0, "step": 26036 }, { "epoch": 0.51, "learning_rate": 4.1579812252133354e-05, "loss": 0.0007, "step": 26038 }, { "epoch": 0.51, "learning_rate": 4.157916529187612e-05, "loss": 0.0, "step": 26040 }, { "epoch": 0.51, "learning_rate": 4.157851833161889e-05, "loss": 0.0045, "step": 26042 }, { "epoch": 0.51, "learning_rate": 4.157787137136166e-05, "loss": 0.0001, "step": 26044 }, { "epoch": 0.51, "learning_rate": 4.1577224411104423e-05, "loss": 0.0047, "step": 26046 }, { "epoch": 0.51, "learning_rate": 4.15765774508472e-05, "loss": 0.0006, "step": 26048 }, { "epoch": 0.51, "learning_rate": 4.157593049058996e-05, "loss": 0.0001, "step": 26050 }, { "epoch": 0.51, "learning_rate": 4.157528353033273e-05, "loss": 0.0061, "step": 26052 }, { "epoch": 0.51, "learning_rate": 4.1574636570075507e-05, "loss": 0.0002, "step": 26054 }, { "epoch": 0.51, "learning_rate": 4.157398960981827e-05, "loss": 0.0021, "step": 26056 }, { "epoch": 0.51, "learning_rate": 4.157334264956104e-05, "loss": 0.0001, "step": 26058 }, { "epoch": 0.51, "learning_rate": 4.157269568930381e-05, "loss": 0.0007, "step": 26060 }, { "epoch": 0.51, "learning_rate": 4.1572048729046576e-05, "loss": 0.0001, "step": 26062 }, { "epoch": 0.51, "learning_rate": 4.1571401768789345e-05, "loss": 0.0, "step": 26064 }, { "epoch": 0.51, "learning_rate": 4.1570754808532114e-05, "loss": 0.0, "step": 26066 }, { "epoch": 0.51, "learning_rate": 4.157010784827488e-05, "loss": 0.0016, "step": 26068 }, { "epoch": 0.51, "learning_rate": 4.156946088801765e-05, "loss": 0.0001, "step": 26070 }, { "epoch": 0.51, "learning_rate": 4.156881392776042e-05, "loss": 0.0003, "step": 26072 }, { "epoch": 0.51, "learning_rate": 4.1568166967503184e-05, "loss": 0.0005, "step": 26074 }, { "epoch": 0.51, "learning_rate": 4.156752000724596e-05, "loss": 0.0003, "step": 26076 }, { "epoch": 0.51, "learning_rate": 4.156687304698872e-05, "loss": 0.0003, "step": 26078 }, { "epoch": 0.51, "learning_rate": 4.15662260867315e-05, "loss": 0.0033, "step": 26080 }, { "epoch": 0.51, "learning_rate": 4.156557912647426e-05, "loss": 0.0205, "step": 26082 }, { "epoch": 0.51, "learning_rate": 4.156493216621703e-05, "loss": 0.0005, "step": 26084 }, { "epoch": 0.51, "learning_rate": 4.1564285205959805e-05, "loss": 0.0029, "step": 26086 }, { "epoch": 0.51, "learning_rate": 4.156363824570257e-05, "loss": 0.0001, "step": 26088 }, { "epoch": 0.51, "learning_rate": 4.1562991285445337e-05, "loss": 0.0015, "step": 26090 }, { "epoch": 0.51, "learning_rate": 4.1562344325188106e-05, "loss": 0.0056, "step": 26092 }, { "epoch": 0.51, "learning_rate": 4.1561697364930875e-05, "loss": 0.0001, "step": 26094 }, { "epoch": 0.51, "learning_rate": 4.1561050404673644e-05, "loss": 0.0001, "step": 26096 }, { "epoch": 0.51, "learning_rate": 4.156040344441641e-05, "loss": 0.0, "step": 26098 }, { "epoch": 0.51, "learning_rate": 4.155975648415918e-05, "loss": 0.0, "step": 26100 }, { "epoch": 0.51, "learning_rate": 4.155910952390195e-05, "loss": 0.0001, "step": 26102 }, { "epoch": 0.51, "learning_rate": 4.155846256364472e-05, "loss": 0.0147, "step": 26104 }, { "epoch": 0.51, "learning_rate": 4.155781560338748e-05, "loss": 0.0003, "step": 26106 }, { "epoch": 0.51, "learning_rate": 4.155716864313026e-05, "loss": 0.0013, "step": 26108 }, { "epoch": 0.51, "learning_rate": 4.155652168287302e-05, "loss": 0.0001, "step": 26110 }, { "epoch": 0.51, "learning_rate": 4.155587472261579e-05, "loss": 0.0001, "step": 26112 }, { "epoch": 0.51, "learning_rate": 4.155522776235856e-05, "loss": 0.0022, "step": 26114 }, { "epoch": 0.51, "learning_rate": 4.155458080210133e-05, "loss": 0.0007, "step": 26116 }, { "epoch": 0.51, "learning_rate": 4.15539338418441e-05, "loss": 0.003, "step": 26118 }, { "epoch": 0.51, "learning_rate": 4.1553286881586866e-05, "loss": 0.0001, "step": 26120 }, { "epoch": 0.51, "learning_rate": 4.1552639921329635e-05, "loss": 0.0011, "step": 26122 }, { "epoch": 0.51, "learning_rate": 4.1551992961072404e-05, "loss": 0.0119, "step": 26124 }, { "epoch": 0.51, "learning_rate": 4.155134600081517e-05, "loss": 0.0002, "step": 26126 }, { "epoch": 0.51, "learning_rate": 4.1550699040557936e-05, "loss": 0.0001, "step": 26128 }, { "epoch": 0.51, "learning_rate": 4.155005208030071e-05, "loss": 0.0504, "step": 26130 }, { "epoch": 0.51, "learning_rate": 4.154940512004348e-05, "loss": 0.0, "step": 26132 }, { "epoch": 0.51, "learning_rate": 4.154875815978624e-05, "loss": 0.0001, "step": 26134 }, { "epoch": 0.51, "learning_rate": 4.154811119952902e-05, "loss": 0.0048, "step": 26136 }, { "epoch": 0.51, "learning_rate": 4.154746423927178e-05, "loss": 0.008, "step": 26138 }, { "epoch": 0.51, "learning_rate": 4.154681727901456e-05, "loss": 0.0003, "step": 26140 }, { "epoch": 0.51, "learning_rate": 4.154617031875732e-05, "loss": 0.027, "step": 26142 }, { "epoch": 0.51, "learning_rate": 4.154552335850009e-05, "loss": 0.0003, "step": 26144 }, { "epoch": 0.51, "learning_rate": 4.154487639824286e-05, "loss": 0.0006, "step": 26146 }, { "epoch": 0.51, "learning_rate": 4.1544229437985626e-05, "loss": 0.0184, "step": 26148 }, { "epoch": 0.51, "learning_rate": 4.1543582477728395e-05, "loss": 0.0001, "step": 26150 }, { "epoch": 0.51, "learning_rate": 4.1542935517471165e-05, "loss": 0.0, "step": 26152 }, { "epoch": 0.51, "learning_rate": 4.1542288557213934e-05, "loss": 0.0085, "step": 26154 }, { "epoch": 0.51, "learning_rate": 4.1541641596956696e-05, "loss": 0.0004, "step": 26156 }, { "epoch": 0.51, "learning_rate": 4.154099463669947e-05, "loss": 0.0004, "step": 26158 }, { "epoch": 0.51, "learning_rate": 4.1540347676442234e-05, "loss": 0.0039, "step": 26160 }, { "epoch": 0.51, "learning_rate": 4.153970071618501e-05, "loss": 0.0001, "step": 26162 }, { "epoch": 0.51, "learning_rate": 4.153905375592778e-05, "loss": 0.0006, "step": 26164 }, { "epoch": 0.51, "learning_rate": 4.153840679567054e-05, "loss": 0.0006, "step": 26166 }, { "epoch": 0.51, "learning_rate": 4.153775983541332e-05, "loss": 0.003, "step": 26168 }, { "epoch": 0.51, "learning_rate": 4.153711287515608e-05, "loss": 0.0004, "step": 26170 }, { "epoch": 0.51, "learning_rate": 4.153646591489885e-05, "loss": 0.0016, "step": 26172 }, { "epoch": 0.51, "learning_rate": 4.153581895464162e-05, "loss": 0.0004, "step": 26174 }, { "epoch": 0.51, "learning_rate": 4.153517199438439e-05, "loss": 0.0075, "step": 26176 }, { "epoch": 0.51, "learning_rate": 4.1534525034127156e-05, "loss": 0.0005, "step": 26178 }, { "epoch": 0.51, "learning_rate": 4.1533878073869925e-05, "loss": 0.0049, "step": 26180 }, { "epoch": 0.51, "learning_rate": 4.1533231113612694e-05, "loss": 0.0006, "step": 26182 }, { "epoch": 0.51, "learning_rate": 4.153258415335546e-05, "loss": 0.0008, "step": 26184 }, { "epoch": 0.51, "learning_rate": 4.153193719309823e-05, "loss": 0.0, "step": 26186 }, { "epoch": 0.51, "learning_rate": 4.1531290232840995e-05, "loss": 0.0008, "step": 26188 }, { "epoch": 0.51, "learning_rate": 4.153064327258377e-05, "loss": 0.0009, "step": 26190 }, { "epoch": 0.51, "learning_rate": 4.152999631232653e-05, "loss": 0.0122, "step": 26192 }, { "epoch": 0.51, "learning_rate": 4.15293493520693e-05, "loss": 0.0024, "step": 26194 }, { "epoch": 0.51, "learning_rate": 4.152870239181208e-05, "loss": 0.0, "step": 26196 }, { "epoch": 0.51, "learning_rate": 4.152805543155484e-05, "loss": 0.0007, "step": 26198 }, { "epoch": 0.51, "learning_rate": 4.152740847129761e-05, "loss": 0.0094, "step": 26200 }, { "epoch": 0.51, "learning_rate": 4.152676151104038e-05, "loss": 0.0001, "step": 26202 }, { "epoch": 0.51, "learning_rate": 4.152611455078315e-05, "loss": 0.0028, "step": 26204 }, { "epoch": 0.51, "learning_rate": 4.1525467590525916e-05, "loss": 0.0131, "step": 26206 }, { "epoch": 0.51, "learning_rate": 4.1524820630268685e-05, "loss": 0.0, "step": 26208 }, { "epoch": 0.51, "learning_rate": 4.1524173670011454e-05, "loss": 0.0, "step": 26210 }, { "epoch": 0.51, "learning_rate": 4.1523526709754224e-05, "loss": 0.0, "step": 26212 }, { "epoch": 0.51, "learning_rate": 4.152287974949699e-05, "loss": 0.0, "step": 26214 }, { "epoch": 0.51, "learning_rate": 4.1522232789239755e-05, "loss": 0.0, "step": 26216 }, { "epoch": 0.51, "learning_rate": 4.152158582898253e-05, "loss": 0.008, "step": 26218 }, { "epoch": 0.51, "learning_rate": 4.152093886872529e-05, "loss": 0.0007, "step": 26220 }, { "epoch": 0.51, "learning_rate": 4.152029190846807e-05, "loss": 0.0, "step": 26222 }, { "epoch": 0.51, "learning_rate": 4.151964494821083e-05, "loss": 0.0, "step": 26224 }, { "epoch": 0.51, "learning_rate": 4.15189979879536e-05, "loss": 0.0015, "step": 26226 }, { "epoch": 0.51, "learning_rate": 4.151835102769637e-05, "loss": 0.0127, "step": 26228 }, { "epoch": 0.51, "learning_rate": 4.151770406743914e-05, "loss": 0.0001, "step": 26230 }, { "epoch": 0.51, "learning_rate": 4.151705710718191e-05, "loss": 0.0033, "step": 26232 }, { "epoch": 0.51, "learning_rate": 4.151641014692468e-05, "loss": 0.0, "step": 26234 }, { "epoch": 0.51, "learning_rate": 4.1515763186667446e-05, "loss": 0.0006, "step": 26236 }, { "epoch": 0.51, "learning_rate": 4.151511622641021e-05, "loss": 0.0017, "step": 26238 }, { "epoch": 0.51, "learning_rate": 4.1514469266152984e-05, "loss": 0.0037, "step": 26240 }, { "epoch": 0.51, "learning_rate": 4.151382230589575e-05, "loss": 0.0, "step": 26242 }, { "epoch": 0.51, "learning_rate": 4.151317534563852e-05, "loss": 0.002, "step": 26244 }, { "epoch": 0.51, "learning_rate": 4.151252838538129e-05, "loss": 0.0, "step": 26246 }, { "epoch": 0.51, "learning_rate": 4.1511881425124054e-05, "loss": 0.0341, "step": 26248 }, { "epoch": 0.51, "learning_rate": 4.151123446486683e-05, "loss": 0.0001, "step": 26250 }, { "epoch": 0.51, "learning_rate": 4.151058750460959e-05, "loss": 0.0, "step": 26252 }, { "epoch": 0.51, "learning_rate": 4.150994054435236e-05, "loss": 0.0, "step": 26254 }, { "epoch": 0.51, "learning_rate": 4.150929358409513e-05, "loss": 0.0001, "step": 26256 }, { "epoch": 0.51, "learning_rate": 4.15086466238379e-05, "loss": 0.0006, "step": 26258 }, { "epoch": 0.51, "learning_rate": 4.150799966358067e-05, "loss": 0.0013, "step": 26260 }, { "epoch": 0.51, "learning_rate": 4.150735270332344e-05, "loss": 0.0001, "step": 26262 }, { "epoch": 0.51, "learning_rate": 4.1506705743066206e-05, "loss": 0.0001, "step": 26264 }, { "epoch": 0.51, "learning_rate": 4.1506058782808975e-05, "loss": 0.0001, "step": 26266 }, { "epoch": 0.51, "learning_rate": 4.1505411822551744e-05, "loss": 0.0004, "step": 26268 }, { "epoch": 0.51, "learning_rate": 4.150476486229451e-05, "loss": 0.0018, "step": 26270 }, { "epoch": 0.51, "learning_rate": 4.150411790203728e-05, "loss": 0.0022, "step": 26272 }, { "epoch": 0.51, "learning_rate": 4.150347094178005e-05, "loss": 0.0002, "step": 26274 }, { "epoch": 0.51, "learning_rate": 4.1502823981522814e-05, "loss": 0.0, "step": 26276 }, { "epoch": 0.51, "learning_rate": 4.150217702126559e-05, "loss": 0.0123, "step": 26278 }, { "epoch": 0.51, "learning_rate": 4.150153006100835e-05, "loss": 0.0001, "step": 26280 }, { "epoch": 0.51, "learning_rate": 4.150088310075113e-05, "loss": 0.0274, "step": 26282 }, { "epoch": 0.51, "learning_rate": 4.150023614049389e-05, "loss": 0.0003, "step": 26284 }, { "epoch": 0.51, "learning_rate": 4.149958918023666e-05, "loss": 0.0014, "step": 26286 }, { "epoch": 0.51, "learning_rate": 4.149894221997943e-05, "loss": 0.0001, "step": 26288 }, { "epoch": 0.51, "learning_rate": 4.14982952597222e-05, "loss": 0.0, "step": 26290 }, { "epoch": 0.51, "learning_rate": 4.1497648299464967e-05, "loss": 0.0002, "step": 26292 }, { "epoch": 0.51, "learning_rate": 4.1497001339207736e-05, "loss": 0.0001, "step": 26294 }, { "epoch": 0.51, "learning_rate": 4.1496354378950505e-05, "loss": 0.0, "step": 26296 }, { "epoch": 0.51, "learning_rate": 4.149570741869327e-05, "loss": 0.005, "step": 26298 }, { "epoch": 0.51, "learning_rate": 4.149506045843604e-05, "loss": 0.0002, "step": 26300 }, { "epoch": 0.51, "learning_rate": 4.1494413498178805e-05, "loss": 0.0073, "step": 26302 }, { "epoch": 0.51, "learning_rate": 4.149376653792158e-05, "loss": 0.0002, "step": 26304 }, { "epoch": 0.51, "learning_rate": 4.1493119577664343e-05, "loss": 0.0002, "step": 26306 }, { "epoch": 0.51, "learning_rate": 4.149247261740711e-05, "loss": 0.0003, "step": 26308 }, { "epoch": 0.51, "learning_rate": 4.149182565714989e-05, "loss": 0.0001, "step": 26310 }, { "epoch": 0.51, "learning_rate": 4.149117869689265e-05, "loss": 0.0001, "step": 26312 }, { "epoch": 0.51, "learning_rate": 4.149053173663542e-05, "loss": 0.0001, "step": 26314 }, { "epoch": 0.51, "learning_rate": 4.148988477637819e-05, "loss": 0.0005, "step": 26316 }, { "epoch": 0.51, "learning_rate": 4.148923781612096e-05, "loss": 0.0001, "step": 26318 }, { "epoch": 0.51, "learning_rate": 4.148859085586373e-05, "loss": 0.0001, "step": 26320 }, { "epoch": 0.51, "learning_rate": 4.1487943895606496e-05, "loss": 0.0001, "step": 26322 }, { "epoch": 0.51, "learning_rate": 4.1487296935349265e-05, "loss": 0.0002, "step": 26324 }, { "epoch": 0.51, "learning_rate": 4.1486649975092034e-05, "loss": 0.0075, "step": 26326 }, { "epoch": 0.51, "learning_rate": 4.14860030148348e-05, "loss": 0.0002, "step": 26328 }, { "epoch": 0.51, "learning_rate": 4.1485356054577566e-05, "loss": 0.0002, "step": 26330 }, { "epoch": 0.51, "learning_rate": 4.148470909432034e-05, "loss": 0.0008, "step": 26332 }, { "epoch": 0.51, "learning_rate": 4.1484062134063104e-05, "loss": 0.0001, "step": 26334 }, { "epoch": 0.51, "learning_rate": 4.148341517380587e-05, "loss": 0.0001, "step": 26336 }, { "epoch": 0.51, "learning_rate": 4.148276821354864e-05, "loss": 0.0012, "step": 26338 }, { "epoch": 0.51, "learning_rate": 4.148212125329141e-05, "loss": 0.0001, "step": 26340 }, { "epoch": 0.51, "learning_rate": 4.148147429303418e-05, "loss": 0.0, "step": 26342 }, { "epoch": 0.51, "learning_rate": 4.148082733277695e-05, "loss": 0.0, "step": 26344 }, { "epoch": 0.51, "learning_rate": 4.148018037251972e-05, "loss": 0.0001, "step": 26346 }, { "epoch": 0.51, "learning_rate": 4.147953341226249e-05, "loss": 0.0001, "step": 26348 }, { "epoch": 0.51, "learning_rate": 4.1478886452005256e-05, "loss": 0.0002, "step": 26350 }, { "epoch": 0.51, "learning_rate": 4.147823949174802e-05, "loss": 0.0001, "step": 26352 }, { "epoch": 0.51, "learning_rate": 4.1477592531490795e-05, "loss": 0.0001, "step": 26354 }, { "epoch": 0.51, "learning_rate": 4.1476945571233564e-05, "loss": 0.0001, "step": 26356 }, { "epoch": 0.51, "learning_rate": 4.1476298610976326e-05, "loss": 0.0001, "step": 26358 }, { "epoch": 0.51, "learning_rate": 4.14756516507191e-05, "loss": 0.0, "step": 26360 }, { "epoch": 0.51, "learning_rate": 4.1475004690461864e-05, "loss": 0.0004, "step": 26362 }, { "epoch": 0.51, "learning_rate": 4.147435773020464e-05, "loss": 0.0001, "step": 26364 }, { "epoch": 0.51, "learning_rate": 4.14737107699474e-05, "loss": 0.0366, "step": 26366 }, { "epoch": 0.51, "learning_rate": 4.147306380969017e-05, "loss": 0.0103, "step": 26368 }, { "epoch": 0.51, "learning_rate": 4.147241684943294e-05, "loss": 0.0002, "step": 26370 }, { "epoch": 0.51, "learning_rate": 4.147176988917571e-05, "loss": 0.0001, "step": 26372 }, { "epoch": 0.51, "learning_rate": 4.147112292891848e-05, "loss": 0.0004, "step": 26374 }, { "epoch": 0.51, "learning_rate": 4.147047596866125e-05, "loss": 0.0001, "step": 26376 }, { "epoch": 0.51, "learning_rate": 4.146982900840402e-05, "loss": 0.0012, "step": 26378 }, { "epoch": 0.51, "learning_rate": 4.146918204814678e-05, "loss": 0.0011, "step": 26380 }, { "epoch": 0.51, "learning_rate": 4.1468535087889555e-05, "loss": 0.0006, "step": 26382 }, { "epoch": 0.51, "learning_rate": 4.146788812763232e-05, "loss": 0.0, "step": 26384 }, { "epoch": 0.51, "learning_rate": 4.146724116737509e-05, "loss": 0.0019, "step": 26386 }, { "epoch": 0.51, "learning_rate": 4.146659420711786e-05, "loss": 0.0001, "step": 26388 }, { "epoch": 0.51, "learning_rate": 4.1465947246860625e-05, "loss": 0.0001, "step": 26390 }, { "epoch": 0.51, "learning_rate": 4.14653002866034e-05, "loss": 0.0036, "step": 26392 }, { "epoch": 0.51, "learning_rate": 4.146465332634616e-05, "loss": 0.0002, "step": 26394 }, { "epoch": 0.51, "learning_rate": 4.146400636608893e-05, "loss": 0.0001, "step": 26396 }, { "epoch": 0.51, "learning_rate": 4.14633594058317e-05, "loss": 0.0001, "step": 26398 }, { "epoch": 0.51, "learning_rate": 4.146271244557447e-05, "loss": 0.0008, "step": 26400 }, { "epoch": 0.51, "learning_rate": 4.146206548531724e-05, "loss": 0.0001, "step": 26402 }, { "epoch": 0.51, "learning_rate": 4.146141852506001e-05, "loss": 0.0001, "step": 26404 }, { "epoch": 0.51, "learning_rate": 4.146077156480278e-05, "loss": 0.0137, "step": 26406 }, { "epoch": 0.51, "learning_rate": 4.1460124604545546e-05, "loss": 0.0001, "step": 26408 }, { "epoch": 0.51, "learning_rate": 4.1459477644288315e-05, "loss": 0.0002, "step": 26410 }, { "epoch": 0.51, "learning_rate": 4.145883068403108e-05, "loss": 0.0081, "step": 26412 }, { "epoch": 0.51, "learning_rate": 4.1458183723773854e-05, "loss": 0.0002, "step": 26414 }, { "epoch": 0.51, "learning_rate": 4.1457536763516616e-05, "loss": 0.0, "step": 26416 }, { "epoch": 0.51, "learning_rate": 4.1456889803259385e-05, "loss": 0.0001, "step": 26418 }, { "epoch": 0.51, "learning_rate": 4.145624284300216e-05, "loss": 0.0, "step": 26420 }, { "epoch": 0.51, "learning_rate": 4.145559588274492e-05, "loss": 0.0001, "step": 26422 }, { "epoch": 0.51, "learning_rate": 4.145494892248769e-05, "loss": 0.0024, "step": 26424 }, { "epoch": 0.51, "learning_rate": 4.145430196223046e-05, "loss": 0.0025, "step": 26426 }, { "epoch": 0.51, "learning_rate": 4.145365500197323e-05, "loss": 0.015, "step": 26428 }, { "epoch": 0.51, "learning_rate": 4.1453008041716e-05, "loss": 0.0001, "step": 26430 }, { "epoch": 0.51, "learning_rate": 4.145236108145877e-05, "loss": 0.0003, "step": 26432 }, { "epoch": 0.51, "learning_rate": 4.145171412120154e-05, "loss": 0.0, "step": 26434 }, { "epoch": 0.51, "learning_rate": 4.145106716094431e-05, "loss": 0.0087, "step": 26436 }, { "epoch": 0.51, "learning_rate": 4.1450420200687076e-05, "loss": 0.0003, "step": 26438 }, { "epoch": 0.51, "learning_rate": 4.144977324042984e-05, "loss": 0.0011, "step": 26440 }, { "epoch": 0.51, "learning_rate": 4.1449126280172614e-05, "loss": 0.0096, "step": 26442 }, { "epoch": 0.51, "learning_rate": 4.1448479319915376e-05, "loss": 0.0009, "step": 26444 }, { "epoch": 0.51, "learning_rate": 4.144783235965815e-05, "loss": 0.0015, "step": 26446 }, { "epoch": 0.51, "learning_rate": 4.1447185399400915e-05, "loss": 0.0, "step": 26448 }, { "epoch": 0.51, "learning_rate": 4.1446538439143684e-05, "loss": 0.0466, "step": 26450 }, { "epoch": 0.51, "learning_rate": 4.144589147888645e-05, "loss": 0.0004, "step": 26452 }, { "epoch": 0.51, "learning_rate": 4.144524451862922e-05, "loss": 0.0002, "step": 26454 }, { "epoch": 0.51, "learning_rate": 4.144459755837199e-05, "loss": 0.0007, "step": 26456 }, { "epoch": 0.51, "learning_rate": 4.144395059811476e-05, "loss": 0.0037, "step": 26458 }, { "epoch": 0.51, "learning_rate": 4.144330363785753e-05, "loss": 0.0124, "step": 26460 }, { "epoch": 0.51, "learning_rate": 4.144265667760029e-05, "loss": 0.0004, "step": 26462 }, { "epoch": 0.51, "learning_rate": 4.144200971734307e-05, "loss": 0.0044, "step": 26464 }, { "epoch": 0.51, "learning_rate": 4.1441362757085836e-05, "loss": 0.0003, "step": 26466 }, { "epoch": 0.51, "learning_rate": 4.1440715796828605e-05, "loss": 0.0001, "step": 26468 }, { "epoch": 0.51, "learning_rate": 4.1440068836571374e-05, "loss": 0.0001, "step": 26470 }, { "epoch": 0.51, "learning_rate": 4.143942187631414e-05, "loss": 0.0001, "step": 26472 }, { "epoch": 0.51, "learning_rate": 4.143877491605691e-05, "loss": 0.0007, "step": 26474 }, { "epoch": 0.51, "learning_rate": 4.1438127955799675e-05, "loss": 0.0089, "step": 26476 }, { "epoch": 0.51, "learning_rate": 4.1437480995542444e-05, "loss": 0.0001, "step": 26478 }, { "epoch": 0.51, "learning_rate": 4.143683403528521e-05, "loss": 0.0001, "step": 26480 }, { "epoch": 0.51, "learning_rate": 4.143618707502798e-05, "loss": 0.0001, "step": 26482 }, { "epoch": 0.51, "learning_rate": 4.143554011477075e-05, "loss": 0.0004, "step": 26484 }, { "epoch": 0.51, "learning_rate": 4.143489315451352e-05, "loss": 0.0, "step": 26486 }, { "epoch": 0.51, "learning_rate": 4.143424619425629e-05, "loss": 0.0019, "step": 26488 }, { "epoch": 0.51, "learning_rate": 4.143359923399906e-05, "loss": 0.0001, "step": 26490 }, { "epoch": 0.51, "learning_rate": 4.143295227374183e-05, "loss": 0.0058, "step": 26492 }, { "epoch": 0.51, "learning_rate": 4.143230531348459e-05, "loss": 0.011, "step": 26494 }, { "epoch": 0.51, "learning_rate": 4.1431658353227366e-05, "loss": 0.0001, "step": 26496 }, { "epoch": 0.51, "learning_rate": 4.1431011392970135e-05, "loss": 0.0166, "step": 26498 }, { "epoch": 0.51, "learning_rate": 4.14303644327129e-05, "loss": 0.0093, "step": 26500 }, { "epoch": 0.51, "learning_rate": 4.142971747245567e-05, "loss": 0.0001, "step": 26502 }, { "epoch": 0.51, "learning_rate": 4.1429070512198435e-05, "loss": 0.0007, "step": 26504 }, { "epoch": 0.51, "learning_rate": 4.142842355194121e-05, "loss": 0.0008, "step": 26506 }, { "epoch": 0.51, "learning_rate": 4.1427776591683973e-05, "loss": 0.0001, "step": 26508 }, { "epoch": 0.51, "learning_rate": 4.142712963142674e-05, "loss": 0.0014, "step": 26510 }, { "epoch": 0.51, "learning_rate": 4.142648267116951e-05, "loss": 0.0003, "step": 26512 }, { "epoch": 0.51, "learning_rate": 4.142583571091228e-05, "loss": 0.0003, "step": 26514 }, { "epoch": 0.51, "learning_rate": 4.142518875065505e-05, "loss": 0.0062, "step": 26516 }, { "epoch": 0.51, "learning_rate": 4.142454179039782e-05, "loss": 0.0111, "step": 26518 }, { "epoch": 0.51, "learning_rate": 4.142389483014059e-05, "loss": 0.0003, "step": 26520 }, { "epoch": 0.51, "learning_rate": 4.142324786988335e-05, "loss": 0.0029, "step": 26522 }, { "epoch": 0.51, "learning_rate": 4.1422600909626126e-05, "loss": 0.0028, "step": 26524 }, { "epoch": 0.51, "learning_rate": 4.142195394936889e-05, "loss": 0.0001, "step": 26526 }, { "epoch": 0.51, "learning_rate": 4.1421306989111664e-05, "loss": 0.0003, "step": 26528 }, { "epoch": 0.51, "learning_rate": 4.142066002885443e-05, "loss": 0.0, "step": 26530 }, { "epoch": 0.51, "learning_rate": 4.1420013068597196e-05, "loss": 0.0012, "step": 26532 }, { "epoch": 0.51, "learning_rate": 4.141936610833997e-05, "loss": 0.0048, "step": 26534 }, { "epoch": 0.52, "learning_rate": 4.1418719148082734e-05, "loss": 0.0038, "step": 26536 }, { "epoch": 0.52, "learning_rate": 4.14180721878255e-05, "loss": 0.0, "step": 26538 }, { "epoch": 0.52, "learning_rate": 4.141742522756827e-05, "loss": 0.0091, "step": 26540 }, { "epoch": 0.52, "learning_rate": 4.141677826731104e-05, "loss": 0.0068, "step": 26542 }, { "epoch": 0.52, "learning_rate": 4.141613130705381e-05, "loss": 0.0071, "step": 26544 }, { "epoch": 0.52, "learning_rate": 4.141548434679658e-05, "loss": 0.0006, "step": 26546 }, { "epoch": 0.52, "learning_rate": 4.141483738653935e-05, "loss": 0.0014, "step": 26548 }, { "epoch": 0.52, "learning_rate": 4.141419042628212e-05, "loss": 0.0001, "step": 26550 }, { "epoch": 0.52, "learning_rate": 4.1413543466024887e-05, "loss": 0.0049, "step": 26552 }, { "epoch": 0.52, "learning_rate": 4.141289650576765e-05, "loss": 0.0301, "step": 26554 }, { "epoch": 0.52, "learning_rate": 4.1412249545510425e-05, "loss": 0.0, "step": 26556 }, { "epoch": 0.52, "learning_rate": 4.141160258525319e-05, "loss": 0.0001, "step": 26558 }, { "epoch": 0.52, "learning_rate": 4.1410955624995956e-05, "loss": 0.0046, "step": 26560 }, { "epoch": 0.52, "learning_rate": 4.1410308664738725e-05, "loss": 0.0129, "step": 26562 }, { "epoch": 0.52, "learning_rate": 4.1409661704481494e-05, "loss": 0.0007, "step": 26564 }, { "epoch": 0.52, "learning_rate": 4.1409014744224263e-05, "loss": 0.0001, "step": 26566 }, { "epoch": 0.52, "learning_rate": 4.140836778396703e-05, "loss": 0.0001, "step": 26568 }, { "epoch": 0.52, "learning_rate": 4.14077208237098e-05, "loss": 0.0001, "step": 26570 }, { "epoch": 0.52, "learning_rate": 4.140707386345257e-05, "loss": 0.0126, "step": 26572 }, { "epoch": 0.52, "learning_rate": 4.140642690319534e-05, "loss": 0.0001, "step": 26574 }, { "epoch": 0.52, "learning_rate": 4.140577994293811e-05, "loss": 0.0039, "step": 26576 }, { "epoch": 0.52, "learning_rate": 4.140513298268088e-05, "loss": 0.0002, "step": 26578 }, { "epoch": 0.52, "learning_rate": 4.140448602242365e-05, "loss": 0.0171, "step": 26580 }, { "epoch": 0.52, "learning_rate": 4.140383906216641e-05, "loss": 0.0, "step": 26582 }, { "epoch": 0.52, "learning_rate": 4.1403192101909185e-05, "loss": 0.0001, "step": 26584 }, { "epoch": 0.52, "learning_rate": 4.140254514165195e-05, "loss": 0.0042, "step": 26586 }, { "epoch": 0.52, "learning_rate": 4.140189818139472e-05, "loss": 0.0007, "step": 26588 }, { "epoch": 0.52, "learning_rate": 4.1401251221137486e-05, "loss": 0.0036, "step": 26590 }, { "epoch": 0.52, "learning_rate": 4.1400604260880255e-05, "loss": 0.0011, "step": 26592 }, { "epoch": 0.52, "learning_rate": 4.1399957300623024e-05, "loss": 0.0001, "step": 26594 }, { "epoch": 0.52, "learning_rate": 4.139931034036579e-05, "loss": 0.0001, "step": 26596 }, { "epoch": 0.52, "learning_rate": 4.139866338010856e-05, "loss": 0.0006, "step": 26598 }, { "epoch": 0.52, "learning_rate": 4.139801641985133e-05, "loss": 0.0011, "step": 26600 }, { "epoch": 0.52, "learning_rate": 4.13973694595941e-05, "loss": 0.0053, "step": 26602 }, { "epoch": 0.52, "learning_rate": 4.139672249933686e-05, "loss": 0.0082, "step": 26604 }, { "epoch": 0.52, "learning_rate": 4.139607553907964e-05, "loss": 0.0001, "step": 26606 }, { "epoch": 0.52, "learning_rate": 4.13954285788224e-05, "loss": 0.0055, "step": 26608 }, { "epoch": 0.52, "learning_rate": 4.1394781618565176e-05, "loss": 0.0038, "step": 26610 }, { "epoch": 0.52, "learning_rate": 4.1394134658307946e-05, "loss": 0.0003, "step": 26612 }, { "epoch": 0.52, "learning_rate": 4.139348769805071e-05, "loss": 0.0202, "step": 26614 }, { "epoch": 0.52, "learning_rate": 4.1392840737793484e-05, "loss": 0.0004, "step": 26616 }, { "epoch": 0.52, "learning_rate": 4.1392193777536246e-05, "loss": 0.0275, "step": 26618 }, { "epoch": 0.52, "learning_rate": 4.1391546817279015e-05, "loss": 0.0001, "step": 26620 }, { "epoch": 0.52, "learning_rate": 4.1390899857021784e-05, "loss": 0.0005, "step": 26622 }, { "epoch": 0.52, "learning_rate": 4.139025289676455e-05, "loss": 0.0001, "step": 26624 }, { "epoch": 0.52, "learning_rate": 4.138960593650732e-05, "loss": 0.0034, "step": 26626 }, { "epoch": 0.52, "learning_rate": 4.138895897625009e-05, "loss": 0.0005, "step": 26628 }, { "epoch": 0.52, "learning_rate": 4.138831201599286e-05, "loss": 0.0001, "step": 26630 }, { "epoch": 0.52, "learning_rate": 4.138766505573563e-05, "loss": 0.0002, "step": 26632 }, { "epoch": 0.52, "learning_rate": 4.13870180954784e-05, "loss": 0.0002, "step": 26634 }, { "epoch": 0.52, "learning_rate": 4.138637113522116e-05, "loss": 0.0001, "step": 26636 }, { "epoch": 0.52, "learning_rate": 4.138572417496394e-05, "loss": 0.0001, "step": 26638 }, { "epoch": 0.52, "learning_rate": 4.13850772147067e-05, "loss": 0.0002, "step": 26640 }, { "epoch": 0.52, "learning_rate": 4.138443025444947e-05, "loss": 0.0006, "step": 26642 }, { "epoch": 0.52, "learning_rate": 4.1383783294192244e-05, "loss": 0.0001, "step": 26644 }, { "epoch": 0.52, "learning_rate": 4.1383136333935006e-05, "loss": 0.0004, "step": 26646 }, { "epoch": 0.52, "learning_rate": 4.138248937367778e-05, "loss": 0.0003, "step": 26648 }, { "epoch": 0.52, "learning_rate": 4.1381842413420545e-05, "loss": 0.0089, "step": 26650 }, { "epoch": 0.52, "learning_rate": 4.1381195453163314e-05, "loss": 0.0038, "step": 26652 }, { "epoch": 0.52, "learning_rate": 4.138054849290608e-05, "loss": 0.0006, "step": 26654 }, { "epoch": 0.52, "learning_rate": 4.137990153264885e-05, "loss": 0.0039, "step": 26656 }, { "epoch": 0.52, "learning_rate": 4.137925457239162e-05, "loss": 0.0002, "step": 26658 }, { "epoch": 0.52, "learning_rate": 4.137860761213439e-05, "loss": 0.0015, "step": 26660 }, { "epoch": 0.52, "learning_rate": 4.137796065187716e-05, "loss": 0.0015, "step": 26662 }, { "epoch": 0.52, "learning_rate": 4.137731369161992e-05, "loss": 0.0009, "step": 26664 }, { "epoch": 0.52, "learning_rate": 4.13766667313627e-05, "loss": 0.0129, "step": 26666 }, { "epoch": 0.52, "learning_rate": 4.137601977110546e-05, "loss": 0.0056, "step": 26668 }, { "epoch": 0.52, "learning_rate": 4.1375372810848235e-05, "loss": 0.0001, "step": 26670 }, { "epoch": 0.52, "learning_rate": 4.1374725850591e-05, "loss": 0.0001, "step": 26672 }, { "epoch": 0.52, "learning_rate": 4.137407889033377e-05, "loss": 0.0026, "step": 26674 }, { "epoch": 0.52, "learning_rate": 4.137343193007654e-05, "loss": 0.0017, "step": 26676 }, { "epoch": 0.52, "learning_rate": 4.1372784969819305e-05, "loss": 0.0093, "step": 26678 }, { "epoch": 0.52, "learning_rate": 4.1372138009562074e-05, "loss": 0.0053, "step": 26680 }, { "epoch": 0.52, "learning_rate": 4.137149104930484e-05, "loss": 0.0001, "step": 26682 }, { "epoch": 0.52, "learning_rate": 4.137084408904761e-05, "loss": 0.0045, "step": 26684 }, { "epoch": 0.52, "learning_rate": 4.1370197128790375e-05, "loss": 0.0059, "step": 26686 }, { "epoch": 0.52, "learning_rate": 4.136955016853315e-05, "loss": 0.0006, "step": 26688 }, { "epoch": 0.52, "learning_rate": 4.136890320827592e-05, "loss": 0.0038, "step": 26690 }, { "epoch": 0.52, "learning_rate": 4.136825624801869e-05, "loss": 0.0005, "step": 26692 }, { "epoch": 0.52, "learning_rate": 4.136760928776146e-05, "loss": 0.0001, "step": 26694 }, { "epoch": 0.52, "learning_rate": 4.136696232750422e-05, "loss": 0.0004, "step": 26696 }, { "epoch": 0.52, "learning_rate": 4.1366315367246996e-05, "loss": 0.0002, "step": 26698 }, { "epoch": 0.52, "learning_rate": 4.136566840698976e-05, "loss": 0.0, "step": 26700 }, { "epoch": 0.52, "learning_rate": 4.136502144673253e-05, "loss": 0.0439, "step": 26702 }, { "epoch": 0.52, "learning_rate": 4.1364374486475296e-05, "loss": 0.0004, "step": 26704 }, { "epoch": 0.52, "learning_rate": 4.1363727526218065e-05, "loss": 0.0006, "step": 26706 }, { "epoch": 0.52, "learning_rate": 4.1363080565960834e-05, "loss": 0.0001, "step": 26708 }, { "epoch": 0.52, "learning_rate": 4.1362433605703604e-05, "loss": 0.001, "step": 26710 }, { "epoch": 0.52, "learning_rate": 4.136178664544637e-05, "loss": 0.0, "step": 26712 }, { "epoch": 0.52, "learning_rate": 4.136113968518914e-05, "loss": 0.0002, "step": 26714 }, { "epoch": 0.52, "learning_rate": 4.136049272493191e-05, "loss": 0.0, "step": 26716 }, { "epoch": 0.52, "learning_rate": 4.135984576467467e-05, "loss": 0.0162, "step": 26718 }, { "epoch": 0.52, "learning_rate": 4.135919880441745e-05, "loss": 0.0001, "step": 26720 }, { "epoch": 0.52, "learning_rate": 4.135855184416022e-05, "loss": 0.0, "step": 26722 }, { "epoch": 0.52, "learning_rate": 4.135790488390298e-05, "loss": 0.0001, "step": 26724 }, { "epoch": 0.52, "learning_rate": 4.1357257923645756e-05, "loss": 0.0004, "step": 26726 }, { "epoch": 0.52, "learning_rate": 4.135661096338852e-05, "loss": 0.0001, "step": 26728 }, { "epoch": 0.52, "learning_rate": 4.1355964003131294e-05, "loss": 0.0001, "step": 26730 }, { "epoch": 0.52, "learning_rate": 4.135531704287406e-05, "loss": 0.0001, "step": 26732 }, { "epoch": 0.52, "learning_rate": 4.1354670082616826e-05, "loss": 0.001, "step": 26734 }, { "epoch": 0.52, "learning_rate": 4.1354023122359595e-05, "loss": 0.0002, "step": 26736 }, { "epoch": 0.52, "learning_rate": 4.1353376162102364e-05, "loss": 0.0, "step": 26738 }, { "epoch": 0.52, "learning_rate": 4.135272920184513e-05, "loss": 0.0002, "step": 26740 }, { "epoch": 0.52, "learning_rate": 4.13520822415879e-05, "loss": 0.0022, "step": 26742 }, { "epoch": 0.52, "learning_rate": 4.135143528133067e-05, "loss": 0.0001, "step": 26744 }, { "epoch": 0.52, "learning_rate": 4.1350788321073434e-05, "loss": 0.0001, "step": 26746 }, { "epoch": 0.52, "learning_rate": 4.135014136081621e-05, "loss": 0.0001, "step": 26748 }, { "epoch": 0.52, "learning_rate": 4.134949440055897e-05, "loss": 0.0, "step": 26750 }, { "epoch": 0.52, "learning_rate": 4.134884744030175e-05, "loss": 0.0075, "step": 26752 }, { "epoch": 0.52, "learning_rate": 4.134820048004452e-05, "loss": 0.0, "step": 26754 }, { "epoch": 0.52, "learning_rate": 4.134755351978728e-05, "loss": 0.0001, "step": 26756 }, { "epoch": 0.52, "learning_rate": 4.1346906559530055e-05, "loss": 0.0027, "step": 26758 }, { "epoch": 0.52, "learning_rate": 4.134625959927282e-05, "loss": 0.0011, "step": 26760 }, { "epoch": 0.52, "learning_rate": 4.1345612639015586e-05, "loss": 0.0081, "step": 26762 }, { "epoch": 0.52, "learning_rate": 4.1344965678758355e-05, "loss": 0.0063, "step": 26764 }, { "epoch": 0.52, "learning_rate": 4.1344318718501124e-05, "loss": 0.0001, "step": 26766 }, { "epoch": 0.52, "learning_rate": 4.1343671758243893e-05, "loss": 0.0002, "step": 26768 }, { "epoch": 0.52, "learning_rate": 4.134302479798666e-05, "loss": 0.0, "step": 26770 }, { "epoch": 0.52, "learning_rate": 4.134237783772943e-05, "loss": 0.0009, "step": 26772 }, { "epoch": 0.52, "learning_rate": 4.13417308774722e-05, "loss": 0.0009, "step": 26774 }, { "epoch": 0.52, "learning_rate": 4.134108391721497e-05, "loss": 0.0, "step": 26776 }, { "epoch": 0.52, "learning_rate": 4.134043695695773e-05, "loss": 0.0239, "step": 26778 }, { "epoch": 0.52, "learning_rate": 4.133978999670051e-05, "loss": 0.0245, "step": 26780 }, { "epoch": 0.52, "learning_rate": 4.133914303644327e-05, "loss": 0.0001, "step": 26782 }, { "epoch": 0.52, "learning_rate": 4.133849607618604e-05, "loss": 0.0033, "step": 26784 }, { "epoch": 0.52, "learning_rate": 4.133784911592881e-05, "loss": 0.0002, "step": 26786 }, { "epoch": 0.52, "learning_rate": 4.133720215567158e-05, "loss": 0.0005, "step": 26788 }, { "epoch": 0.52, "learning_rate": 4.1336555195414353e-05, "loss": 0.0048, "step": 26790 }, { "epoch": 0.52, "learning_rate": 4.1335908235157116e-05, "loss": 0.0035, "step": 26792 }, { "epoch": 0.52, "learning_rate": 4.1335261274899885e-05, "loss": 0.0276, "step": 26794 }, { "epoch": 0.52, "learning_rate": 4.1334614314642654e-05, "loss": 0.0016, "step": 26796 }, { "epoch": 0.52, "learning_rate": 4.133396735438542e-05, "loss": 0.0006, "step": 26798 }, { "epoch": 0.52, "learning_rate": 4.133332039412819e-05, "loss": 0.0013, "step": 26800 }, { "epoch": 0.52, "learning_rate": 4.133267343387096e-05, "loss": 0.0, "step": 26802 }, { "epoch": 0.52, "learning_rate": 4.133202647361373e-05, "loss": 0.0002, "step": 26804 }, { "epoch": 0.52, "learning_rate": 4.133137951335649e-05, "loss": 0.0005, "step": 26806 }, { "epoch": 0.52, "learning_rate": 4.133073255309927e-05, "loss": 0.0286, "step": 26808 }, { "epoch": 0.52, "learning_rate": 4.133008559284203e-05, "loss": 0.0141, "step": 26810 }, { "epoch": 0.52, "learning_rate": 4.1329438632584807e-05, "loss": 0.0, "step": 26812 }, { "epoch": 0.52, "learning_rate": 4.132879167232757e-05, "loss": 0.0032, "step": 26814 }, { "epoch": 0.52, "learning_rate": 4.132814471207034e-05, "loss": 0.0013, "step": 26816 }, { "epoch": 0.52, "learning_rate": 4.132749775181311e-05, "loss": 0.0006, "step": 26818 }, { "epoch": 0.52, "learning_rate": 4.1326850791555876e-05, "loss": 0.0001, "step": 26820 }, { "epoch": 0.52, "learning_rate": 4.1326203831298645e-05, "loss": 0.0001, "step": 26822 }, { "epoch": 0.52, "learning_rate": 4.1325556871041414e-05, "loss": 0.0007, "step": 26824 }, { "epoch": 0.52, "learning_rate": 4.132490991078418e-05, "loss": 0.0003, "step": 26826 }, { "epoch": 0.52, "learning_rate": 4.1324262950526946e-05, "loss": 0.0001, "step": 26828 }, { "epoch": 0.52, "learning_rate": 4.132361599026972e-05, "loss": 0.0002, "step": 26830 }, { "epoch": 0.52, "learning_rate": 4.1322969030012484e-05, "loss": 0.0019, "step": 26832 }, { "epoch": 0.52, "learning_rate": 4.132232206975526e-05, "loss": 0.0001, "step": 26834 }, { "epoch": 0.52, "learning_rate": 4.132167510949803e-05, "loss": 0.0001, "step": 26836 }, { "epoch": 0.52, "learning_rate": 4.132102814924079e-05, "loss": 0.0058, "step": 26838 }, { "epoch": 0.52, "learning_rate": 4.132038118898357e-05, "loss": 0.0001, "step": 26840 }, { "epoch": 0.52, "learning_rate": 4.131973422872633e-05, "loss": 0.0002, "step": 26842 }, { "epoch": 0.52, "learning_rate": 4.13190872684691e-05, "loss": 0.0099, "step": 26844 }, { "epoch": 0.52, "learning_rate": 4.131844030821187e-05, "loss": 0.0002, "step": 26846 }, { "epoch": 0.52, "learning_rate": 4.1317793347954637e-05, "loss": 0.0152, "step": 26848 }, { "epoch": 0.52, "learning_rate": 4.1317146387697406e-05, "loss": 0.0001, "step": 26850 }, { "epoch": 0.52, "learning_rate": 4.1316499427440175e-05, "loss": 0.0008, "step": 26852 }, { "epoch": 0.52, "learning_rate": 4.1315852467182944e-05, "loss": 0.0006, "step": 26854 }, { "epoch": 0.52, "learning_rate": 4.131520550692571e-05, "loss": 0.0004, "step": 26856 }, { "epoch": 0.52, "learning_rate": 4.131455854666848e-05, "loss": 0.0001, "step": 26858 }, { "epoch": 0.52, "learning_rate": 4.1313911586411244e-05, "loss": 0.0, "step": 26860 }, { "epoch": 0.52, "learning_rate": 4.131326462615402e-05, "loss": 0.0, "step": 26862 }, { "epoch": 0.52, "learning_rate": 4.131261766589678e-05, "loss": 0.0003, "step": 26864 }, { "epoch": 0.52, "learning_rate": 4.131197070563955e-05, "loss": 0.0001, "step": 26866 }, { "epoch": 0.52, "learning_rate": 4.131132374538233e-05, "loss": 0.0001, "step": 26868 }, { "epoch": 0.52, "learning_rate": 4.131067678512509e-05, "loss": 0.0008, "step": 26870 }, { "epoch": 0.52, "learning_rate": 4.1310029824867866e-05, "loss": 0.0001, "step": 26872 }, { "epoch": 0.52, "learning_rate": 4.130938286461063e-05, "loss": 0.0001, "step": 26874 }, { "epoch": 0.52, "learning_rate": 4.13087359043534e-05, "loss": 0.0001, "step": 26876 }, { "epoch": 0.52, "learning_rate": 4.1308088944096166e-05, "loss": 0.0001, "step": 26878 }, { "epoch": 0.52, "learning_rate": 4.1307441983838935e-05, "loss": 0.0004, "step": 26880 }, { "epoch": 0.52, "learning_rate": 4.1306795023581704e-05, "loss": 0.0022, "step": 26882 }, { "epoch": 0.52, "learning_rate": 4.130614806332447e-05, "loss": 0.0, "step": 26884 }, { "epoch": 0.52, "learning_rate": 4.130550110306724e-05, "loss": 0.0005, "step": 26886 }, { "epoch": 0.52, "learning_rate": 4.1304854142810005e-05, "loss": 0.0012, "step": 26888 }, { "epoch": 0.52, "learning_rate": 4.130420718255278e-05, "loss": 0.0001, "step": 26890 }, { "epoch": 0.52, "learning_rate": 4.130356022229554e-05, "loss": 0.0016, "step": 26892 }, { "epoch": 0.52, "learning_rate": 4.130291326203832e-05, "loss": 0.0057, "step": 26894 }, { "epoch": 0.52, "learning_rate": 4.130226630178108e-05, "loss": 0.0223, "step": 26896 }, { "epoch": 0.52, "learning_rate": 4.130161934152385e-05, "loss": 0.0001, "step": 26898 }, { "epoch": 0.52, "learning_rate": 4.1300972381266626e-05, "loss": 0.0113, "step": 26900 }, { "epoch": 0.52, "learning_rate": 4.130032542100939e-05, "loss": 0.0003, "step": 26902 }, { "epoch": 0.52, "learning_rate": 4.129967846075216e-05, "loss": 0.0, "step": 26904 }, { "epoch": 0.52, "learning_rate": 4.1299031500494926e-05, "loss": 0.0001, "step": 26906 }, { "epoch": 0.52, "learning_rate": 4.1298384540237696e-05, "loss": 0.0031, "step": 26908 }, { "epoch": 0.52, "learning_rate": 4.129773757998046e-05, "loss": 0.0044, "step": 26910 }, { "epoch": 0.52, "learning_rate": 4.1297090619723234e-05, "loss": 0.0, "step": 26912 }, { "epoch": 0.52, "learning_rate": 4.1296443659466e-05, "loss": 0.0025, "step": 26914 }, { "epoch": 0.52, "learning_rate": 4.129579669920877e-05, "loss": 0.0002, "step": 26916 }, { "epoch": 0.52, "learning_rate": 4.129514973895154e-05, "loss": 0.0001, "step": 26918 }, { "epoch": 0.52, "learning_rate": 4.12945027786943e-05, "loss": 0.0001, "step": 26920 }, { "epoch": 0.52, "learning_rate": 4.129385581843708e-05, "loss": 0.0002, "step": 26922 }, { "epoch": 0.52, "learning_rate": 4.129320885817984e-05, "loss": 0.0041, "step": 26924 }, { "epoch": 0.52, "learning_rate": 4.129256189792261e-05, "loss": 0.007, "step": 26926 }, { "epoch": 0.52, "learning_rate": 4.129191493766538e-05, "loss": 0.0, "step": 26928 }, { "epoch": 0.52, "learning_rate": 4.129126797740815e-05, "loss": 0.0001, "step": 26930 }, { "epoch": 0.52, "learning_rate": 4.1290621017150925e-05, "loss": 0.0016, "step": 26932 }, { "epoch": 0.52, "learning_rate": 4.128997405689369e-05, "loss": 0.0002, "step": 26934 }, { "epoch": 0.52, "learning_rate": 4.1289327096636456e-05, "loss": 0.0593, "step": 26936 }, { "epoch": 0.52, "learning_rate": 4.1288680136379225e-05, "loss": 0.0063, "step": 26938 }, { "epoch": 0.52, "learning_rate": 4.1288033176121994e-05, "loss": 0.0029, "step": 26940 }, { "epoch": 0.52, "learning_rate": 4.1287386215864756e-05, "loss": 0.0001, "step": 26942 }, { "epoch": 0.52, "learning_rate": 4.128673925560753e-05, "loss": 0.0001, "step": 26944 }, { "epoch": 0.52, "learning_rate": 4.12860922953503e-05, "loss": 0.0001, "step": 26946 }, { "epoch": 0.52, "learning_rate": 4.1285445335093064e-05, "loss": 0.0164, "step": 26948 }, { "epoch": 0.52, "learning_rate": 4.128479837483584e-05, "loss": 0.0003, "step": 26950 }, { "epoch": 0.52, "learning_rate": 4.12841514145786e-05, "loss": 0.0056, "step": 26952 }, { "epoch": 0.52, "learning_rate": 4.128350445432138e-05, "loss": 0.0, "step": 26954 }, { "epoch": 0.52, "learning_rate": 4.128285749406414e-05, "loss": 0.0, "step": 26956 }, { "epoch": 0.52, "learning_rate": 4.128221053380691e-05, "loss": 0.0001, "step": 26958 }, { "epoch": 0.52, "learning_rate": 4.128156357354968e-05, "loss": 0.0, "step": 26960 }, { "epoch": 0.52, "learning_rate": 4.128091661329245e-05, "loss": 0.0004, "step": 26962 }, { "epoch": 0.52, "learning_rate": 4.1280269653035216e-05, "loss": 0.0009, "step": 26964 }, { "epoch": 0.52, "learning_rate": 4.1279622692777985e-05, "loss": 0.0009, "step": 26966 }, { "epoch": 0.52, "learning_rate": 4.1278975732520754e-05, "loss": 0.0036, "step": 26968 }, { "epoch": 0.52, "learning_rate": 4.127832877226352e-05, "loss": 0.0055, "step": 26970 }, { "epoch": 0.52, "learning_rate": 4.127768181200629e-05, "loss": 0.0001, "step": 26972 }, { "epoch": 0.52, "learning_rate": 4.1277034851749055e-05, "loss": 0.0054, "step": 26974 }, { "epoch": 0.52, "learning_rate": 4.127638789149183e-05, "loss": 0.0001, "step": 26976 }, { "epoch": 0.52, "learning_rate": 4.12757409312346e-05, "loss": 0.0026, "step": 26978 }, { "epoch": 0.52, "learning_rate": 4.127509397097736e-05, "loss": 0.0063, "step": 26980 }, { "epoch": 0.52, "learning_rate": 4.127444701072014e-05, "loss": 0.0005, "step": 26982 }, { "epoch": 0.52, "learning_rate": 4.12738000504629e-05, "loss": 0.0001, "step": 26984 }, { "epoch": 0.52, "learning_rate": 4.127315309020567e-05, "loss": 0.0021, "step": 26986 }, { "epoch": 0.52, "learning_rate": 4.127250612994844e-05, "loss": 0.0002, "step": 26988 }, { "epoch": 0.52, "learning_rate": 4.127185916969121e-05, "loss": 0.0003, "step": 26990 }, { "epoch": 0.52, "learning_rate": 4.127121220943398e-05, "loss": 0.0022, "step": 26992 }, { "epoch": 0.52, "learning_rate": 4.1270565249176746e-05, "loss": 0.0001, "step": 26994 }, { "epoch": 0.52, "learning_rate": 4.1269918288919515e-05, "loss": 0.0001, "step": 26996 }, { "epoch": 0.52, "learning_rate": 4.1269271328662284e-05, "loss": 0.0001, "step": 26998 }, { "epoch": 0.52, "learning_rate": 4.126862436840505e-05, "loss": 0.0208, "step": 27000 }, { "epoch": 0.52, "learning_rate": 4.1267977408147815e-05, "loss": 0.0006, "step": 27002 }, { "epoch": 0.52, "learning_rate": 4.126733044789059e-05, "loss": 0.0001, "step": 27004 }, { "epoch": 0.52, "learning_rate": 4.1266683487633354e-05, "loss": 0.0, "step": 27006 }, { "epoch": 0.52, "learning_rate": 4.126603652737612e-05, "loss": 0.0023, "step": 27008 }, { "epoch": 0.52, "learning_rate": 4.126538956711889e-05, "loss": 0.0001, "step": 27010 }, { "epoch": 0.52, "learning_rate": 4.126474260686166e-05, "loss": 0.0009, "step": 27012 }, { "epoch": 0.52, "learning_rate": 4.126409564660444e-05, "loss": 0.0101, "step": 27014 }, { "epoch": 0.52, "learning_rate": 4.12634486863472e-05, "loss": 0.0018, "step": 27016 }, { "epoch": 0.52, "learning_rate": 4.126280172608997e-05, "loss": 0.0001, "step": 27018 }, { "epoch": 0.52, "learning_rate": 4.126215476583274e-05, "loss": 0.0, "step": 27020 }, { "epoch": 0.52, "learning_rate": 4.1261507805575506e-05, "loss": 0.0139, "step": 27022 }, { "epoch": 0.52, "learning_rate": 4.1260860845318275e-05, "loss": 0.0001, "step": 27024 }, { "epoch": 0.52, "learning_rate": 4.1260213885061044e-05, "loss": 0.0181, "step": 27026 }, { "epoch": 0.52, "learning_rate": 4.1259566924803813e-05, "loss": 0.0001, "step": 27028 }, { "epoch": 0.52, "learning_rate": 4.1258919964546576e-05, "loss": 0.0001, "step": 27030 }, { "epoch": 0.52, "learning_rate": 4.125827300428935e-05, "loss": 0.0003, "step": 27032 }, { "epoch": 0.52, "learning_rate": 4.1257626044032114e-05, "loss": 0.0, "step": 27034 }, { "epoch": 0.52, "learning_rate": 4.125697908377489e-05, "loss": 0.0032, "step": 27036 }, { "epoch": 0.52, "learning_rate": 4.125633212351765e-05, "loss": 0.0001, "step": 27038 }, { "epoch": 0.52, "learning_rate": 4.125568516326042e-05, "loss": 0.0, "step": 27040 }, { "epoch": 0.52, "learning_rate": 4.125503820300319e-05, "loss": 0.0002, "step": 27042 }, { "epoch": 0.52, "learning_rate": 4.125439124274596e-05, "loss": 0.0001, "step": 27044 }, { "epoch": 0.52, "learning_rate": 4.125374428248873e-05, "loss": 0.0, "step": 27046 }, { "epoch": 0.52, "learning_rate": 4.12530973222315e-05, "loss": 0.0005, "step": 27048 }, { "epoch": 0.53, "learning_rate": 4.1252450361974267e-05, "loss": 0.0001, "step": 27050 }, { "epoch": 0.53, "learning_rate": 4.125180340171703e-05, "loss": 0.0005, "step": 27052 }, { "epoch": 0.53, "learning_rate": 4.1251156441459805e-05, "loss": 0.0057, "step": 27054 }, { "epoch": 0.53, "learning_rate": 4.1250509481202574e-05, "loss": 0.0, "step": 27056 }, { "epoch": 0.53, "learning_rate": 4.124986252094534e-05, "loss": 0.0045, "step": 27058 }, { "epoch": 0.53, "learning_rate": 4.124921556068811e-05, "loss": 0.0339, "step": 27060 }, { "epoch": 0.53, "learning_rate": 4.1248568600430874e-05, "loss": 0.0001, "step": 27062 }, { "epoch": 0.53, "learning_rate": 4.124792164017365e-05, "loss": 0.0001, "step": 27064 }, { "epoch": 0.53, "learning_rate": 4.124727467991641e-05, "loss": 0.0001, "step": 27066 }, { "epoch": 0.53, "learning_rate": 4.124662771965918e-05, "loss": 0.0001, "step": 27068 }, { "epoch": 0.53, "learning_rate": 4.124598075940195e-05, "loss": 0.0143, "step": 27070 }, { "epoch": 0.53, "learning_rate": 4.124533379914472e-05, "loss": 0.0013, "step": 27072 }, { "epoch": 0.53, "learning_rate": 4.124468683888749e-05, "loss": 0.0003, "step": 27074 }, { "epoch": 0.53, "learning_rate": 4.124403987863026e-05, "loss": 0.0736, "step": 27076 }, { "epoch": 0.53, "learning_rate": 4.124339291837303e-05, "loss": 0.005, "step": 27078 }, { "epoch": 0.53, "learning_rate": 4.1242745958115796e-05, "loss": 0.0001, "step": 27080 }, { "epoch": 0.53, "learning_rate": 4.1242098997858565e-05, "loss": 0.0, "step": 27082 }, { "epoch": 0.53, "learning_rate": 4.124145203760133e-05, "loss": 0.0001, "step": 27084 }, { "epoch": 0.53, "learning_rate": 4.12408050773441e-05, "loss": 0.0053, "step": 27086 }, { "epoch": 0.53, "learning_rate": 4.1240158117086866e-05, "loss": 0.0003, "step": 27088 }, { "epoch": 0.53, "learning_rate": 4.1239511156829635e-05, "loss": 0.0001, "step": 27090 }, { "epoch": 0.53, "learning_rate": 4.123886419657241e-05, "loss": 0.0045, "step": 27092 }, { "epoch": 0.53, "learning_rate": 4.123821723631517e-05, "loss": 0.0, "step": 27094 }, { "epoch": 0.53, "learning_rate": 4.123757027605795e-05, "loss": 0.0001, "step": 27096 }, { "epoch": 0.53, "learning_rate": 4.123692331580071e-05, "loss": 0.0003, "step": 27098 }, { "epoch": 0.53, "learning_rate": 4.123627635554348e-05, "loss": 0.0241, "step": 27100 }, { "epoch": 0.53, "learning_rate": 4.123562939528625e-05, "loss": 0.0028, "step": 27102 }, { "epoch": 0.53, "learning_rate": 4.123498243502902e-05, "loss": 0.0, "step": 27104 }, { "epoch": 0.53, "learning_rate": 4.123433547477179e-05, "loss": 0.0041, "step": 27106 }, { "epoch": 0.53, "learning_rate": 4.1233688514514557e-05, "loss": 0.0004, "step": 27108 }, { "epoch": 0.53, "learning_rate": 4.1233041554257326e-05, "loss": 0.0001, "step": 27110 }, { "epoch": 0.53, "learning_rate": 4.123239459400009e-05, "loss": 0.0005, "step": 27112 }, { "epoch": 0.53, "learning_rate": 4.1231747633742864e-05, "loss": 0.0007, "step": 27114 }, { "epoch": 0.53, "learning_rate": 4.1231100673485626e-05, "loss": 0.0066, "step": 27116 }, { "epoch": 0.53, "learning_rate": 4.12304537132284e-05, "loss": 0.0, "step": 27118 }, { "epoch": 0.53, "learning_rate": 4.1229806752971164e-05, "loss": 0.0008, "step": 27120 }, { "epoch": 0.53, "learning_rate": 4.122915979271393e-05, "loss": 0.0001, "step": 27122 }, { "epoch": 0.53, "learning_rate": 4.122851283245671e-05, "loss": 0.0, "step": 27124 }, { "epoch": 0.53, "learning_rate": 4.122786587219947e-05, "loss": 0.0101, "step": 27126 }, { "epoch": 0.53, "learning_rate": 4.122721891194224e-05, "loss": 0.0076, "step": 27128 }, { "epoch": 0.53, "learning_rate": 4.122657195168501e-05, "loss": 0.0056, "step": 27130 }, { "epoch": 0.53, "learning_rate": 4.122592499142778e-05, "loss": 0.0028, "step": 27132 }, { "epoch": 0.53, "learning_rate": 4.122527803117054e-05, "loss": 0.0029, "step": 27134 }, { "epoch": 0.53, "learning_rate": 4.122463107091332e-05, "loss": 0.0002, "step": 27136 }, { "epoch": 0.53, "learning_rate": 4.1223984110656086e-05, "loss": 0.0002, "step": 27138 }, { "epoch": 0.53, "learning_rate": 4.1223337150398855e-05, "loss": 0.0073, "step": 27140 }, { "epoch": 0.53, "learning_rate": 4.1222690190141624e-05, "loss": 0.0, "step": 27142 }, { "epoch": 0.53, "learning_rate": 4.1222043229884386e-05, "loss": 0.0001, "step": 27144 }, { "epoch": 0.53, "learning_rate": 4.122139626962716e-05, "loss": 0.0, "step": 27146 }, { "epoch": 0.53, "learning_rate": 4.1220749309369925e-05, "loss": 0.0076, "step": 27148 }, { "epoch": 0.53, "learning_rate": 4.1220102349112694e-05, "loss": 0.0027, "step": 27150 }, { "epoch": 0.53, "learning_rate": 4.121945538885546e-05, "loss": 0.0006, "step": 27152 }, { "epoch": 0.53, "learning_rate": 4.121880842859823e-05, "loss": 0.0003, "step": 27154 }, { "epoch": 0.53, "learning_rate": 4.121816146834101e-05, "loss": 0.0006, "step": 27156 }, { "epoch": 0.53, "learning_rate": 4.121751450808377e-05, "loss": 0.0168, "step": 27158 }, { "epoch": 0.53, "learning_rate": 4.121686754782654e-05, "loss": 0.0006, "step": 27160 }, { "epoch": 0.53, "learning_rate": 4.121622058756931e-05, "loss": 0.0002, "step": 27162 }, { "epoch": 0.53, "learning_rate": 4.121557362731208e-05, "loss": 0.0, "step": 27164 }, { "epoch": 0.53, "learning_rate": 4.121492666705484e-05, "loss": 0.0, "step": 27166 }, { "epoch": 0.53, "learning_rate": 4.1214279706797615e-05, "loss": 0.0002, "step": 27168 }, { "epoch": 0.53, "learning_rate": 4.1213632746540385e-05, "loss": 0.0035, "step": 27170 }, { "epoch": 0.53, "learning_rate": 4.121298578628315e-05, "loss": 0.0002, "step": 27172 }, { "epoch": 0.53, "learning_rate": 4.121233882602592e-05, "loss": 0.0001, "step": 27174 }, { "epoch": 0.53, "learning_rate": 4.1211691865768685e-05, "loss": 0.0013, "step": 27176 }, { "epoch": 0.53, "learning_rate": 4.121104490551146e-05, "loss": 0.0016, "step": 27178 }, { "epoch": 0.53, "learning_rate": 4.121039794525422e-05, "loss": 0.0005, "step": 27180 }, { "epoch": 0.53, "learning_rate": 4.120975098499699e-05, "loss": 0.0002, "step": 27182 }, { "epoch": 0.53, "learning_rate": 4.120910402473976e-05, "loss": 0.0005, "step": 27184 }, { "epoch": 0.53, "learning_rate": 4.120845706448253e-05, "loss": 0.0, "step": 27186 }, { "epoch": 0.53, "learning_rate": 4.12078101042253e-05, "loss": 0.0107, "step": 27188 }, { "epoch": 0.53, "learning_rate": 4.120716314396807e-05, "loss": 0.0001, "step": 27190 }, { "epoch": 0.53, "learning_rate": 4.120651618371084e-05, "loss": 0.0001, "step": 27192 }, { "epoch": 0.53, "learning_rate": 4.12058692234536e-05, "loss": 0.0008, "step": 27194 }, { "epoch": 0.53, "learning_rate": 4.1205222263196376e-05, "loss": 0.0162, "step": 27196 }, { "epoch": 0.53, "learning_rate": 4.120457530293914e-05, "loss": 0.0029, "step": 27198 }, { "epoch": 0.53, "learning_rate": 4.1203928342681914e-05, "loss": 0.0, "step": 27200 }, { "epoch": 0.53, "learning_rate": 4.120328138242468e-05, "loss": 0.0065, "step": 27202 }, { "epoch": 0.53, "learning_rate": 4.1202634422167445e-05, "loss": 0.0006, "step": 27204 }, { "epoch": 0.53, "learning_rate": 4.120198746191022e-05, "loss": 0.0002, "step": 27206 }, { "epoch": 0.53, "learning_rate": 4.1201340501652984e-05, "loss": 0.0038, "step": 27208 }, { "epoch": 0.53, "learning_rate": 4.120069354139575e-05, "loss": 0.0066, "step": 27210 }, { "epoch": 0.53, "learning_rate": 4.120004658113852e-05, "loss": 0.0, "step": 27212 }, { "epoch": 0.53, "learning_rate": 4.119939962088129e-05, "loss": 0.0192, "step": 27214 }, { "epoch": 0.53, "learning_rate": 4.119875266062406e-05, "loss": 0.0002, "step": 27216 }, { "epoch": 0.53, "learning_rate": 4.119810570036683e-05, "loss": 0.0056, "step": 27218 }, { "epoch": 0.53, "learning_rate": 4.11974587401096e-05, "loss": 0.0322, "step": 27220 }, { "epoch": 0.53, "learning_rate": 4.119681177985237e-05, "loss": 0.0001, "step": 27222 }, { "epoch": 0.53, "learning_rate": 4.1196164819595136e-05, "loss": 0.0, "step": 27224 }, { "epoch": 0.53, "learning_rate": 4.11955178593379e-05, "loss": 0.0, "step": 27226 }, { "epoch": 0.53, "learning_rate": 4.1194870899080674e-05, "loss": 0.0051, "step": 27228 }, { "epoch": 0.53, "learning_rate": 4.119422393882344e-05, "loss": 0.02, "step": 27230 }, { "epoch": 0.53, "learning_rate": 4.1193576978566206e-05, "loss": 0.0, "step": 27232 }, { "epoch": 0.53, "learning_rate": 4.119293001830898e-05, "loss": 0.0002, "step": 27234 }, { "epoch": 0.53, "learning_rate": 4.1192283058051744e-05, "loss": 0.0032, "step": 27236 }, { "epoch": 0.53, "learning_rate": 4.119163609779452e-05, "loss": 0.0037, "step": 27238 }, { "epoch": 0.53, "learning_rate": 4.119098913753728e-05, "loss": 0.0, "step": 27240 }, { "epoch": 0.53, "learning_rate": 4.119034217728005e-05, "loss": 0.0004, "step": 27242 }, { "epoch": 0.53, "learning_rate": 4.118969521702282e-05, "loss": 0.0018, "step": 27244 }, { "epoch": 0.53, "learning_rate": 4.118904825676559e-05, "loss": 0.0, "step": 27246 }, { "epoch": 0.53, "learning_rate": 4.118840129650836e-05, "loss": 0.0002, "step": 27248 }, { "epoch": 0.53, "learning_rate": 4.118775433625113e-05, "loss": 0.0001, "step": 27250 }, { "epoch": 0.53, "learning_rate": 4.11871073759939e-05, "loss": 0.0003, "step": 27252 }, { "epoch": 0.53, "learning_rate": 4.118646041573666e-05, "loss": 0.0001, "step": 27254 }, { "epoch": 0.53, "learning_rate": 4.1185813455479435e-05, "loss": 0.0033, "step": 27256 }, { "epoch": 0.53, "learning_rate": 4.11851664952222e-05, "loss": 0.0001, "step": 27258 }, { "epoch": 0.53, "learning_rate": 4.118451953496497e-05, "loss": 0.0072, "step": 27260 }, { "epoch": 0.53, "learning_rate": 4.1183872574707735e-05, "loss": 0.0, "step": 27262 }, { "epoch": 0.53, "learning_rate": 4.1183225614450504e-05, "loss": 0.0, "step": 27264 }, { "epoch": 0.53, "learning_rate": 4.1182578654193274e-05, "loss": 0.0006, "step": 27266 }, { "epoch": 0.53, "learning_rate": 4.118193169393604e-05, "loss": 0.0, "step": 27268 }, { "epoch": 0.53, "learning_rate": 4.118128473367881e-05, "loss": 0.0, "step": 27270 }, { "epoch": 0.53, "learning_rate": 4.118063777342158e-05, "loss": 0.0094, "step": 27272 }, { "epoch": 0.53, "learning_rate": 4.117999081316435e-05, "loss": 0.0, "step": 27274 }, { "epoch": 0.53, "learning_rate": 4.117934385290711e-05, "loss": 0.0, "step": 27276 }, { "epoch": 0.53, "learning_rate": 4.117869689264989e-05, "loss": 0.0094, "step": 27278 }, { "epoch": 0.53, "learning_rate": 4.117804993239266e-05, "loss": 0.0081, "step": 27280 }, { "epoch": 0.53, "learning_rate": 4.1177402972135426e-05, "loss": 0.0032, "step": 27282 }, { "epoch": 0.53, "learning_rate": 4.1176756011878195e-05, "loss": 0.003, "step": 27284 }, { "epoch": 0.53, "learning_rate": 4.117610905162096e-05, "loss": 0.0002, "step": 27286 }, { "epoch": 0.53, "learning_rate": 4.1175462091363733e-05, "loss": 0.0012, "step": 27288 }, { "epoch": 0.53, "learning_rate": 4.1174815131106496e-05, "loss": 0.0001, "step": 27290 }, { "epoch": 0.53, "learning_rate": 4.1174168170849265e-05, "loss": 0.0001, "step": 27292 }, { "epoch": 0.53, "learning_rate": 4.1173521210592034e-05, "loss": 0.0, "step": 27294 }, { "epoch": 0.53, "learning_rate": 4.11728742503348e-05, "loss": 0.0027, "step": 27296 }, { "epoch": 0.53, "learning_rate": 4.117222729007757e-05, "loss": 0.0001, "step": 27298 }, { "epoch": 0.53, "learning_rate": 4.117158032982034e-05, "loss": 0.0001, "step": 27300 }, { "epoch": 0.53, "learning_rate": 4.117093336956311e-05, "loss": 0.0084, "step": 27302 }, { "epoch": 0.53, "learning_rate": 4.117028640930588e-05, "loss": 0.0001, "step": 27304 }, { "epoch": 0.53, "learning_rate": 4.116963944904865e-05, "loss": 0.0001, "step": 27306 }, { "epoch": 0.53, "learning_rate": 4.116899248879141e-05, "loss": 0.0001, "step": 27308 }, { "epoch": 0.53, "learning_rate": 4.1168345528534187e-05, "loss": 0.0015, "step": 27310 }, { "epoch": 0.53, "learning_rate": 4.116769856827695e-05, "loss": 0.0139, "step": 27312 }, { "epoch": 0.53, "learning_rate": 4.116705160801972e-05, "loss": 0.0023, "step": 27314 }, { "epoch": 0.53, "learning_rate": 4.1166404647762494e-05, "loss": 0.0, "step": 27316 }, { "epoch": 0.53, "learning_rate": 4.1165757687505256e-05, "loss": 0.0001, "step": 27318 }, { "epoch": 0.53, "learning_rate": 4.116511072724803e-05, "loss": 0.0053, "step": 27320 }, { "epoch": 0.53, "learning_rate": 4.1164463766990794e-05, "loss": 0.0001, "step": 27322 }, { "epoch": 0.53, "learning_rate": 4.1163816806733563e-05, "loss": 0.0028, "step": 27324 }, { "epoch": 0.53, "learning_rate": 4.116316984647633e-05, "loss": 0.0, "step": 27326 }, { "epoch": 0.53, "learning_rate": 4.11625228862191e-05, "loss": 0.0001, "step": 27328 }, { "epoch": 0.53, "learning_rate": 4.116187592596187e-05, "loss": 0.0002, "step": 27330 }, { "epoch": 0.53, "learning_rate": 4.116122896570464e-05, "loss": 0.0006, "step": 27332 }, { "epoch": 0.53, "learning_rate": 4.116058200544741e-05, "loss": 0.0017, "step": 27334 }, { "epoch": 0.53, "learning_rate": 4.115993504519017e-05, "loss": 0.0032, "step": 27336 }, { "epoch": 0.53, "learning_rate": 4.115928808493295e-05, "loss": 0.0, "step": 27338 }, { "epoch": 0.53, "learning_rate": 4.115864112467571e-05, "loss": 0.0001, "step": 27340 }, { "epoch": 0.53, "learning_rate": 4.1157994164418485e-05, "loss": 0.0004, "step": 27342 }, { "epoch": 0.53, "learning_rate": 4.115734720416125e-05, "loss": 0.0, "step": 27344 }, { "epoch": 0.53, "learning_rate": 4.1156700243904017e-05, "loss": 0.0002, "step": 27346 }, { "epoch": 0.53, "learning_rate": 4.115605328364679e-05, "loss": 0.0004, "step": 27348 }, { "epoch": 0.53, "learning_rate": 4.1155406323389555e-05, "loss": 0.0002, "step": 27350 }, { "epoch": 0.53, "learning_rate": 4.1154759363132324e-05, "loss": 0.0001, "step": 27352 }, { "epoch": 0.53, "learning_rate": 4.115411240287509e-05, "loss": 0.0002, "step": 27354 }, { "epoch": 0.53, "learning_rate": 4.115346544261786e-05, "loss": 0.0003, "step": 27356 }, { "epoch": 0.53, "learning_rate": 4.115281848236063e-05, "loss": 0.0006, "step": 27358 }, { "epoch": 0.53, "learning_rate": 4.11521715221034e-05, "loss": 0.0091, "step": 27360 }, { "epoch": 0.53, "learning_rate": 4.115152456184617e-05, "loss": 0.0002, "step": 27362 }, { "epoch": 0.53, "learning_rate": 4.115087760158894e-05, "loss": 0.0, "step": 27364 }, { "epoch": 0.53, "learning_rate": 4.115023064133171e-05, "loss": 0.0001, "step": 27366 }, { "epoch": 0.53, "learning_rate": 4.114958368107447e-05, "loss": 0.0015, "step": 27368 }, { "epoch": 0.53, "learning_rate": 4.1148936720817246e-05, "loss": 0.0003, "step": 27370 }, { "epoch": 0.53, "learning_rate": 4.114828976056001e-05, "loss": 0.0012, "step": 27372 }, { "epoch": 0.53, "learning_rate": 4.114764280030278e-05, "loss": 0.0125, "step": 27374 }, { "epoch": 0.53, "learning_rate": 4.1146995840045546e-05, "loss": 0.0001, "step": 27376 }, { "epoch": 0.53, "learning_rate": 4.1146348879788315e-05, "loss": 0.0003, "step": 27378 }, { "epoch": 0.53, "learning_rate": 4.114570191953109e-05, "loss": 0.0003, "step": 27380 }, { "epoch": 0.53, "learning_rate": 4.114505495927385e-05, "loss": 0.0005, "step": 27382 }, { "epoch": 0.53, "learning_rate": 4.114440799901662e-05, "loss": 0.0, "step": 27384 }, { "epoch": 0.53, "learning_rate": 4.114376103875939e-05, "loss": 0.001, "step": 27386 }, { "epoch": 0.53, "learning_rate": 4.114311407850216e-05, "loss": 0.01, "step": 27388 }, { "epoch": 0.53, "learning_rate": 4.114246711824492e-05, "loss": 0.0009, "step": 27390 }, { "epoch": 0.53, "learning_rate": 4.11418201579877e-05, "loss": 0.0003, "step": 27392 }, { "epoch": 0.53, "learning_rate": 4.114117319773047e-05, "loss": 0.0008, "step": 27394 }, { "epoch": 0.53, "learning_rate": 4.114052623747323e-05, "loss": 0.0063, "step": 27396 }, { "epoch": 0.53, "learning_rate": 4.1139879277216006e-05, "loss": 0.0, "step": 27398 }, { "epoch": 0.53, "learning_rate": 4.113923231695877e-05, "loss": 0.0001, "step": 27400 }, { "epoch": 0.53, "learning_rate": 4.1138585356701544e-05, "loss": 0.0007, "step": 27402 }, { "epoch": 0.53, "learning_rate": 4.1137938396444306e-05, "loss": 0.0028, "step": 27404 }, { "epoch": 0.53, "learning_rate": 4.1137291436187076e-05, "loss": 0.0, "step": 27406 }, { "epoch": 0.53, "learning_rate": 4.1136644475929845e-05, "loss": 0.0003, "step": 27408 }, { "epoch": 0.53, "learning_rate": 4.1135997515672614e-05, "loss": 0.0, "step": 27410 }, { "epoch": 0.53, "learning_rate": 4.113535055541538e-05, "loss": 0.0001, "step": 27412 }, { "epoch": 0.53, "learning_rate": 4.113470359515815e-05, "loss": 0.0, "step": 27414 }, { "epoch": 0.53, "learning_rate": 4.113405663490092e-05, "loss": 0.009, "step": 27416 }, { "epoch": 0.53, "learning_rate": 4.113340967464368e-05, "loss": 0.0, "step": 27418 }, { "epoch": 0.53, "learning_rate": 4.113276271438646e-05, "loss": 0.0001, "step": 27420 }, { "epoch": 0.53, "learning_rate": 4.113211575412922e-05, "loss": 0.0, "step": 27422 }, { "epoch": 0.53, "learning_rate": 4.1131468793872e-05, "loss": 0.016, "step": 27424 }, { "epoch": 0.53, "learning_rate": 4.1130821833614766e-05, "loss": 0.0, "step": 27426 }, { "epoch": 0.53, "learning_rate": 4.113017487335753e-05, "loss": 0.0001, "step": 27428 }, { "epoch": 0.53, "learning_rate": 4.1129527913100305e-05, "loss": 0.0, "step": 27430 }, { "epoch": 0.53, "learning_rate": 4.112888095284307e-05, "loss": 0.0042, "step": 27432 }, { "epoch": 0.53, "learning_rate": 4.1128233992585836e-05, "loss": 0.0024, "step": 27434 }, { "epoch": 0.53, "learning_rate": 4.1127587032328605e-05, "loss": 0.0031, "step": 27436 }, { "epoch": 0.53, "learning_rate": 4.1126940072071374e-05, "loss": 0.0003, "step": 27438 }, { "epoch": 0.53, "learning_rate": 4.112629311181414e-05, "loss": 0.0002, "step": 27440 }, { "epoch": 0.53, "learning_rate": 4.112564615155691e-05, "loss": 0.0, "step": 27442 }, { "epoch": 0.53, "learning_rate": 4.112499919129968e-05, "loss": 0.0002, "step": 27444 }, { "epoch": 0.53, "learning_rate": 4.112435223104245e-05, "loss": 0.0, "step": 27446 }, { "epoch": 0.53, "learning_rate": 4.112370527078522e-05, "loss": 0.0001, "step": 27448 }, { "epoch": 0.53, "learning_rate": 4.112305831052798e-05, "loss": 0.0, "step": 27450 }, { "epoch": 0.53, "learning_rate": 4.112241135027076e-05, "loss": 0.0097, "step": 27452 }, { "epoch": 0.53, "learning_rate": 4.112176439001352e-05, "loss": 0.0001, "step": 27454 }, { "epoch": 0.53, "learning_rate": 4.112111742975629e-05, "loss": 0.0001, "step": 27456 }, { "epoch": 0.53, "learning_rate": 4.1120470469499065e-05, "loss": 0.0028, "step": 27458 }, { "epoch": 0.53, "learning_rate": 4.111982350924183e-05, "loss": 0.0004, "step": 27460 }, { "epoch": 0.53, "learning_rate": 4.11191765489846e-05, "loss": 0.0031, "step": 27462 }, { "epoch": 0.53, "learning_rate": 4.1118529588727365e-05, "loss": 0.0, "step": 27464 }, { "epoch": 0.53, "learning_rate": 4.1117882628470135e-05, "loss": 0.0, "step": 27466 }, { "epoch": 0.53, "learning_rate": 4.1117235668212904e-05, "loss": 0.0, "step": 27468 }, { "epoch": 0.53, "learning_rate": 4.111658870795567e-05, "loss": 0.0013, "step": 27470 }, { "epoch": 0.53, "learning_rate": 4.111594174769844e-05, "loss": 0.0, "step": 27472 }, { "epoch": 0.53, "learning_rate": 4.111529478744121e-05, "loss": 0.0034, "step": 27474 }, { "epoch": 0.53, "learning_rate": 4.111464782718398e-05, "loss": 0.0, "step": 27476 }, { "epoch": 0.53, "learning_rate": 4.111400086692674e-05, "loss": 0.004, "step": 27478 }, { "epoch": 0.53, "learning_rate": 4.111335390666952e-05, "loss": 0.001, "step": 27480 }, { "epoch": 0.53, "learning_rate": 4.111270694641228e-05, "loss": 0.0002, "step": 27482 }, { "epoch": 0.53, "learning_rate": 4.1112059986155056e-05, "loss": 0.0006, "step": 27484 }, { "epoch": 0.53, "learning_rate": 4.111141302589782e-05, "loss": 0.0028, "step": 27486 }, { "epoch": 0.53, "learning_rate": 4.111076606564059e-05, "loss": 0.0317, "step": 27488 }, { "epoch": 0.53, "learning_rate": 4.111011910538336e-05, "loss": 0.0033, "step": 27490 }, { "epoch": 0.53, "learning_rate": 4.1109472145126126e-05, "loss": 0.003, "step": 27492 }, { "epoch": 0.53, "learning_rate": 4.1108825184868895e-05, "loss": 0.0047, "step": 27494 }, { "epoch": 0.53, "learning_rate": 4.1108178224611664e-05, "loss": 0.0134, "step": 27496 }, { "epoch": 0.53, "learning_rate": 4.110753126435443e-05, "loss": 0.0043, "step": 27498 }, { "epoch": 0.53, "learning_rate": 4.11068843040972e-05, "loss": 0.0008, "step": 27500 }, { "epoch": 0.53, "learning_rate": 4.110623734383997e-05, "loss": 0.0019, "step": 27502 }, { "epoch": 0.53, "learning_rate": 4.110559038358274e-05, "loss": 0.0007, "step": 27504 }, { "epoch": 0.53, "learning_rate": 4.110494342332551e-05, "loss": 0.0001, "step": 27506 }, { "epoch": 0.53, "learning_rate": 4.110429646306828e-05, "loss": 0.0001, "step": 27508 }, { "epoch": 0.53, "learning_rate": 4.110364950281104e-05, "loss": 0.0137, "step": 27510 }, { "epoch": 0.53, "learning_rate": 4.110300254255382e-05, "loss": 0.0006, "step": 27512 }, { "epoch": 0.53, "learning_rate": 4.110235558229658e-05, "loss": 0.0003, "step": 27514 }, { "epoch": 0.53, "learning_rate": 4.110170862203935e-05, "loss": 0.0004, "step": 27516 }, { "epoch": 0.53, "learning_rate": 4.110106166178212e-05, "loss": 0.0023, "step": 27518 }, { "epoch": 0.53, "learning_rate": 4.1100414701524886e-05, "loss": 0.0013, "step": 27520 }, { "epoch": 0.53, "learning_rate": 4.1099767741267655e-05, "loss": 0.001, "step": 27522 }, { "epoch": 0.53, "learning_rate": 4.1099120781010424e-05, "loss": 0.0087, "step": 27524 }, { "epoch": 0.53, "learning_rate": 4.1098473820753193e-05, "loss": 0.0021, "step": 27526 }, { "epoch": 0.53, "learning_rate": 4.109782686049596e-05, "loss": 0.0, "step": 27528 }, { "epoch": 0.53, "learning_rate": 4.109717990023873e-05, "loss": 0.0002, "step": 27530 }, { "epoch": 0.53, "learning_rate": 4.1096532939981494e-05, "loss": 0.0001, "step": 27532 }, { "epoch": 0.53, "learning_rate": 4.109588597972427e-05, "loss": 0.0005, "step": 27534 }, { "epoch": 0.53, "learning_rate": 4.109523901946704e-05, "loss": 0.0014, "step": 27536 }, { "epoch": 0.53, "learning_rate": 4.10945920592098e-05, "loss": 0.0, "step": 27538 }, { "epoch": 0.53, "learning_rate": 4.109394509895258e-05, "loss": 0.0059, "step": 27540 }, { "epoch": 0.53, "learning_rate": 4.109329813869534e-05, "loss": 0.0078, "step": 27542 }, { "epoch": 0.53, "learning_rate": 4.1092651178438115e-05, "loss": 0.0002, "step": 27544 }, { "epoch": 0.53, "learning_rate": 4.109200421818088e-05, "loss": 0.0001, "step": 27546 }, { "epoch": 0.53, "learning_rate": 4.109135725792365e-05, "loss": 0.0193, "step": 27548 }, { "epoch": 0.53, "learning_rate": 4.1090710297666416e-05, "loss": 0.0003, "step": 27550 }, { "epoch": 0.53, "learning_rate": 4.1090063337409185e-05, "loss": 0.0264, "step": 27552 }, { "epoch": 0.53, "learning_rate": 4.1089416377151954e-05, "loss": 0.0006, "step": 27554 }, { "epoch": 0.53, "learning_rate": 4.108876941689472e-05, "loss": 0.0, "step": 27556 }, { "epoch": 0.53, "learning_rate": 4.108812245663749e-05, "loss": 0.0002, "step": 27558 }, { "epoch": 0.53, "learning_rate": 4.1087475496380254e-05, "loss": 0.0171, "step": 27560 }, { "epoch": 0.53, "learning_rate": 4.108682853612303e-05, "loss": 0.0001, "step": 27562 }, { "epoch": 0.53, "learning_rate": 4.108618157586579e-05, "loss": 0.0001, "step": 27564 }, { "epoch": 0.54, "learning_rate": 4.108553461560857e-05, "loss": 0.0001, "step": 27566 }, { "epoch": 0.54, "learning_rate": 4.108488765535133e-05, "loss": 0.0005, "step": 27568 }, { "epoch": 0.54, "learning_rate": 4.10842406950941e-05, "loss": 0.008, "step": 27570 }, { "epoch": 0.54, "learning_rate": 4.1083593734836876e-05, "loss": 0.0002, "step": 27572 }, { "epoch": 0.54, "learning_rate": 4.108294677457964e-05, "loss": 0.0013, "step": 27574 }, { "epoch": 0.54, "learning_rate": 4.108229981432241e-05, "loss": 0.0049, "step": 27576 }, { "epoch": 0.54, "learning_rate": 4.1081652854065176e-05, "loss": 0.0001, "step": 27578 }, { "epoch": 0.54, "learning_rate": 4.1081005893807945e-05, "loss": 0.0106, "step": 27580 }, { "epoch": 0.54, "learning_rate": 4.1080358933550714e-05, "loss": 0.0174, "step": 27582 }, { "epoch": 0.54, "learning_rate": 4.1079711973293483e-05, "loss": 0.0002, "step": 27584 }, { "epoch": 0.54, "learning_rate": 4.107906501303625e-05, "loss": 0.0078, "step": 27586 }, { "epoch": 0.54, "learning_rate": 4.107841805277902e-05, "loss": 0.0019, "step": 27588 }, { "epoch": 0.54, "learning_rate": 4.107777109252179e-05, "loss": 0.0005, "step": 27590 }, { "epoch": 0.54, "learning_rate": 4.107712413226455e-05, "loss": 0.0, "step": 27592 }, { "epoch": 0.54, "learning_rate": 4.107647717200733e-05, "loss": 0.0, "step": 27594 }, { "epoch": 0.54, "learning_rate": 4.107583021175009e-05, "loss": 0.0011, "step": 27596 }, { "epoch": 0.54, "learning_rate": 4.107518325149286e-05, "loss": 0.0023, "step": 27598 }, { "epoch": 0.54, "learning_rate": 4.107453629123563e-05, "loss": 0.0, "step": 27600 }, { "epoch": 0.54, "learning_rate": 4.10738893309784e-05, "loss": 0.0001, "step": 27602 }, { "epoch": 0.54, "learning_rate": 4.1073242370721174e-05, "loss": 0.0006, "step": 27604 }, { "epoch": 0.54, "learning_rate": 4.1072595410463937e-05, "loss": 0.0, "step": 27606 }, { "epoch": 0.54, "learning_rate": 4.1071948450206706e-05, "loss": 0.0001, "step": 27608 }, { "epoch": 0.54, "learning_rate": 4.1071301489949475e-05, "loss": 0.0148, "step": 27610 }, { "epoch": 0.54, "learning_rate": 4.1070654529692244e-05, "loss": 0.0, "step": 27612 }, { "epoch": 0.54, "learning_rate": 4.107000756943501e-05, "loss": 0.0031, "step": 27614 }, { "epoch": 0.54, "learning_rate": 4.106936060917778e-05, "loss": 0.0051, "step": 27616 }, { "epoch": 0.54, "learning_rate": 4.106871364892055e-05, "loss": 0.0, "step": 27618 }, { "epoch": 0.54, "learning_rate": 4.106806668866331e-05, "loss": 0.0, "step": 27620 }, { "epoch": 0.54, "learning_rate": 4.106741972840609e-05, "loss": 0.0029, "step": 27622 }, { "epoch": 0.54, "learning_rate": 4.106677276814885e-05, "loss": 0.0025, "step": 27624 }, { "epoch": 0.54, "learning_rate": 4.106612580789163e-05, "loss": 0.0002, "step": 27626 }, { "epoch": 0.54, "learning_rate": 4.106547884763439e-05, "loss": 0.0022, "step": 27628 }, { "epoch": 0.54, "learning_rate": 4.106483188737716e-05, "loss": 0.0, "step": 27630 }, { "epoch": 0.54, "learning_rate": 4.106418492711993e-05, "loss": 0.0004, "step": 27632 }, { "epoch": 0.54, "learning_rate": 4.10635379668627e-05, "loss": 0.0, "step": 27634 }, { "epoch": 0.54, "learning_rate": 4.1062891006605466e-05, "loss": 0.0001, "step": 27636 }, { "epoch": 0.54, "learning_rate": 4.1062244046348235e-05, "loss": 0.0016, "step": 27638 }, { "epoch": 0.54, "learning_rate": 4.1061597086091004e-05, "loss": 0.0, "step": 27640 }, { "epoch": 0.54, "learning_rate": 4.106095012583377e-05, "loss": 0.0001, "step": 27642 }, { "epoch": 0.54, "learning_rate": 4.106030316557654e-05, "loss": 0.0094, "step": 27644 }, { "epoch": 0.54, "learning_rate": 4.1059656205319305e-05, "loss": 0.0071, "step": 27646 }, { "epoch": 0.54, "learning_rate": 4.105900924506208e-05, "loss": 0.0, "step": 27648 }, { "epoch": 0.54, "learning_rate": 4.105836228480485e-05, "loss": 0.01, "step": 27650 }, { "epoch": 0.54, "learning_rate": 4.105771532454761e-05, "loss": 0.029, "step": 27652 }, { "epoch": 0.54, "learning_rate": 4.105706836429039e-05, "loss": 0.0002, "step": 27654 }, { "epoch": 0.54, "learning_rate": 4.105642140403315e-05, "loss": 0.0001, "step": 27656 }, { "epoch": 0.54, "learning_rate": 4.105577444377592e-05, "loss": 0.0001, "step": 27658 }, { "epoch": 0.54, "learning_rate": 4.105512748351869e-05, "loss": 0.0, "step": 27660 }, { "epoch": 0.54, "learning_rate": 4.105448052326146e-05, "loss": 0.002, "step": 27662 }, { "epoch": 0.54, "learning_rate": 4.1053833563004226e-05, "loss": 0.0002, "step": 27664 }, { "epoch": 0.54, "learning_rate": 4.1053186602746996e-05, "loss": 0.0122, "step": 27666 }, { "epoch": 0.54, "learning_rate": 4.1052539642489765e-05, "loss": 0.0014, "step": 27668 }, { "epoch": 0.54, "learning_rate": 4.1051892682232534e-05, "loss": 0.0, "step": 27670 }, { "epoch": 0.54, "learning_rate": 4.10512457219753e-05, "loss": 0.0, "step": 27672 }, { "epoch": 0.54, "learning_rate": 4.1050598761718065e-05, "loss": 0.0001, "step": 27674 }, { "epoch": 0.54, "learning_rate": 4.104995180146084e-05, "loss": 0.0035, "step": 27676 }, { "epoch": 0.54, "learning_rate": 4.10493048412036e-05, "loss": 0.0025, "step": 27678 }, { "epoch": 0.54, "learning_rate": 4.104865788094637e-05, "loss": 0.0014, "step": 27680 }, { "epoch": 0.54, "learning_rate": 4.104801092068915e-05, "loss": 0.0006, "step": 27682 }, { "epoch": 0.54, "learning_rate": 4.104736396043191e-05, "loss": 0.0005, "step": 27684 }, { "epoch": 0.54, "learning_rate": 4.1046717000174686e-05, "loss": 0.0245, "step": 27686 }, { "epoch": 0.54, "learning_rate": 4.104607003991745e-05, "loss": 0.0172, "step": 27688 }, { "epoch": 0.54, "learning_rate": 4.104542307966022e-05, "loss": 0.0013, "step": 27690 }, { "epoch": 0.54, "learning_rate": 4.104477611940299e-05, "loss": 0.0028, "step": 27692 }, { "epoch": 0.54, "learning_rate": 4.1044129159145756e-05, "loss": 0.001, "step": 27694 }, { "epoch": 0.54, "learning_rate": 4.1043482198888525e-05, "loss": 0.0029, "step": 27696 }, { "epoch": 0.54, "learning_rate": 4.1042835238631294e-05, "loss": 0.0, "step": 27698 }, { "epoch": 0.54, "learning_rate": 4.104218827837406e-05, "loss": 0.0003, "step": 27700 }, { "epoch": 0.54, "learning_rate": 4.1041541318116825e-05, "loss": 0.0001, "step": 27702 }, { "epoch": 0.54, "learning_rate": 4.10408943578596e-05, "loss": 0.0063, "step": 27704 }, { "epoch": 0.54, "learning_rate": 4.1040247397602364e-05, "loss": 0.0001, "step": 27706 }, { "epoch": 0.54, "learning_rate": 4.103960043734514e-05, "loss": 0.0032, "step": 27708 }, { "epoch": 0.54, "learning_rate": 4.10389534770879e-05, "loss": 0.0003, "step": 27710 }, { "epoch": 0.54, "learning_rate": 4.103830651683067e-05, "loss": 0.0, "step": 27712 }, { "epoch": 0.54, "learning_rate": 4.103765955657345e-05, "loss": 0.0032, "step": 27714 }, { "epoch": 0.54, "learning_rate": 4.103701259631621e-05, "loss": 0.0009, "step": 27716 }, { "epoch": 0.54, "learning_rate": 4.103636563605898e-05, "loss": 0.0005, "step": 27718 }, { "epoch": 0.54, "learning_rate": 4.103571867580175e-05, "loss": 0.0005, "step": 27720 }, { "epoch": 0.54, "learning_rate": 4.1035071715544516e-05, "loss": 0.0048, "step": 27722 }, { "epoch": 0.54, "learning_rate": 4.1034424755287285e-05, "loss": 0.0, "step": 27724 }, { "epoch": 0.54, "learning_rate": 4.1033777795030054e-05, "loss": 0.0006, "step": 27726 }, { "epoch": 0.54, "learning_rate": 4.1033130834772824e-05, "loss": 0.0, "step": 27728 }, { "epoch": 0.54, "learning_rate": 4.103248387451559e-05, "loss": 0.0001, "step": 27730 }, { "epoch": 0.54, "learning_rate": 4.103183691425836e-05, "loss": 0.0001, "step": 27732 }, { "epoch": 0.54, "learning_rate": 4.1031189954001124e-05, "loss": 0.0004, "step": 27734 }, { "epoch": 0.54, "learning_rate": 4.10305429937439e-05, "loss": 0.0, "step": 27736 }, { "epoch": 0.54, "learning_rate": 4.102989603348666e-05, "loss": 0.0106, "step": 27738 }, { "epoch": 0.54, "learning_rate": 4.102924907322943e-05, "loss": 0.0071, "step": 27740 }, { "epoch": 0.54, "learning_rate": 4.10286021129722e-05, "loss": 0.0033, "step": 27742 }, { "epoch": 0.54, "learning_rate": 4.102795515271497e-05, "loss": 0.0001, "step": 27744 }, { "epoch": 0.54, "learning_rate": 4.102730819245774e-05, "loss": 0.0047, "step": 27746 }, { "epoch": 0.54, "learning_rate": 4.102666123220051e-05, "loss": 0.0, "step": 27748 }, { "epoch": 0.54, "learning_rate": 4.102601427194328e-05, "loss": 0.0032, "step": 27750 }, { "epoch": 0.54, "learning_rate": 4.1025367311686046e-05, "loss": 0.0001, "step": 27752 }, { "epoch": 0.54, "learning_rate": 4.1024720351428815e-05, "loss": 0.0, "step": 27754 }, { "epoch": 0.54, "learning_rate": 4.102407339117158e-05, "loss": 0.0, "step": 27756 }, { "epoch": 0.54, "learning_rate": 4.102342643091435e-05, "loss": 0.0018, "step": 27758 }, { "epoch": 0.54, "learning_rate": 4.102277947065712e-05, "loss": 0.0017, "step": 27760 }, { "epoch": 0.54, "learning_rate": 4.1022132510399884e-05, "loss": 0.0, "step": 27762 }, { "epoch": 0.54, "learning_rate": 4.102148555014266e-05, "loss": 0.0001, "step": 27764 }, { "epoch": 0.54, "learning_rate": 4.102083858988542e-05, "loss": 0.0001, "step": 27766 }, { "epoch": 0.54, "learning_rate": 4.10201916296282e-05, "loss": 0.0, "step": 27768 }, { "epoch": 0.54, "learning_rate": 4.101954466937096e-05, "loss": 0.0007, "step": 27770 }, { "epoch": 0.54, "learning_rate": 4.101889770911373e-05, "loss": 0.0003, "step": 27772 }, { "epoch": 0.54, "learning_rate": 4.10182507488565e-05, "loss": 0.0018, "step": 27774 }, { "epoch": 0.54, "learning_rate": 4.101760378859927e-05, "loss": 0.0001, "step": 27776 }, { "epoch": 0.54, "learning_rate": 4.101695682834204e-05, "loss": 0.0, "step": 27778 }, { "epoch": 0.54, "learning_rate": 4.1016309868084806e-05, "loss": 0.0, "step": 27780 }, { "epoch": 0.54, "learning_rate": 4.1015662907827575e-05, "loss": 0.0001, "step": 27782 }, { "epoch": 0.54, "learning_rate": 4.1015015947570344e-05, "loss": 0.0003, "step": 27784 }, { "epoch": 0.54, "learning_rate": 4.1014368987313113e-05, "loss": 0.0005, "step": 27786 }, { "epoch": 0.54, "learning_rate": 4.1013722027055876e-05, "loss": 0.0, "step": 27788 }, { "epoch": 0.54, "learning_rate": 4.101307506679865e-05, "loss": 0.0001, "step": 27790 }, { "epoch": 0.54, "learning_rate": 4.1012428106541414e-05, "loss": 0.0, "step": 27792 }, { "epoch": 0.54, "learning_rate": 4.101178114628418e-05, "loss": 0.0036, "step": 27794 }, { "epoch": 0.54, "learning_rate": 4.101113418602696e-05, "loss": 0.0116, "step": 27796 }, { "epoch": 0.54, "learning_rate": 4.101048722576972e-05, "loss": 0.0001, "step": 27798 }, { "epoch": 0.54, "learning_rate": 4.100984026551249e-05, "loss": 0.0158, "step": 27800 }, { "epoch": 0.54, "learning_rate": 4.100919330525526e-05, "loss": 0.0001, "step": 27802 }, { "epoch": 0.54, "learning_rate": 4.100854634499803e-05, "loss": 0.0001, "step": 27804 }, { "epoch": 0.54, "learning_rate": 4.10078993847408e-05, "loss": 0.0084, "step": 27806 }, { "epoch": 0.54, "learning_rate": 4.100725242448357e-05, "loss": 0.0, "step": 27808 }, { "epoch": 0.54, "learning_rate": 4.1006605464226336e-05, "loss": 0.0, "step": 27810 }, { "epoch": 0.54, "learning_rate": 4.1005958503969105e-05, "loss": 0.0002, "step": 27812 }, { "epoch": 0.54, "learning_rate": 4.1005311543711874e-05, "loss": 0.0018, "step": 27814 }, { "epoch": 0.54, "learning_rate": 4.1004664583454636e-05, "loss": 0.0001, "step": 27816 }, { "epoch": 0.54, "learning_rate": 4.100401762319741e-05, "loss": 0.0, "step": 27818 }, { "epoch": 0.54, "learning_rate": 4.1003370662940174e-05, "loss": 0.0001, "step": 27820 }, { "epoch": 0.54, "learning_rate": 4.1002723702682943e-05, "loss": 0.0006, "step": 27822 }, { "epoch": 0.54, "learning_rate": 4.100207674242571e-05, "loss": 0.0002, "step": 27824 }, { "epoch": 0.54, "learning_rate": 4.100142978216848e-05, "loss": 0.0, "step": 27826 }, { "epoch": 0.54, "learning_rate": 4.100078282191126e-05, "loss": 0.0015, "step": 27828 }, { "epoch": 0.54, "learning_rate": 4.100013586165402e-05, "loss": 0.0001, "step": 27830 }, { "epoch": 0.54, "learning_rate": 4.099948890139679e-05, "loss": 0.0134, "step": 27832 }, { "epoch": 0.54, "learning_rate": 4.099884194113956e-05, "loss": 0.0004, "step": 27834 }, { "epoch": 0.54, "learning_rate": 4.099819498088233e-05, "loss": 0.004, "step": 27836 }, { "epoch": 0.54, "learning_rate": 4.0997548020625096e-05, "loss": 0.0107, "step": 27838 }, { "epoch": 0.54, "learning_rate": 4.0996901060367865e-05, "loss": 0.0014, "step": 27840 }, { "epoch": 0.54, "learning_rate": 4.0996254100110634e-05, "loss": 0.0002, "step": 27842 }, { "epoch": 0.54, "learning_rate": 4.0995607139853397e-05, "loss": 0.0, "step": 27844 }, { "epoch": 0.54, "learning_rate": 4.099496017959617e-05, "loss": 0.0001, "step": 27846 }, { "epoch": 0.54, "learning_rate": 4.0994313219338935e-05, "loss": 0.0427, "step": 27848 }, { "epoch": 0.54, "learning_rate": 4.099366625908171e-05, "loss": 0.0002, "step": 27850 }, { "epoch": 0.54, "learning_rate": 4.099301929882447e-05, "loss": 0.0, "step": 27852 }, { "epoch": 0.54, "learning_rate": 4.099237233856724e-05, "loss": 0.0303, "step": 27854 }, { "epoch": 0.54, "learning_rate": 4.099172537831001e-05, "loss": 0.0002, "step": 27856 }, { "epoch": 0.54, "learning_rate": 4.099107841805278e-05, "loss": 0.0254, "step": 27858 }, { "epoch": 0.54, "learning_rate": 4.099043145779555e-05, "loss": 0.0003, "step": 27860 }, { "epoch": 0.54, "learning_rate": 4.098978449753832e-05, "loss": 0.0001, "step": 27862 }, { "epoch": 0.54, "learning_rate": 4.098913753728109e-05, "loss": 0.0045, "step": 27864 }, { "epoch": 0.54, "learning_rate": 4.0988490577023857e-05, "loss": 0.0407, "step": 27866 }, { "epoch": 0.54, "learning_rate": 4.0987843616766626e-05, "loss": 0.0, "step": 27868 }, { "epoch": 0.54, "learning_rate": 4.098719665650939e-05, "loss": 0.0001, "step": 27870 }, { "epoch": 0.54, "learning_rate": 4.0986549696252164e-05, "loss": 0.0001, "step": 27872 }, { "epoch": 0.54, "learning_rate": 4.098590273599493e-05, "loss": 0.0038, "step": 27874 }, { "epoch": 0.54, "learning_rate": 4.0985255775737695e-05, "loss": 0.0, "step": 27876 }, { "epoch": 0.54, "learning_rate": 4.098460881548047e-05, "loss": 0.0002, "step": 27878 }, { "epoch": 0.54, "learning_rate": 4.098396185522323e-05, "loss": 0.0016, "step": 27880 }, { "epoch": 0.54, "learning_rate": 4.0983314894966e-05, "loss": 0.0, "step": 27882 }, { "epoch": 0.54, "learning_rate": 4.098266793470877e-05, "loss": 0.0001, "step": 27884 }, { "epoch": 0.54, "learning_rate": 4.098202097445154e-05, "loss": 0.0, "step": 27886 }, { "epoch": 0.54, "learning_rate": 4.098137401419431e-05, "loss": 0.0003, "step": 27888 }, { "epoch": 0.54, "learning_rate": 4.098072705393708e-05, "loss": 0.0001, "step": 27890 }, { "epoch": 0.54, "learning_rate": 4.098008009367985e-05, "loss": 0.0012, "step": 27892 }, { "epoch": 0.54, "learning_rate": 4.097943313342262e-05, "loss": 0.0001, "step": 27894 }, { "epoch": 0.54, "learning_rate": 4.0978786173165386e-05, "loss": 0.0001, "step": 27896 }, { "epoch": 0.54, "learning_rate": 4.097813921290815e-05, "loss": 0.001, "step": 27898 }, { "epoch": 0.54, "learning_rate": 4.0977492252650924e-05, "loss": 0.0016, "step": 27900 }, { "epoch": 0.54, "learning_rate": 4.0976845292393686e-05, "loss": 0.0236, "step": 27902 }, { "epoch": 0.54, "learning_rate": 4.0976198332136456e-05, "loss": 0.0043, "step": 27904 }, { "epoch": 0.54, "learning_rate": 4.097555137187923e-05, "loss": 0.0002, "step": 27906 }, { "epoch": 0.54, "learning_rate": 4.0974904411621994e-05, "loss": 0.0011, "step": 27908 }, { "epoch": 0.54, "learning_rate": 4.097425745136477e-05, "loss": 0.0009, "step": 27910 }, { "epoch": 0.54, "learning_rate": 4.097361049110753e-05, "loss": 0.0013, "step": 27912 }, { "epoch": 0.54, "learning_rate": 4.09729635308503e-05, "loss": 0.0, "step": 27914 }, { "epoch": 0.54, "learning_rate": 4.097231657059307e-05, "loss": 0.0065, "step": 27916 }, { "epoch": 0.54, "learning_rate": 4.097166961033584e-05, "loss": 0.0015, "step": 27918 }, { "epoch": 0.54, "learning_rate": 4.097102265007861e-05, "loss": 0.0002, "step": 27920 }, { "epoch": 0.54, "learning_rate": 4.097037568982138e-05, "loss": 0.0042, "step": 27922 }, { "epoch": 0.54, "learning_rate": 4.0969728729564146e-05, "loss": 0.0005, "step": 27924 }, { "epoch": 0.54, "learning_rate": 4.096908176930691e-05, "loss": 0.0, "step": 27926 }, { "epoch": 0.54, "learning_rate": 4.0968434809049685e-05, "loss": 0.0002, "step": 27928 }, { "epoch": 0.54, "learning_rate": 4.096778784879245e-05, "loss": 0.0046, "step": 27930 }, { "epoch": 0.54, "learning_rate": 4.096714088853522e-05, "loss": 0.0003, "step": 27932 }, { "epoch": 0.54, "learning_rate": 4.0966493928277985e-05, "loss": 0.0114, "step": 27934 }, { "epoch": 0.54, "learning_rate": 4.0965846968020754e-05, "loss": 0.0001, "step": 27936 }, { "epoch": 0.54, "learning_rate": 4.096520000776353e-05, "loss": 0.0061, "step": 27938 }, { "epoch": 0.54, "learning_rate": 4.096455304750629e-05, "loss": 0.0002, "step": 27940 }, { "epoch": 0.54, "learning_rate": 4.096390608724906e-05, "loss": 0.0001, "step": 27942 }, { "epoch": 0.54, "learning_rate": 4.096325912699183e-05, "loss": 0.0021, "step": 27944 }, { "epoch": 0.54, "learning_rate": 4.09626121667346e-05, "loss": 0.0063, "step": 27946 }, { "epoch": 0.54, "learning_rate": 4.096196520647737e-05, "loss": 0.0004, "step": 27948 }, { "epoch": 0.54, "learning_rate": 4.096131824622014e-05, "loss": 0.0066, "step": 27950 }, { "epoch": 0.54, "learning_rate": 4.096067128596291e-05, "loss": 0.0024, "step": 27952 }, { "epoch": 0.54, "learning_rate": 4.0960024325705676e-05, "loss": 0.0001, "step": 27954 }, { "epoch": 0.54, "learning_rate": 4.0959377365448445e-05, "loss": 0.0001, "step": 27956 }, { "epoch": 0.54, "learning_rate": 4.095873040519121e-05, "loss": 0.0002, "step": 27958 }, { "epoch": 0.54, "learning_rate": 4.095808344493398e-05, "loss": 0.0001, "step": 27960 }, { "epoch": 0.54, "learning_rate": 4.0957436484676745e-05, "loss": 0.0001, "step": 27962 }, { "epoch": 0.54, "learning_rate": 4.0956789524419515e-05, "loss": 0.0005, "step": 27964 }, { "epoch": 0.54, "learning_rate": 4.0956142564162284e-05, "loss": 0.001, "step": 27966 }, { "epoch": 0.54, "learning_rate": 4.095549560390505e-05, "loss": 0.0055, "step": 27968 }, { "epoch": 0.54, "learning_rate": 4.095484864364782e-05, "loss": 0.0, "step": 27970 }, { "epoch": 0.54, "learning_rate": 4.095420168339059e-05, "loss": 0.0001, "step": 27972 }, { "epoch": 0.54, "learning_rate": 4.095355472313336e-05, "loss": 0.005, "step": 27974 }, { "epoch": 0.54, "learning_rate": 4.095290776287613e-05, "loss": 0.0123, "step": 27976 }, { "epoch": 0.54, "learning_rate": 4.09522608026189e-05, "loss": 0.0001, "step": 27978 }, { "epoch": 0.54, "learning_rate": 4.095161384236166e-05, "loss": 0.0249, "step": 27980 }, { "epoch": 0.54, "learning_rate": 4.0950966882104436e-05, "loss": 0.0021, "step": 27982 }, { "epoch": 0.54, "learning_rate": 4.0950319921847205e-05, "loss": 0.0012, "step": 27984 }, { "epoch": 0.54, "learning_rate": 4.094967296158997e-05, "loss": 0.0025, "step": 27986 }, { "epoch": 0.54, "learning_rate": 4.0949026001332744e-05, "loss": 0.001, "step": 27988 }, { "epoch": 0.54, "learning_rate": 4.0948379041075506e-05, "loss": 0.0001, "step": 27990 }, { "epoch": 0.54, "learning_rate": 4.094773208081828e-05, "loss": 0.0001, "step": 27992 }, { "epoch": 0.54, "learning_rate": 4.0947085120561044e-05, "loss": 0.0003, "step": 27994 }, { "epoch": 0.54, "learning_rate": 4.094643816030381e-05, "loss": 0.0001, "step": 27996 }, { "epoch": 0.54, "learning_rate": 4.094579120004658e-05, "loss": 0.0001, "step": 27998 }, { "epoch": 0.54, "learning_rate": 4.094514423978935e-05, "loss": 0.0011, "step": 28000 }, { "epoch": 0.54, "learning_rate": 4.094449727953212e-05, "loss": 0.0077, "step": 28002 }, { "epoch": 0.54, "learning_rate": 4.094385031927489e-05, "loss": 0.0002, "step": 28004 }, { "epoch": 0.54, "learning_rate": 4.094320335901766e-05, "loss": 0.0, "step": 28006 }, { "epoch": 0.54, "learning_rate": 4.094255639876043e-05, "loss": 0.0076, "step": 28008 }, { "epoch": 0.54, "learning_rate": 4.09419094385032e-05, "loss": 0.0019, "step": 28010 }, { "epoch": 0.54, "learning_rate": 4.094126247824596e-05, "loss": 0.0, "step": 28012 }, { "epoch": 0.54, "learning_rate": 4.0940615517988735e-05, "loss": 0.001, "step": 28014 }, { "epoch": 0.54, "learning_rate": 4.0939968557731504e-05, "loss": 0.0041, "step": 28016 }, { "epoch": 0.54, "learning_rate": 4.0939321597474266e-05, "loss": 0.0001, "step": 28018 }, { "epoch": 0.54, "learning_rate": 4.093867463721704e-05, "loss": 0.0001, "step": 28020 }, { "epoch": 0.54, "learning_rate": 4.0938027676959804e-05, "loss": 0.0147, "step": 28022 }, { "epoch": 0.54, "learning_rate": 4.0937380716702574e-05, "loss": 0.0072, "step": 28024 }, { "epoch": 0.54, "learning_rate": 4.093673375644534e-05, "loss": 0.0, "step": 28026 }, { "epoch": 0.54, "learning_rate": 4.093608679618811e-05, "loss": 0.0001, "step": 28028 }, { "epoch": 0.54, "learning_rate": 4.093543983593088e-05, "loss": 0.0064, "step": 28030 }, { "epoch": 0.54, "learning_rate": 4.093479287567365e-05, "loss": 0.0001, "step": 28032 }, { "epoch": 0.54, "learning_rate": 4.093414591541642e-05, "loss": 0.0, "step": 28034 }, { "epoch": 0.54, "learning_rate": 4.093349895515919e-05, "loss": 0.0, "step": 28036 }, { "epoch": 0.54, "learning_rate": 4.093285199490196e-05, "loss": 0.0081, "step": 28038 }, { "epoch": 0.54, "learning_rate": 4.093220503464472e-05, "loss": 0.001, "step": 28040 }, { "epoch": 0.54, "learning_rate": 4.0931558074387495e-05, "loss": 0.0061, "step": 28042 }, { "epoch": 0.54, "learning_rate": 4.093091111413026e-05, "loss": 0.0122, "step": 28044 }, { "epoch": 0.54, "learning_rate": 4.093026415387303e-05, "loss": 0.0042, "step": 28046 }, { "epoch": 0.54, "learning_rate": 4.0929617193615796e-05, "loss": 0.0, "step": 28048 }, { "epoch": 0.54, "learning_rate": 4.0928970233358565e-05, "loss": 0.0, "step": 28050 }, { "epoch": 0.54, "learning_rate": 4.092832327310134e-05, "loss": 0.0001, "step": 28052 }, { "epoch": 0.54, "learning_rate": 4.09276763128441e-05, "loss": 0.0, "step": 28054 }, { "epoch": 0.54, "learning_rate": 4.092702935258687e-05, "loss": 0.0032, "step": 28056 }, { "epoch": 0.54, "learning_rate": 4.092638239232964e-05, "loss": 0.0002, "step": 28058 }, { "epoch": 0.54, "learning_rate": 4.092573543207241e-05, "loss": 0.0, "step": 28060 }, { "epoch": 0.54, "learning_rate": 4.092508847181518e-05, "loss": 0.0, "step": 28062 }, { "epoch": 0.54, "learning_rate": 4.092444151155795e-05, "loss": 0.0, "step": 28064 }, { "epoch": 0.54, "learning_rate": 4.092379455130072e-05, "loss": 0.0051, "step": 28066 }, { "epoch": 0.54, "learning_rate": 4.092314759104348e-05, "loss": 0.0004, "step": 28068 }, { "epoch": 0.54, "learning_rate": 4.0922500630786256e-05, "loss": 0.0, "step": 28070 }, { "epoch": 0.54, "learning_rate": 4.092185367052902e-05, "loss": 0.0001, "step": 28072 }, { "epoch": 0.54, "learning_rate": 4.0921206710271794e-05, "loss": 0.0026, "step": 28074 }, { "epoch": 0.54, "learning_rate": 4.0920559750014556e-05, "loss": 0.0001, "step": 28076 }, { "epoch": 0.54, "learning_rate": 4.0919912789757325e-05, "loss": 0.0024, "step": 28078 }, { "epoch": 0.54, "learning_rate": 4.0919265829500094e-05, "loss": 0.0219, "step": 28080 }, { "epoch": 0.55, "learning_rate": 4.0918618869242863e-05, "loss": 0.0, "step": 28082 }, { "epoch": 0.55, "learning_rate": 4.091797190898563e-05, "loss": 0.0011, "step": 28084 }, { "epoch": 0.55, "learning_rate": 4.09173249487284e-05, "loss": 0.0006, "step": 28086 }, { "epoch": 0.55, "learning_rate": 4.091667798847117e-05, "loss": 0.0254, "step": 28088 }, { "epoch": 0.55, "learning_rate": 4.091603102821394e-05, "loss": 0.0001, "step": 28090 }, { "epoch": 0.55, "learning_rate": 4.091538406795671e-05, "loss": 0.0004, "step": 28092 }, { "epoch": 0.55, "learning_rate": 4.091473710769948e-05, "loss": 0.0005, "step": 28094 }, { "epoch": 0.55, "learning_rate": 4.091409014744225e-05, "loss": 0.0, "step": 28096 }, { "epoch": 0.55, "learning_rate": 4.0913443187185016e-05, "loss": 0.0, "step": 28098 }, { "epoch": 0.55, "learning_rate": 4.091279622692778e-05, "loss": 0.0184, "step": 28100 }, { "epoch": 0.55, "learning_rate": 4.0912149266670554e-05, "loss": 0.0061, "step": 28102 }, { "epoch": 0.55, "learning_rate": 4.0911502306413317e-05, "loss": 0.0031, "step": 28104 }, { "epoch": 0.55, "learning_rate": 4.0910855346156086e-05, "loss": 0.0002, "step": 28106 }, { "epoch": 0.55, "learning_rate": 4.0910208385898855e-05, "loss": 0.0011, "step": 28108 }, { "epoch": 0.55, "learning_rate": 4.0909561425641624e-05, "loss": 0.0, "step": 28110 }, { "epoch": 0.55, "learning_rate": 4.090891446538439e-05, "loss": 0.0002, "step": 28112 }, { "epoch": 0.55, "learning_rate": 4.090826750512716e-05, "loss": 0.0, "step": 28114 }, { "epoch": 0.55, "learning_rate": 4.090762054486993e-05, "loss": 0.0, "step": 28116 }, { "epoch": 0.55, "learning_rate": 4.09069735846127e-05, "loss": 0.0026, "step": 28118 }, { "epoch": 0.55, "learning_rate": 4.090632662435547e-05, "loss": 0.0059, "step": 28120 }, { "epoch": 0.55, "learning_rate": 4.090567966409823e-05, "loss": 0.0004, "step": 28122 }, { "epoch": 0.55, "learning_rate": 4.090503270384101e-05, "loss": 0.0001, "step": 28124 }, { "epoch": 0.55, "learning_rate": 4.090438574358377e-05, "loss": 0.0003, "step": 28126 }, { "epoch": 0.55, "learning_rate": 4.090373878332654e-05, "loss": 0.0007, "step": 28128 }, { "epoch": 0.55, "learning_rate": 4.0903091823069315e-05, "loss": 0.0003, "step": 28130 }, { "epoch": 0.55, "learning_rate": 4.090244486281208e-05, "loss": 0.0001, "step": 28132 }, { "epoch": 0.55, "learning_rate": 4.090179790255485e-05, "loss": 0.0001, "step": 28134 }, { "epoch": 0.55, "learning_rate": 4.0901150942297615e-05, "loss": 0.0, "step": 28136 }, { "epoch": 0.55, "learning_rate": 4.0900503982040384e-05, "loss": 0.0188, "step": 28138 }, { "epoch": 0.55, "learning_rate": 4.089985702178315e-05, "loss": 0.0001, "step": 28140 }, { "epoch": 0.55, "learning_rate": 4.089921006152592e-05, "loss": 0.0004, "step": 28142 }, { "epoch": 0.55, "learning_rate": 4.089856310126869e-05, "loss": 0.0002, "step": 28144 }, { "epoch": 0.55, "learning_rate": 4.089791614101146e-05, "loss": 0.0169, "step": 28146 }, { "epoch": 0.55, "learning_rate": 4.089726918075423e-05, "loss": 0.0009, "step": 28148 }, { "epoch": 0.55, "learning_rate": 4.0896622220497e-05, "loss": 0.0, "step": 28150 }, { "epoch": 0.55, "learning_rate": 4.089597526023977e-05, "loss": 0.0052, "step": 28152 }, { "epoch": 0.55, "learning_rate": 4.089532829998253e-05, "loss": 0.0025, "step": 28154 }, { "epoch": 0.55, "learning_rate": 4.0894681339725306e-05, "loss": 0.0, "step": 28156 }, { "epoch": 0.55, "learning_rate": 4.089403437946807e-05, "loss": 0.0, "step": 28158 }, { "epoch": 0.55, "learning_rate": 4.089338741921084e-05, "loss": 0.0, "step": 28160 }, { "epoch": 0.55, "learning_rate": 4.089274045895361e-05, "loss": 0.0023, "step": 28162 }, { "epoch": 0.55, "learning_rate": 4.0892093498696376e-05, "loss": 0.0259, "step": 28164 }, { "epoch": 0.55, "learning_rate": 4.0891446538439145e-05, "loss": 0.0, "step": 28166 }, { "epoch": 0.55, "learning_rate": 4.0890799578181914e-05, "loss": 0.0051, "step": 28168 }, { "epoch": 0.55, "learning_rate": 4.089015261792468e-05, "loss": 0.001, "step": 28170 }, { "epoch": 0.55, "learning_rate": 4.088950565766745e-05, "loss": 0.0007, "step": 28172 }, { "epoch": 0.55, "learning_rate": 4.088885869741022e-05, "loss": 0.0001, "step": 28174 }, { "epoch": 0.55, "learning_rate": 4.088821173715299e-05, "loss": 0.0002, "step": 28176 }, { "epoch": 0.55, "learning_rate": 4.088756477689576e-05, "loss": 0.0001, "step": 28178 }, { "epoch": 0.55, "learning_rate": 4.088691781663853e-05, "loss": 0.0001, "step": 28180 }, { "epoch": 0.55, "learning_rate": 4.088627085638129e-05, "loss": 0.0031, "step": 28182 }, { "epoch": 0.55, "learning_rate": 4.0885623896124066e-05, "loss": 0.0084, "step": 28184 }, { "epoch": 0.55, "learning_rate": 4.088497693586683e-05, "loss": 0.0001, "step": 28186 }, { "epoch": 0.55, "learning_rate": 4.08843299756096e-05, "loss": 0.0002, "step": 28188 }, { "epoch": 0.55, "learning_rate": 4.088368301535237e-05, "loss": 0.0101, "step": 28190 }, { "epoch": 0.55, "learning_rate": 4.0883036055095136e-05, "loss": 0.0, "step": 28192 }, { "epoch": 0.55, "learning_rate": 4.088238909483791e-05, "loss": 0.0038, "step": 28194 }, { "epoch": 0.55, "learning_rate": 4.0881742134580674e-05, "loss": 0.0, "step": 28196 }, { "epoch": 0.55, "learning_rate": 4.088109517432344e-05, "loss": 0.0055, "step": 28198 }, { "epoch": 0.55, "learning_rate": 4.088044821406621e-05, "loss": 0.0013, "step": 28200 }, { "epoch": 0.55, "learning_rate": 4.087980125380898e-05, "loss": 0.0059, "step": 28202 }, { "epoch": 0.55, "learning_rate": 4.0879154293551744e-05, "loss": 0.0001, "step": 28204 }, { "epoch": 0.55, "learning_rate": 4.087850733329452e-05, "loss": 0.0012, "step": 28206 }, { "epoch": 0.55, "learning_rate": 4.087786037303729e-05, "loss": 0.0, "step": 28208 }, { "epoch": 0.55, "learning_rate": 4.087721341278005e-05, "loss": 0.0001, "step": 28210 }, { "epoch": 0.55, "learning_rate": 4.087656645252283e-05, "loss": 0.0002, "step": 28212 }, { "epoch": 0.55, "learning_rate": 4.087591949226559e-05, "loss": 0.0024, "step": 28214 }, { "epoch": 0.55, "learning_rate": 4.0875272532008365e-05, "loss": 0.0045, "step": 28216 }, { "epoch": 0.55, "learning_rate": 4.087462557175113e-05, "loss": 0.0056, "step": 28218 }, { "epoch": 0.55, "learning_rate": 4.0873978611493896e-05, "loss": 0.0001, "step": 28220 }, { "epoch": 0.55, "learning_rate": 4.0873331651236665e-05, "loss": 0.0001, "step": 28222 }, { "epoch": 0.55, "learning_rate": 4.0872684690979435e-05, "loss": 0.0001, "step": 28224 }, { "epoch": 0.55, "learning_rate": 4.0872037730722204e-05, "loss": 0.0133, "step": 28226 }, { "epoch": 0.55, "learning_rate": 4.087139077046497e-05, "loss": 0.0001, "step": 28228 }, { "epoch": 0.55, "learning_rate": 4.087074381020774e-05, "loss": 0.0234, "step": 28230 }, { "epoch": 0.55, "learning_rate": 4.087009684995051e-05, "loss": 0.0, "step": 28232 }, { "epoch": 0.55, "learning_rate": 4.086944988969328e-05, "loss": 0.0003, "step": 28234 }, { "epoch": 0.55, "learning_rate": 4.086880292943604e-05, "loss": 0.0004, "step": 28236 }, { "epoch": 0.55, "learning_rate": 4.086815596917882e-05, "loss": 0.0, "step": 28238 }, { "epoch": 0.55, "learning_rate": 4.086750900892159e-05, "loss": 0.0, "step": 28240 }, { "epoch": 0.55, "learning_rate": 4.086686204866435e-05, "loss": 0.0018, "step": 28242 }, { "epoch": 0.55, "learning_rate": 4.0866215088407125e-05, "loss": 0.0001, "step": 28244 }, { "epoch": 0.55, "learning_rate": 4.086556812814989e-05, "loss": 0.0001, "step": 28246 }, { "epoch": 0.55, "learning_rate": 4.086492116789266e-05, "loss": 0.0, "step": 28248 }, { "epoch": 0.55, "learning_rate": 4.0864274207635426e-05, "loss": 0.0002, "step": 28250 }, { "epoch": 0.55, "learning_rate": 4.0863627247378195e-05, "loss": 0.0083, "step": 28252 }, { "epoch": 0.55, "learning_rate": 4.0862980287120964e-05, "loss": 0.0, "step": 28254 }, { "epoch": 0.55, "learning_rate": 4.086233332686373e-05, "loss": 0.0073, "step": 28256 }, { "epoch": 0.55, "learning_rate": 4.08616863666065e-05, "loss": 0.0001, "step": 28258 }, { "epoch": 0.55, "learning_rate": 4.086103940634927e-05, "loss": 0.0001, "step": 28260 }, { "epoch": 0.55, "learning_rate": 4.086039244609204e-05, "loss": 0.0, "step": 28262 }, { "epoch": 0.55, "learning_rate": 4.08597454858348e-05, "loss": 0.0045, "step": 28264 }, { "epoch": 0.55, "learning_rate": 4.085909852557758e-05, "loss": 0.0005, "step": 28266 }, { "epoch": 0.55, "learning_rate": 4.085845156532034e-05, "loss": 0.0003, "step": 28268 }, { "epoch": 0.55, "learning_rate": 4.085780460506311e-05, "loss": 0.0024, "step": 28270 }, { "epoch": 0.55, "learning_rate": 4.085715764480588e-05, "loss": 0.0001, "step": 28272 }, { "epoch": 0.55, "learning_rate": 4.085651068454865e-05, "loss": 0.0001, "step": 28274 }, { "epoch": 0.55, "learning_rate": 4.0855863724291424e-05, "loss": 0.0, "step": 28276 }, { "epoch": 0.55, "learning_rate": 4.0855216764034186e-05, "loss": 0.0, "step": 28278 }, { "epoch": 0.55, "learning_rate": 4.0854569803776955e-05, "loss": 0.0014, "step": 28280 }, { "epoch": 0.55, "learning_rate": 4.0853922843519724e-05, "loss": 0.002, "step": 28282 }, { "epoch": 0.55, "learning_rate": 4.0853275883262494e-05, "loss": 0.0001, "step": 28284 }, { "epoch": 0.55, "learning_rate": 4.085262892300526e-05, "loss": 0.0004, "step": 28286 }, { "epoch": 0.55, "learning_rate": 4.085198196274803e-05, "loss": 0.0, "step": 28288 }, { "epoch": 0.55, "learning_rate": 4.0851658482619416e-05, "loss": 0.0138, "step": 28290 }, { "epoch": 0.55, "learning_rate": 4.085101152236218e-05, "loss": 0.0029, "step": 28292 }, { "epoch": 0.55, "learning_rate": 4.0850364562104954e-05, "loss": 0.0, "step": 28294 }, { "epoch": 0.55, "learning_rate": 4.0849717601847723e-05, "loss": 0.0031, "step": 28296 }, { "epoch": 0.55, "learning_rate": 4.0849070641590486e-05, "loss": 0.0, "step": 28298 }, { "epoch": 0.55, "learning_rate": 4.084842368133326e-05, "loss": 0.0005, "step": 28300 }, { "epoch": 0.55, "learning_rate": 4.0847776721076024e-05, "loss": 0.0001, "step": 28302 }, { "epoch": 0.55, "learning_rate": 4.08471297608188e-05, "loss": 0.0001, "step": 28304 }, { "epoch": 0.55, "learning_rate": 4.084648280056156e-05, "loss": 0.0356, "step": 28306 }, { "epoch": 0.55, "learning_rate": 4.084583584030433e-05, "loss": 0.0018, "step": 28308 }, { "epoch": 0.55, "learning_rate": 4.08451888800471e-05, "loss": 0.0, "step": 28310 }, { "epoch": 0.55, "learning_rate": 4.084454191978987e-05, "loss": 0.0046, "step": 28312 }, { "epoch": 0.55, "learning_rate": 4.084389495953264e-05, "loss": 0.0013, "step": 28314 }, { "epoch": 0.55, "learning_rate": 4.084324799927541e-05, "loss": 0.0, "step": 28316 }, { "epoch": 0.55, "learning_rate": 4.084260103901818e-05, "loss": 0.0081, "step": 28318 }, { "epoch": 0.55, "learning_rate": 4.084195407876094e-05, "loss": 0.0001, "step": 28320 }, { "epoch": 0.55, "learning_rate": 4.0841307118503715e-05, "loss": 0.0001, "step": 28322 }, { "epoch": 0.55, "learning_rate": 4.084066015824648e-05, "loss": 0.0001, "step": 28324 }, { "epoch": 0.55, "learning_rate": 4.084001319798925e-05, "loss": 0.0, "step": 28326 }, { "epoch": 0.55, "learning_rate": 4.0839366237732015e-05, "loss": 0.0016, "step": 28328 }, { "epoch": 0.55, "learning_rate": 4.0838719277474784e-05, "loss": 0.0006, "step": 28330 }, { "epoch": 0.55, "learning_rate": 4.083807231721756e-05, "loss": 0.0037, "step": 28332 }, { "epoch": 0.55, "learning_rate": 4.083742535696032e-05, "loss": 0.0001, "step": 28334 }, { "epoch": 0.55, "learning_rate": 4.083677839670309e-05, "loss": 0.0, "step": 28336 }, { "epoch": 0.55, "learning_rate": 4.083613143644586e-05, "loss": 0.0003, "step": 28338 }, { "epoch": 0.55, "learning_rate": 4.083548447618863e-05, "loss": 0.0064, "step": 28340 }, { "epoch": 0.55, "learning_rate": 4.08348375159314e-05, "loss": 0.0009, "step": 28342 }, { "epoch": 0.55, "learning_rate": 4.083419055567417e-05, "loss": 0.0, "step": 28344 }, { "epoch": 0.55, "learning_rate": 4.083354359541694e-05, "loss": 0.0053, "step": 28346 }, { "epoch": 0.55, "learning_rate": 4.0832896635159706e-05, "loss": 0.0, "step": 28348 }, { "epoch": 0.55, "learning_rate": 4.0832249674902475e-05, "loss": 0.0003, "step": 28350 }, { "epoch": 0.55, "learning_rate": 4.083160271464524e-05, "loss": 0.0141, "step": 28352 }, { "epoch": 0.55, "learning_rate": 4.083095575438801e-05, "loss": 0.0142, "step": 28354 }, { "epoch": 0.55, "learning_rate": 4.0830308794130776e-05, "loss": 0.0148, "step": 28356 }, { "epoch": 0.55, "learning_rate": 4.0829661833873545e-05, "loss": 0.0, "step": 28358 }, { "epoch": 0.55, "learning_rate": 4.0829014873616314e-05, "loss": 0.0001, "step": 28360 }, { "epoch": 0.55, "learning_rate": 4.082836791335908e-05, "loss": 0.0, "step": 28362 }, { "epoch": 0.55, "learning_rate": 4.082772095310186e-05, "loss": 0.0005, "step": 28364 }, { "epoch": 0.55, "learning_rate": 4.082707399284462e-05, "loss": 0.0001, "step": 28366 }, { "epoch": 0.55, "learning_rate": 4.082642703258739e-05, "loss": 0.0005, "step": 28368 }, { "epoch": 0.55, "learning_rate": 4.082578007233016e-05, "loss": 0.002, "step": 28370 }, { "epoch": 0.55, "learning_rate": 4.082513311207293e-05, "loss": 0.0051, "step": 28372 }, { "epoch": 0.55, "learning_rate": 4.082448615181569e-05, "loss": 0.0289, "step": 28374 }, { "epoch": 0.55, "learning_rate": 4.0823839191558467e-05, "loss": 0.0014, "step": 28376 }, { "epoch": 0.55, "learning_rate": 4.0823192231301236e-05, "loss": 0.0001, "step": 28378 }, { "epoch": 0.55, "learning_rate": 4.0822545271044e-05, "loss": 0.0, "step": 28380 }, { "epoch": 0.55, "learning_rate": 4.0821898310786774e-05, "loss": 0.0001, "step": 28382 }, { "epoch": 0.55, "learning_rate": 4.0821251350529536e-05, "loss": 0.0002, "step": 28384 }, { "epoch": 0.55, "learning_rate": 4.082060439027231e-05, "loss": 0.0008, "step": 28386 }, { "epoch": 0.55, "learning_rate": 4.0819957430015074e-05, "loss": 0.0, "step": 28388 }, { "epoch": 0.55, "learning_rate": 4.081931046975784e-05, "loss": 0.0026, "step": 28390 }, { "epoch": 0.55, "learning_rate": 4.081866350950061e-05, "loss": 0.0001, "step": 28392 }, { "epoch": 0.55, "learning_rate": 4.081801654924338e-05, "loss": 0.0021, "step": 28394 }, { "epoch": 0.55, "learning_rate": 4.081736958898615e-05, "loss": 0.0, "step": 28396 }, { "epoch": 0.55, "learning_rate": 4.081672262872892e-05, "loss": 0.0001, "step": 28398 }, { "epoch": 0.55, "learning_rate": 4.081607566847169e-05, "loss": 0.0001, "step": 28400 }, { "epoch": 0.55, "learning_rate": 4.081542870821445e-05, "loss": 0.0001, "step": 28402 }, { "epoch": 0.55, "learning_rate": 4.081478174795723e-05, "loss": 0.0, "step": 28404 }, { "epoch": 0.55, "learning_rate": 4.081413478769999e-05, "loss": 0.0067, "step": 28406 }, { "epoch": 0.55, "learning_rate": 4.0813487827442765e-05, "loss": 0.0, "step": 28408 }, { "epoch": 0.55, "learning_rate": 4.0812840867185534e-05, "loss": 0.0001, "step": 28410 }, { "epoch": 0.55, "learning_rate": 4.0812193906928296e-05, "loss": 0.0005, "step": 28412 }, { "epoch": 0.55, "learning_rate": 4.081154694667107e-05, "loss": 0.0, "step": 28414 }, { "epoch": 0.55, "learning_rate": 4.0810899986413835e-05, "loss": 0.0, "step": 28416 }, { "epoch": 0.55, "learning_rate": 4.0810253026156604e-05, "loss": 0.0021, "step": 28418 }, { "epoch": 0.55, "learning_rate": 4.080960606589937e-05, "loss": 0.0, "step": 28420 }, { "epoch": 0.55, "learning_rate": 4.080895910564214e-05, "loss": 0.0, "step": 28422 }, { "epoch": 0.55, "learning_rate": 4.080831214538491e-05, "loss": 0.0028, "step": 28424 }, { "epoch": 0.55, "learning_rate": 4.080766518512768e-05, "loss": 0.0003, "step": 28426 }, { "epoch": 0.55, "learning_rate": 4.080701822487045e-05, "loss": 0.0, "step": 28428 }, { "epoch": 0.55, "learning_rate": 4.080637126461322e-05, "loss": 0.0021, "step": 28430 }, { "epoch": 0.55, "learning_rate": 4.080572430435599e-05, "loss": 0.0001, "step": 28432 }, { "epoch": 0.55, "learning_rate": 4.080507734409875e-05, "loss": 0.0015, "step": 28434 }, { "epoch": 0.55, "learning_rate": 4.0804430383841525e-05, "loss": 0.0001, "step": 28436 }, { "epoch": 0.55, "learning_rate": 4.080378342358429e-05, "loss": 0.0049, "step": 28438 }, { "epoch": 0.55, "learning_rate": 4.080313646332706e-05, "loss": 0.0016, "step": 28440 }, { "epoch": 0.55, "learning_rate": 4.080248950306983e-05, "loss": 0.0019, "step": 28442 }, { "epoch": 0.55, "learning_rate": 4.0801842542812595e-05, "loss": 0.0, "step": 28444 }, { "epoch": 0.55, "learning_rate": 4.080119558255537e-05, "loss": 0.0077, "step": 28446 }, { "epoch": 0.55, "learning_rate": 4.080054862229813e-05, "loss": 0.0001, "step": 28448 }, { "epoch": 0.55, "learning_rate": 4.07999016620409e-05, "loss": 0.0, "step": 28450 }, { "epoch": 0.55, "learning_rate": 4.079925470178367e-05, "loss": 0.0304, "step": 28452 }, { "epoch": 0.55, "learning_rate": 4.079860774152644e-05, "loss": 0.0, "step": 28454 }, { "epoch": 0.55, "learning_rate": 4.079796078126921e-05, "loss": 0.0037, "step": 28456 }, { "epoch": 0.55, "learning_rate": 4.079731382101198e-05, "loss": 0.0011, "step": 28458 }, { "epoch": 0.55, "learning_rate": 4.079666686075475e-05, "loss": 0.0, "step": 28460 }, { "epoch": 0.55, "learning_rate": 4.079601990049751e-05, "loss": 0.0, "step": 28462 }, { "epoch": 0.55, "learning_rate": 4.0795372940240286e-05, "loss": 0.0094, "step": 28464 }, { "epoch": 0.55, "learning_rate": 4.079472597998305e-05, "loss": 0.0037, "step": 28466 }, { "epoch": 0.55, "learning_rate": 4.0794079019725824e-05, "loss": 0.0129, "step": 28468 }, { "epoch": 0.55, "learning_rate": 4.0793432059468586e-05, "loss": 0.0001, "step": 28470 }, { "epoch": 0.55, "learning_rate": 4.0792785099211355e-05, "loss": 0.0001, "step": 28472 }, { "epoch": 0.55, "learning_rate": 4.079213813895413e-05, "loss": 0.0004, "step": 28474 }, { "epoch": 0.55, "learning_rate": 4.0791491178696894e-05, "loss": 0.0014, "step": 28476 }, { "epoch": 0.55, "learning_rate": 4.079084421843966e-05, "loss": 0.0039, "step": 28478 }, { "epoch": 0.55, "learning_rate": 4.079019725818243e-05, "loss": 0.0138, "step": 28480 }, { "epoch": 0.55, "learning_rate": 4.07895502979252e-05, "loss": 0.0006, "step": 28482 }, { "epoch": 0.55, "learning_rate": 4.078890333766796e-05, "loss": 0.0001, "step": 28484 }, { "epoch": 0.55, "learning_rate": 4.078825637741074e-05, "loss": 0.0009, "step": 28486 }, { "epoch": 0.55, "learning_rate": 4.078760941715351e-05, "loss": 0.0, "step": 28488 }, { "epoch": 0.55, "learning_rate": 4.078696245689628e-05, "loss": 0.0002, "step": 28490 }, { "epoch": 0.55, "learning_rate": 4.0786315496639046e-05, "loss": 0.0, "step": 28492 }, { "epoch": 0.55, "learning_rate": 4.078566853638181e-05, "loss": 0.0015, "step": 28494 }, { "epoch": 0.55, "learning_rate": 4.0785021576124584e-05, "loss": 0.0, "step": 28496 }, { "epoch": 0.55, "learning_rate": 4.078437461586735e-05, "loss": 0.0005, "step": 28498 }, { "epoch": 0.55, "learning_rate": 4.0783727655610116e-05, "loss": 0.0001, "step": 28500 }, { "epoch": 0.55, "learning_rate": 4.0783080695352885e-05, "loss": 0.0, "step": 28502 }, { "epoch": 0.55, "learning_rate": 4.0782433735095654e-05, "loss": 0.0, "step": 28504 }, { "epoch": 0.55, "learning_rate": 4.078178677483842e-05, "loss": 0.0018, "step": 28506 }, { "epoch": 0.55, "learning_rate": 4.078113981458119e-05, "loss": 0.0091, "step": 28508 }, { "epoch": 0.55, "learning_rate": 4.078049285432396e-05, "loss": 0.0017, "step": 28510 }, { "epoch": 0.55, "learning_rate": 4.077984589406673e-05, "loss": 0.0001, "step": 28512 }, { "epoch": 0.55, "learning_rate": 4.07791989338095e-05, "loss": 0.0002, "step": 28514 }, { "epoch": 0.55, "learning_rate": 4.077855197355226e-05, "loss": 0.0024, "step": 28516 }, { "epoch": 0.55, "learning_rate": 4.077790501329504e-05, "loss": 0.0001, "step": 28518 }, { "epoch": 0.55, "learning_rate": 4.077725805303781e-05, "loss": 0.0, "step": 28520 }, { "epoch": 0.55, "learning_rate": 4.077661109278057e-05, "loss": 0.0, "step": 28522 }, { "epoch": 0.55, "learning_rate": 4.0775964132523345e-05, "loss": 0.0, "step": 28524 }, { "epoch": 0.55, "learning_rate": 4.077531717226611e-05, "loss": 0.0011, "step": 28526 }, { "epoch": 0.55, "learning_rate": 4.077467021200888e-05, "loss": 0.0067, "step": 28528 }, { "epoch": 0.55, "learning_rate": 4.0774023251751645e-05, "loss": 0.0001, "step": 28530 }, { "epoch": 0.55, "learning_rate": 4.0773376291494414e-05, "loss": 0.0087, "step": 28532 }, { "epoch": 0.55, "learning_rate": 4.0772729331237184e-05, "loss": 0.0007, "step": 28534 }, { "epoch": 0.55, "learning_rate": 4.077208237097995e-05, "loss": 0.0001, "step": 28536 }, { "epoch": 0.55, "learning_rate": 4.077143541072272e-05, "loss": 0.0015, "step": 28538 }, { "epoch": 0.55, "learning_rate": 4.077078845046549e-05, "loss": 0.0005, "step": 28540 }, { "epoch": 0.55, "learning_rate": 4.077014149020826e-05, "loss": 0.008, "step": 28542 }, { "epoch": 0.55, "learning_rate": 4.076949452995102e-05, "loss": 0.003, "step": 28544 }, { "epoch": 0.55, "learning_rate": 4.07688475696938e-05, "loss": 0.0001, "step": 28546 }, { "epoch": 0.55, "learning_rate": 4.076820060943656e-05, "loss": 0.0007, "step": 28548 }, { "epoch": 0.55, "learning_rate": 4.0767553649179336e-05, "loss": 0.0333, "step": 28550 }, { "epoch": 0.55, "learning_rate": 4.07669066889221e-05, "loss": 0.0001, "step": 28552 }, { "epoch": 0.55, "learning_rate": 4.076625972866487e-05, "loss": 0.0, "step": 28554 }, { "epoch": 0.55, "learning_rate": 4.0765612768407643e-05, "loss": 0.0051, "step": 28556 }, { "epoch": 0.55, "learning_rate": 4.0764965808150406e-05, "loss": 0.0, "step": 28558 }, { "epoch": 0.55, "learning_rate": 4.0764318847893175e-05, "loss": 0.0001, "step": 28560 }, { "epoch": 0.55, "learning_rate": 4.0763671887635944e-05, "loss": 0.0001, "step": 28562 }, { "epoch": 0.55, "learning_rate": 4.076302492737871e-05, "loss": 0.0009, "step": 28564 }, { "epoch": 0.55, "learning_rate": 4.076237796712148e-05, "loss": 0.0027, "step": 28566 }, { "epoch": 0.55, "learning_rate": 4.076173100686425e-05, "loss": 0.0004, "step": 28568 }, { "epoch": 0.55, "learning_rate": 4.076108404660702e-05, "loss": 0.0001, "step": 28570 }, { "epoch": 0.55, "learning_rate": 4.076043708634979e-05, "loss": 0.0001, "step": 28572 }, { "epoch": 0.55, "learning_rate": 4.075979012609256e-05, "loss": 0.0045, "step": 28574 }, { "epoch": 0.55, "learning_rate": 4.075914316583532e-05, "loss": 0.0001, "step": 28576 }, { "epoch": 0.55, "learning_rate": 4.0758496205578097e-05, "loss": 0.0, "step": 28578 }, { "epoch": 0.55, "learning_rate": 4.075784924532086e-05, "loss": 0.0113, "step": 28580 }, { "epoch": 0.55, "learning_rate": 4.075720228506363e-05, "loss": 0.0006, "step": 28582 }, { "epoch": 0.55, "learning_rate": 4.07565553248064e-05, "loss": 0.0044, "step": 28584 }, { "epoch": 0.55, "learning_rate": 4.0755908364549166e-05, "loss": 0.0001, "step": 28586 }, { "epoch": 0.55, "learning_rate": 4.075526140429194e-05, "loss": 0.0039, "step": 28588 }, { "epoch": 0.55, "learning_rate": 4.0754614444034704e-05, "loss": 0.0001, "step": 28590 }, { "epoch": 0.55, "learning_rate": 4.0753967483777473e-05, "loss": 0.0, "step": 28592 }, { "epoch": 0.55, "learning_rate": 4.075332052352024e-05, "loss": 0.003, "step": 28594 }, { "epoch": 0.56, "learning_rate": 4.075267356326301e-05, "loss": 0.0, "step": 28596 }, { "epoch": 0.56, "learning_rate": 4.075202660300578e-05, "loss": 0.0001, "step": 28598 }, { "epoch": 0.56, "learning_rate": 4.075137964274855e-05, "loss": 0.0, "step": 28600 }, { "epoch": 0.56, "learning_rate": 4.075073268249132e-05, "loss": 0.0, "step": 28602 }, { "epoch": 0.56, "learning_rate": 4.075008572223408e-05, "loss": 0.0004, "step": 28604 }, { "epoch": 0.56, "learning_rate": 4.074943876197686e-05, "loss": 0.0037, "step": 28606 }, { "epoch": 0.56, "learning_rate": 4.074879180171962e-05, "loss": 0.0159, "step": 28608 }, { "epoch": 0.56, "learning_rate": 4.0748144841462395e-05, "loss": 0.0149, "step": 28610 }, { "epoch": 0.56, "learning_rate": 4.074749788120516e-05, "loss": 0.0002, "step": 28612 }, { "epoch": 0.56, "learning_rate": 4.0746850920947927e-05, "loss": 0.0005, "step": 28614 }, { "epoch": 0.56, "learning_rate": 4.0746203960690696e-05, "loss": 0.003, "step": 28616 }, { "epoch": 0.56, "learning_rate": 4.0745557000433465e-05, "loss": 0.0035, "step": 28618 }, { "epoch": 0.56, "learning_rate": 4.0744910040176234e-05, "loss": 0.0001, "step": 28620 }, { "epoch": 0.56, "learning_rate": 4.0744263079919e-05, "loss": 0.0124, "step": 28622 }, { "epoch": 0.56, "learning_rate": 4.074361611966177e-05, "loss": 0.0001, "step": 28624 }, { "epoch": 0.56, "learning_rate": 4.0742969159404534e-05, "loss": 0.0, "step": 28626 }, { "epoch": 0.56, "learning_rate": 4.074232219914731e-05, "loss": 0.0001, "step": 28628 }, { "epoch": 0.56, "learning_rate": 4.074167523889007e-05, "loss": 0.0002, "step": 28630 }, { "epoch": 0.56, "learning_rate": 4.074102827863285e-05, "loss": 0.0063, "step": 28632 }, { "epoch": 0.56, "learning_rate": 4.074038131837562e-05, "loss": 0.0, "step": 28634 }, { "epoch": 0.56, "learning_rate": 4.073973435811838e-05, "loss": 0.0, "step": 28636 }, { "epoch": 0.56, "learning_rate": 4.0739087397861156e-05, "loss": 0.0, "step": 28638 }, { "epoch": 0.56, "learning_rate": 4.073844043760392e-05, "loss": 0.0113, "step": 28640 }, { "epoch": 0.56, "learning_rate": 4.073779347734669e-05, "loss": 0.0001, "step": 28642 }, { "epoch": 0.56, "learning_rate": 4.0737146517089456e-05, "loss": 0.0001, "step": 28644 }, { "epoch": 0.56, "learning_rate": 4.0736499556832225e-05, "loss": 0.0001, "step": 28646 }, { "epoch": 0.56, "learning_rate": 4.0735852596574994e-05, "loss": 0.0043, "step": 28648 }, { "epoch": 0.56, "learning_rate": 4.073520563631776e-05, "loss": 0.0002, "step": 28650 }, { "epoch": 0.56, "learning_rate": 4.073455867606053e-05, "loss": 0.0039, "step": 28652 }, { "epoch": 0.56, "learning_rate": 4.07339117158033e-05, "loss": 0.0004, "step": 28654 }, { "epoch": 0.56, "learning_rate": 4.073326475554607e-05, "loss": 0.0021, "step": 28656 }, { "epoch": 0.56, "learning_rate": 4.073261779528883e-05, "loss": 0.001, "step": 28658 }, { "epoch": 0.56, "learning_rate": 4.073197083503161e-05, "loss": 0.0, "step": 28660 }, { "epoch": 0.56, "learning_rate": 4.073132387477437e-05, "loss": 0.0002, "step": 28662 }, { "epoch": 0.56, "learning_rate": 4.073067691451714e-05, "loss": 0.0001, "step": 28664 }, { "epoch": 0.56, "learning_rate": 4.0730029954259916e-05, "loss": 0.0003, "step": 28666 }, { "epoch": 0.56, "learning_rate": 4.072938299400268e-05, "loss": 0.0001, "step": 28668 }, { "epoch": 0.56, "learning_rate": 4.0728736033745454e-05, "loss": 0.0087, "step": 28670 }, { "epoch": 0.56, "learning_rate": 4.0728089073488216e-05, "loss": 0.0, "step": 28672 }, { "epoch": 0.56, "learning_rate": 4.0727442113230986e-05, "loss": 0.0025, "step": 28674 }, { "epoch": 0.56, "learning_rate": 4.0726795152973755e-05, "loss": 0.0, "step": 28676 }, { "epoch": 0.56, "learning_rate": 4.0726148192716524e-05, "loss": 0.0001, "step": 28678 }, { "epoch": 0.56, "learning_rate": 4.072550123245929e-05, "loss": 0.0003, "step": 28680 }, { "epoch": 0.56, "learning_rate": 4.072485427220206e-05, "loss": 0.0021, "step": 28682 }, { "epoch": 0.56, "learning_rate": 4.072420731194483e-05, "loss": 0.0009, "step": 28684 }, { "epoch": 0.56, "learning_rate": 4.072356035168759e-05, "loss": 0.0025, "step": 28686 }, { "epoch": 0.56, "learning_rate": 4.072291339143037e-05, "loss": 0.0002, "step": 28688 }, { "epoch": 0.56, "learning_rate": 4.072226643117313e-05, "loss": 0.0065, "step": 28690 }, { "epoch": 0.56, "learning_rate": 4.072161947091591e-05, "loss": 0.0, "step": 28692 }, { "epoch": 0.56, "learning_rate": 4.072097251065867e-05, "loss": 0.0009, "step": 28694 }, { "epoch": 0.56, "learning_rate": 4.072032555040144e-05, "loss": 0.0, "step": 28696 }, { "epoch": 0.56, "learning_rate": 4.0719678590144215e-05, "loss": 0.0008, "step": 28698 }, { "epoch": 0.56, "learning_rate": 4.071903162988698e-05, "loss": 0.0175, "step": 28700 }, { "epoch": 0.56, "learning_rate": 4.0718384669629746e-05, "loss": 0.0, "step": 28702 }, { "epoch": 0.56, "learning_rate": 4.0717737709372515e-05, "loss": 0.0005, "step": 28704 }, { "epoch": 0.56, "learning_rate": 4.0717090749115284e-05, "loss": 0.0004, "step": 28706 }, { "epoch": 0.56, "learning_rate": 4.0716443788858046e-05, "loss": 0.0015, "step": 28708 }, { "epoch": 0.56, "learning_rate": 4.071579682860082e-05, "loss": 0.0, "step": 28710 }, { "epoch": 0.56, "learning_rate": 4.071514986834359e-05, "loss": 0.0036, "step": 28712 }, { "epoch": 0.56, "learning_rate": 4.071450290808636e-05, "loss": 0.002, "step": 28714 }, { "epoch": 0.56, "learning_rate": 4.071385594782913e-05, "loss": 0.0001, "step": 28716 }, { "epoch": 0.56, "learning_rate": 4.071320898757189e-05, "loss": 0.0, "step": 28718 }, { "epoch": 0.56, "learning_rate": 4.071256202731467e-05, "loss": 0.0002, "step": 28720 }, { "epoch": 0.56, "learning_rate": 4.071191506705743e-05, "loss": 0.0, "step": 28722 }, { "epoch": 0.56, "learning_rate": 4.07112681068002e-05, "loss": 0.0005, "step": 28724 }, { "epoch": 0.56, "learning_rate": 4.071062114654297e-05, "loss": 0.0104, "step": 28726 }, { "epoch": 0.56, "learning_rate": 4.070997418628574e-05, "loss": 0.0, "step": 28728 }, { "epoch": 0.56, "learning_rate": 4.0709327226028506e-05, "loss": 0.0001, "step": 28730 }, { "epoch": 0.56, "learning_rate": 4.0708680265771275e-05, "loss": 0.0, "step": 28732 }, { "epoch": 0.56, "learning_rate": 4.0708033305514045e-05, "loss": 0.0019, "step": 28734 }, { "epoch": 0.56, "learning_rate": 4.0707386345256814e-05, "loss": 0.0019, "step": 28736 }, { "epoch": 0.56, "learning_rate": 4.070673938499958e-05, "loss": 0.0001, "step": 28738 }, { "epoch": 0.56, "learning_rate": 4.0706092424742345e-05, "loss": 0.0753, "step": 28740 }, { "epoch": 0.56, "learning_rate": 4.070544546448512e-05, "loss": 0.0019, "step": 28742 }, { "epoch": 0.56, "learning_rate": 4.070479850422789e-05, "loss": 0.0015, "step": 28744 }, { "epoch": 0.56, "learning_rate": 4.070415154397065e-05, "loss": 0.0023, "step": 28746 }, { "epoch": 0.56, "learning_rate": 4.070350458371343e-05, "loss": 0.0, "step": 28748 }, { "epoch": 0.56, "learning_rate": 4.070285762345619e-05, "loss": 0.0098, "step": 28750 }, { "epoch": 0.56, "learning_rate": 4.0702210663198966e-05, "loss": 0.0075, "step": 28752 }, { "epoch": 0.56, "learning_rate": 4.070156370294173e-05, "loss": 0.0, "step": 28754 }, { "epoch": 0.56, "learning_rate": 4.07009167426845e-05, "loss": 0.0028, "step": 28756 }, { "epoch": 0.56, "learning_rate": 4.070026978242727e-05, "loss": 0.0218, "step": 28758 }, { "epoch": 0.56, "learning_rate": 4.0699622822170036e-05, "loss": 0.0643, "step": 28760 }, { "epoch": 0.56, "learning_rate": 4.0698975861912805e-05, "loss": 0.001, "step": 28762 }, { "epoch": 0.56, "learning_rate": 4.0698328901655574e-05, "loss": 0.0001, "step": 28764 }, { "epoch": 0.56, "learning_rate": 4.069768194139834e-05, "loss": 0.0, "step": 28766 }, { "epoch": 0.56, "learning_rate": 4.0697034981141105e-05, "loss": 0.0156, "step": 28768 }, { "epoch": 0.56, "learning_rate": 4.069638802088388e-05, "loss": 0.0001, "step": 28770 }, { "epoch": 0.56, "learning_rate": 4.0695741060626644e-05, "loss": 0.0001, "step": 28772 }, { "epoch": 0.56, "learning_rate": 4.069509410036942e-05, "loss": 0.0001, "step": 28774 }, { "epoch": 0.56, "learning_rate": 4.069444714011219e-05, "loss": 0.0002, "step": 28776 }, { "epoch": 0.56, "learning_rate": 4.069380017985495e-05, "loss": 0.0019, "step": 28778 }, { "epoch": 0.56, "learning_rate": 4.069315321959773e-05, "loss": 0.0001, "step": 28780 }, { "epoch": 0.56, "learning_rate": 4.069250625934049e-05, "loss": 0.0002, "step": 28782 }, { "epoch": 0.56, "learning_rate": 4.069185929908326e-05, "loss": 0.0025, "step": 28784 }, { "epoch": 0.56, "learning_rate": 4.069121233882603e-05, "loss": 0.0002, "step": 28786 }, { "epoch": 0.56, "learning_rate": 4.0690565378568796e-05, "loss": 0.0015, "step": 28788 }, { "epoch": 0.56, "learning_rate": 4.0689918418311565e-05, "loss": 0.0004, "step": 28790 }, { "epoch": 0.56, "learning_rate": 4.0689271458054334e-05, "loss": 0.0003, "step": 28792 }, { "epoch": 0.56, "learning_rate": 4.0688624497797104e-05, "loss": 0.0002, "step": 28794 }, { "epoch": 0.56, "learning_rate": 4.068797753753987e-05, "loss": 0.0001, "step": 28796 }, { "epoch": 0.56, "learning_rate": 4.068733057728264e-05, "loss": 0.0001, "step": 28798 }, { "epoch": 0.56, "learning_rate": 4.0686683617025404e-05, "loss": 0.0, "step": 28800 }, { "epoch": 0.56, "learning_rate": 4.068603665676818e-05, "loss": 0.0, "step": 28802 }, { "epoch": 0.56, "learning_rate": 4.068538969651094e-05, "loss": 0.0005, "step": 28804 }, { "epoch": 0.56, "learning_rate": 4.068474273625371e-05, "loss": 0.0, "step": 28806 }, { "epoch": 0.56, "learning_rate": 4.068409577599648e-05, "loss": 0.0003, "step": 28808 }, { "epoch": 0.56, "learning_rate": 4.068344881573925e-05, "loss": 0.0054, "step": 28810 }, { "epoch": 0.56, "learning_rate": 4.0682801855482025e-05, "loss": 0.0001, "step": 28812 }, { "epoch": 0.56, "learning_rate": 4.068215489522479e-05, "loss": 0.0, "step": 28814 }, { "epoch": 0.56, "learning_rate": 4.068150793496756e-05, "loss": 0.001, "step": 28816 }, { "epoch": 0.56, "learning_rate": 4.0680860974710326e-05, "loss": 0.0088, "step": 28818 }, { "epoch": 0.56, "learning_rate": 4.0680214014453095e-05, "loss": 0.0001, "step": 28820 }, { "epoch": 0.56, "learning_rate": 4.0679567054195864e-05, "loss": 0.0063, "step": 28822 }, { "epoch": 0.56, "learning_rate": 4.067892009393863e-05, "loss": 0.0001, "step": 28824 }, { "epoch": 0.56, "learning_rate": 4.06782731336814e-05, "loss": 0.0003, "step": 28826 }, { "epoch": 0.56, "learning_rate": 4.0677626173424164e-05, "loss": 0.0002, "step": 28828 }, { "epoch": 0.56, "learning_rate": 4.067697921316694e-05, "loss": 0.0, "step": 28830 }, { "epoch": 0.56, "learning_rate": 4.06763322529097e-05, "loss": 0.0006, "step": 28832 }, { "epoch": 0.56, "learning_rate": 4.067568529265248e-05, "loss": 0.0008, "step": 28834 }, { "epoch": 0.56, "learning_rate": 4.067503833239524e-05, "loss": 0.0, "step": 28836 }, { "epoch": 0.56, "learning_rate": 4.067439137213801e-05, "loss": 0.0, "step": 28838 }, { "epoch": 0.56, "learning_rate": 4.067374441188078e-05, "loss": 0.0001, "step": 28840 }, { "epoch": 0.56, "learning_rate": 4.067309745162355e-05, "loss": 0.0065, "step": 28842 }, { "epoch": 0.56, "learning_rate": 4.067245049136632e-05, "loss": 0.0063, "step": 28844 }, { "epoch": 0.56, "learning_rate": 4.0671803531109086e-05, "loss": 0.0402, "step": 28846 }, { "epoch": 0.56, "learning_rate": 4.0671156570851855e-05, "loss": 0.0, "step": 28848 }, { "epoch": 0.56, "learning_rate": 4.067050961059462e-05, "loss": 0.0019, "step": 28850 }, { "epoch": 0.56, "learning_rate": 4.0669862650337393e-05, "loss": 0.0012, "step": 28852 }, { "epoch": 0.56, "learning_rate": 4.0669215690080156e-05, "loss": 0.0, "step": 28854 }, { "epoch": 0.56, "learning_rate": 4.066856872982293e-05, "loss": 0.0004, "step": 28856 }, { "epoch": 0.56, "learning_rate": 4.06679217695657e-05, "loss": 0.0665, "step": 28858 }, { "epoch": 0.56, "learning_rate": 4.066727480930846e-05, "loss": 0.0009, "step": 28860 }, { "epoch": 0.56, "learning_rate": 4.066662784905124e-05, "loss": 0.0009, "step": 28862 }, { "epoch": 0.56, "learning_rate": 4.0665980888794e-05, "loss": 0.0, "step": 28864 }, { "epoch": 0.56, "learning_rate": 4.066533392853677e-05, "loss": 0.0015, "step": 28866 }, { "epoch": 0.56, "learning_rate": 4.066468696827954e-05, "loss": 0.0, "step": 28868 }, { "epoch": 0.56, "learning_rate": 4.066404000802231e-05, "loss": 0.0288, "step": 28870 }, { "epoch": 0.56, "learning_rate": 4.066339304776508e-05, "loss": 0.0001, "step": 28872 }, { "epoch": 0.56, "learning_rate": 4.0662746087507847e-05, "loss": 0.0035, "step": 28874 }, { "epoch": 0.56, "learning_rate": 4.0662099127250616e-05, "loss": 0.0, "step": 28876 }, { "epoch": 0.56, "learning_rate": 4.0661452166993385e-05, "loss": 0.0002, "step": 28878 }, { "epoch": 0.56, "learning_rate": 4.0660805206736154e-05, "loss": 0.0001, "step": 28880 }, { "epoch": 0.56, "learning_rate": 4.0660158246478916e-05, "loss": 0.0001, "step": 28882 }, { "epoch": 0.56, "learning_rate": 4.065951128622169e-05, "loss": 0.0008, "step": 28884 }, { "epoch": 0.56, "learning_rate": 4.0658864325964454e-05, "loss": 0.0005, "step": 28886 }, { "epoch": 0.56, "learning_rate": 4.065821736570722e-05, "loss": 0.0, "step": 28888 }, { "epoch": 0.56, "learning_rate": 4.065757040545e-05, "loss": 0.0065, "step": 28890 }, { "epoch": 0.56, "learning_rate": 4.065692344519276e-05, "loss": 0.0132, "step": 28892 }, { "epoch": 0.56, "learning_rate": 4.065627648493554e-05, "loss": 0.003, "step": 28894 }, { "epoch": 0.56, "learning_rate": 4.06556295246783e-05, "loss": 0.0012, "step": 28896 }, { "epoch": 0.56, "learning_rate": 4.065498256442107e-05, "loss": 0.003, "step": 28898 }, { "epoch": 0.56, "learning_rate": 4.065433560416384e-05, "loss": 0.0005, "step": 28900 }, { "epoch": 0.56, "learning_rate": 4.065368864390661e-05, "loss": 0.0, "step": 28902 }, { "epoch": 0.56, "learning_rate": 4.0653041683649376e-05, "loss": 0.0109, "step": 28904 }, { "epoch": 0.56, "learning_rate": 4.0652394723392145e-05, "loss": 0.0003, "step": 28906 }, { "epoch": 0.56, "learning_rate": 4.0651747763134914e-05, "loss": 0.0027, "step": 28908 }, { "epoch": 0.56, "learning_rate": 4.0651100802877677e-05, "loss": 0.0, "step": 28910 }, { "epoch": 0.56, "learning_rate": 4.065045384262045e-05, "loss": 0.0, "step": 28912 }, { "epoch": 0.56, "learning_rate": 4.0649806882363215e-05, "loss": 0.0, "step": 28914 }, { "epoch": 0.56, "learning_rate": 4.064915992210599e-05, "loss": 0.0004, "step": 28916 }, { "epoch": 0.56, "learning_rate": 4.064851296184875e-05, "loss": 0.0075, "step": 28918 }, { "epoch": 0.56, "learning_rate": 4.064786600159152e-05, "loss": 0.0, "step": 28920 }, { "epoch": 0.56, "learning_rate": 4.06472190413343e-05, "loss": 0.0, "step": 28922 }, { "epoch": 0.56, "learning_rate": 4.064657208107706e-05, "loss": 0.0001, "step": 28924 }, { "epoch": 0.56, "learning_rate": 4.064592512081983e-05, "loss": 0.0, "step": 28926 }, { "epoch": 0.56, "learning_rate": 4.06452781605626e-05, "loss": 0.0143, "step": 28928 }, { "epoch": 0.56, "learning_rate": 4.064463120030537e-05, "loss": 0.0, "step": 28930 }, { "epoch": 0.56, "learning_rate": 4.0643984240048136e-05, "loss": 0.0, "step": 28932 }, { "epoch": 0.56, "learning_rate": 4.0643337279790906e-05, "loss": 0.0006, "step": 28934 }, { "epoch": 0.56, "learning_rate": 4.0642690319533675e-05, "loss": 0.0011, "step": 28936 }, { "epoch": 0.56, "learning_rate": 4.0642043359276444e-05, "loss": 0.0002, "step": 28938 }, { "epoch": 0.56, "learning_rate": 4.064139639901921e-05, "loss": 0.005, "step": 28940 }, { "epoch": 0.56, "learning_rate": 4.0640749438761975e-05, "loss": 0.004, "step": 28942 }, { "epoch": 0.56, "learning_rate": 4.064010247850475e-05, "loss": 0.0, "step": 28944 }, { "epoch": 0.56, "learning_rate": 4.063945551824751e-05, "loss": 0.0, "step": 28946 }, { "epoch": 0.56, "learning_rate": 4.063880855799028e-05, "loss": 0.0001, "step": 28948 }, { "epoch": 0.56, "learning_rate": 4.063816159773305e-05, "loss": 0.0, "step": 28950 }, { "epoch": 0.56, "learning_rate": 4.063751463747582e-05, "loss": 0.0001, "step": 28952 }, { "epoch": 0.56, "learning_rate": 4.0636867677218596e-05, "loss": 0.0, "step": 28954 }, { "epoch": 0.56, "learning_rate": 4.063622071696136e-05, "loss": 0.0, "step": 28956 }, { "epoch": 0.56, "learning_rate": 4.063557375670413e-05, "loss": 0.0001, "step": 28958 }, { "epoch": 0.56, "learning_rate": 4.06349267964469e-05, "loss": 0.0001, "step": 28960 }, { "epoch": 0.56, "learning_rate": 4.0634279836189666e-05, "loss": 0.0002, "step": 28962 }, { "epoch": 0.56, "learning_rate": 4.063363287593243e-05, "loss": 0.0001, "step": 28964 }, { "epoch": 0.56, "learning_rate": 4.0632985915675204e-05, "loss": 0.0, "step": 28966 }, { "epoch": 0.56, "learning_rate": 4.063233895541797e-05, "loss": 0.0, "step": 28968 }, { "epoch": 0.56, "learning_rate": 4.0631691995160736e-05, "loss": 0.0003, "step": 28970 }, { "epoch": 0.56, "learning_rate": 4.063104503490351e-05, "loss": 0.0, "step": 28972 }, { "epoch": 0.56, "learning_rate": 4.0630398074646274e-05, "loss": 0.0, "step": 28974 }, { "epoch": 0.56, "learning_rate": 4.062975111438905e-05, "loss": 0.0001, "step": 28976 }, { "epoch": 0.56, "learning_rate": 4.062910415413181e-05, "loss": 0.0001, "step": 28978 }, { "epoch": 0.56, "learning_rate": 4.062845719387458e-05, "loss": 0.0001, "step": 28980 }, { "epoch": 0.56, "learning_rate": 4.062781023361735e-05, "loss": 0.0001, "step": 28982 }, { "epoch": 0.56, "learning_rate": 4.062716327336012e-05, "loss": 0.0006, "step": 28984 }, { "epoch": 0.56, "learning_rate": 4.062651631310289e-05, "loss": 0.0003, "step": 28986 }, { "epoch": 0.56, "learning_rate": 4.062586935284566e-05, "loss": 0.0002, "step": 28988 }, { "epoch": 0.56, "learning_rate": 4.0625222392588426e-05, "loss": 0.0033, "step": 28990 }, { "epoch": 0.56, "learning_rate": 4.062457543233119e-05, "loss": 0.0, "step": 28992 }, { "epoch": 0.56, "learning_rate": 4.0623928472073965e-05, "loss": 0.0071, "step": 28994 }, { "epoch": 0.56, "learning_rate": 4.062328151181673e-05, "loss": 0.0008, "step": 28996 }, { "epoch": 0.56, "learning_rate": 4.06226345515595e-05, "loss": 0.0002, "step": 28998 }, { "epoch": 0.56, "learning_rate": 4.062198759130227e-05, "loss": 0.0, "step": 29000 }, { "epoch": 0.56, "learning_rate": 4.0621340631045034e-05, "loss": 0.0033, "step": 29002 }, { "epoch": 0.56, "learning_rate": 4.062069367078781e-05, "loss": 0.0, "step": 29004 }, { "epoch": 0.56, "learning_rate": 4.062004671053057e-05, "loss": 0.0001, "step": 29006 }, { "epoch": 0.56, "learning_rate": 4.061939975027334e-05, "loss": 0.0, "step": 29008 }, { "epoch": 0.56, "learning_rate": 4.061875279001611e-05, "loss": 0.0001, "step": 29010 }, { "epoch": 0.56, "learning_rate": 4.061810582975888e-05, "loss": 0.0004, "step": 29012 }, { "epoch": 0.56, "learning_rate": 4.061745886950165e-05, "loss": 0.0001, "step": 29014 }, { "epoch": 0.56, "learning_rate": 4.061681190924442e-05, "loss": 0.0023, "step": 29016 }, { "epoch": 0.56, "learning_rate": 4.061616494898719e-05, "loss": 0.0, "step": 29018 }, { "epoch": 0.56, "learning_rate": 4.0615517988729956e-05, "loss": 0.0002, "step": 29020 }, { "epoch": 0.56, "learning_rate": 4.0614871028472725e-05, "loss": 0.0004, "step": 29022 }, { "epoch": 0.56, "learning_rate": 4.061422406821549e-05, "loss": 0.0, "step": 29024 }, { "epoch": 0.56, "learning_rate": 4.061357710795826e-05, "loss": 0.0003, "step": 29026 }, { "epoch": 0.56, "learning_rate": 4.0612930147701025e-05, "loss": 0.0008, "step": 29028 }, { "epoch": 0.56, "learning_rate": 4.0612283187443794e-05, "loss": 0.0, "step": 29030 }, { "epoch": 0.56, "learning_rate": 4.0611636227186564e-05, "loss": 0.0001, "step": 29032 }, { "epoch": 0.56, "learning_rate": 4.061098926692933e-05, "loss": 0.0, "step": 29034 }, { "epoch": 0.56, "learning_rate": 4.061034230667211e-05, "loss": 0.0029, "step": 29036 }, { "epoch": 0.56, "learning_rate": 4.060969534641487e-05, "loss": 0.0002, "step": 29038 }, { "epoch": 0.56, "learning_rate": 4.060904838615764e-05, "loss": 0.0001, "step": 29040 }, { "epoch": 0.56, "learning_rate": 4.060840142590041e-05, "loss": 0.0082, "step": 29042 }, { "epoch": 0.56, "learning_rate": 4.060775446564318e-05, "loss": 0.0015, "step": 29044 }, { "epoch": 0.56, "learning_rate": 4.060710750538595e-05, "loss": 0.0001, "step": 29046 }, { "epoch": 0.56, "learning_rate": 4.0606460545128716e-05, "loss": 0.0007, "step": 29048 }, { "epoch": 0.56, "learning_rate": 4.0605813584871485e-05, "loss": 0.0002, "step": 29050 }, { "epoch": 0.56, "learning_rate": 4.060516662461425e-05, "loss": 0.0034, "step": 29052 }, { "epoch": 0.56, "learning_rate": 4.0604519664357023e-05, "loss": 0.0, "step": 29054 }, { "epoch": 0.56, "learning_rate": 4.0603872704099786e-05, "loss": 0.0048, "step": 29056 }, { "epoch": 0.56, "learning_rate": 4.060322574384256e-05, "loss": 0.0032, "step": 29058 }, { "epoch": 0.56, "learning_rate": 4.0602578783585324e-05, "loss": 0.0002, "step": 29060 }, { "epoch": 0.56, "learning_rate": 4.060193182332809e-05, "loss": 0.0008, "step": 29062 }, { "epoch": 0.56, "learning_rate": 4.060128486307086e-05, "loss": 0.0657, "step": 29064 }, { "epoch": 0.56, "learning_rate": 4.060063790281363e-05, "loss": 0.0001, "step": 29066 }, { "epoch": 0.56, "learning_rate": 4.05999909425564e-05, "loss": 0.0003, "step": 29068 }, { "epoch": 0.56, "learning_rate": 4.059934398229917e-05, "loss": 0.0001, "step": 29070 }, { "epoch": 0.56, "learning_rate": 4.059869702204194e-05, "loss": 0.0083, "step": 29072 }, { "epoch": 0.56, "learning_rate": 4.059805006178471e-05, "loss": 0.0001, "step": 29074 }, { "epoch": 0.56, "learning_rate": 4.059740310152748e-05, "loss": 0.0001, "step": 29076 }, { "epoch": 0.56, "learning_rate": 4.0596756141270246e-05, "loss": 0.0005, "step": 29078 }, { "epoch": 0.56, "learning_rate": 4.0596109181013015e-05, "loss": 0.0001, "step": 29080 }, { "epoch": 0.56, "learning_rate": 4.0595462220755784e-05, "loss": 0.001, "step": 29082 }, { "epoch": 0.56, "learning_rate": 4.0594815260498546e-05, "loss": 0.0009, "step": 29084 }, { "epoch": 0.56, "learning_rate": 4.059416830024132e-05, "loss": 0.0, "step": 29086 }, { "epoch": 0.56, "learning_rate": 4.0593521339984084e-05, "loss": 0.0025, "step": 29088 }, { "epoch": 0.56, "learning_rate": 4.0592874379726853e-05, "loss": 0.0001, "step": 29090 }, { "epoch": 0.56, "learning_rate": 4.059222741946962e-05, "loss": 0.0152, "step": 29092 }, { "epoch": 0.56, "learning_rate": 4.059158045921239e-05, "loss": 0.0043, "step": 29094 }, { "epoch": 0.56, "learning_rate": 4.059093349895516e-05, "loss": 0.0, "step": 29096 }, { "epoch": 0.56, "learning_rate": 4.059028653869793e-05, "loss": 0.0, "step": 29098 }, { "epoch": 0.56, "learning_rate": 4.05896395784407e-05, "loss": 0.0002, "step": 29100 }, { "epoch": 0.56, "learning_rate": 4.058899261818347e-05, "loss": 0.0008, "step": 29102 }, { "epoch": 0.56, "learning_rate": 4.058834565792624e-05, "loss": 0.0001, "step": 29104 }, { "epoch": 0.56, "learning_rate": 4.0587698697669e-05, "loss": 0.0002, "step": 29106 }, { "epoch": 0.56, "learning_rate": 4.0587051737411775e-05, "loss": 0.0, "step": 29108 }, { "epoch": 0.56, "learning_rate": 4.058640477715454e-05, "loss": 0.0, "step": 29110 }, { "epoch": 0.57, "learning_rate": 4.0585757816897307e-05, "loss": 0.0, "step": 29112 }, { "epoch": 0.57, "learning_rate": 4.058511085664008e-05, "loss": 0.0, "step": 29114 }, { "epoch": 0.57, "learning_rate": 4.0584463896382845e-05, "loss": 0.0039, "step": 29116 }, { "epoch": 0.57, "learning_rate": 4.058381693612562e-05, "loss": 0.0001, "step": 29118 }, { "epoch": 0.57, "learning_rate": 4.058316997586838e-05, "loss": 0.0, "step": 29120 }, { "epoch": 0.57, "learning_rate": 4.058252301561115e-05, "loss": 0.0, "step": 29122 }, { "epoch": 0.57, "learning_rate": 4.058187605535392e-05, "loss": 0.0102, "step": 29124 }, { "epoch": 0.57, "learning_rate": 4.058122909509669e-05, "loss": 0.0001, "step": 29126 }, { "epoch": 0.57, "learning_rate": 4.058058213483946e-05, "loss": 0.0002, "step": 29128 }, { "epoch": 0.57, "learning_rate": 4.057993517458223e-05, "loss": 0.0001, "step": 29130 }, { "epoch": 0.57, "learning_rate": 4.0579288214325e-05, "loss": 0.0152, "step": 29132 }, { "epoch": 0.57, "learning_rate": 4.057864125406776e-05, "loss": 0.0001, "step": 29134 }, { "epoch": 0.57, "learning_rate": 4.0577994293810536e-05, "loss": 0.0006, "step": 29136 }, { "epoch": 0.57, "learning_rate": 4.05773473335533e-05, "loss": 0.0001, "step": 29138 }, { "epoch": 0.57, "learning_rate": 4.0576700373296074e-05, "loss": 0.0047, "step": 29140 }, { "epoch": 0.57, "learning_rate": 4.0576053413038836e-05, "loss": 0.0086, "step": 29142 }, { "epoch": 0.57, "learning_rate": 4.0575406452781605e-05, "loss": 0.0, "step": 29144 }, { "epoch": 0.57, "learning_rate": 4.057475949252438e-05, "loss": 0.0144, "step": 29146 }, { "epoch": 0.57, "learning_rate": 4.057411253226714e-05, "loss": 0.0001, "step": 29148 }, { "epoch": 0.57, "learning_rate": 4.057346557200991e-05, "loss": 0.0, "step": 29150 }, { "epoch": 0.57, "learning_rate": 4.057281861175268e-05, "loss": 0.0034, "step": 29152 }, { "epoch": 0.57, "learning_rate": 4.057217165149545e-05, "loss": 0.0, "step": 29154 }, { "epoch": 0.57, "learning_rate": 4.057152469123822e-05, "loss": 0.0003, "step": 29156 }, { "epoch": 0.57, "learning_rate": 4.057087773098099e-05, "loss": 0.0163, "step": 29158 }, { "epoch": 0.57, "learning_rate": 4.057023077072376e-05, "loss": 0.0003, "step": 29160 }, { "epoch": 0.57, "learning_rate": 4.056958381046653e-05, "loss": 0.0001, "step": 29162 }, { "epoch": 0.57, "learning_rate": 4.0568936850209296e-05, "loss": 0.0002, "step": 29164 }, { "epoch": 0.57, "learning_rate": 4.056828988995206e-05, "loss": 0.0006, "step": 29166 }, { "epoch": 0.57, "learning_rate": 4.0567642929694834e-05, "loss": 0.0034, "step": 29168 }, { "epoch": 0.57, "learning_rate": 4.0566995969437597e-05, "loss": 0.0, "step": 29170 }, { "epoch": 0.57, "learning_rate": 4.0566349009180366e-05, "loss": 0.0091, "step": 29172 }, { "epoch": 0.57, "learning_rate": 4.0565702048923135e-05, "loss": 0.0, "step": 29174 }, { "epoch": 0.57, "learning_rate": 4.0565055088665904e-05, "loss": 0.0, "step": 29176 }, { "epoch": 0.57, "learning_rate": 4.056440812840868e-05, "loss": 0.0, "step": 29178 }, { "epoch": 0.57, "learning_rate": 4.056376116815144e-05, "loss": 0.0003, "step": 29180 }, { "epoch": 0.57, "learning_rate": 4.056311420789421e-05, "loss": 0.005, "step": 29182 }, { "epoch": 0.57, "learning_rate": 4.056246724763698e-05, "loss": 0.0001, "step": 29184 }, { "epoch": 0.57, "learning_rate": 4.056182028737975e-05, "loss": 0.0051, "step": 29186 }, { "epoch": 0.57, "learning_rate": 4.056117332712251e-05, "loss": 0.0003, "step": 29188 }, { "epoch": 0.57, "learning_rate": 4.056052636686529e-05, "loss": 0.0, "step": 29190 }, { "epoch": 0.57, "learning_rate": 4.0559879406608056e-05, "loss": 0.0001, "step": 29192 }, { "epoch": 0.57, "learning_rate": 4.055923244635082e-05, "loss": 0.0004, "step": 29194 }, { "epoch": 0.57, "learning_rate": 4.0558585486093595e-05, "loss": 0.0001, "step": 29196 }, { "epoch": 0.57, "learning_rate": 4.055793852583636e-05, "loss": 0.0107, "step": 29198 }, { "epoch": 0.57, "learning_rate": 4.055729156557913e-05, "loss": 0.0061, "step": 29200 }, { "epoch": 0.57, "learning_rate": 4.0556644605321895e-05, "loss": 0.0001, "step": 29202 }, { "epoch": 0.57, "learning_rate": 4.0555997645064664e-05, "loss": 0.0002, "step": 29204 }, { "epoch": 0.57, "learning_rate": 4.055535068480743e-05, "loss": 0.0006, "step": 29206 }, { "epoch": 0.57, "learning_rate": 4.05547037245502e-05, "loss": 0.0028, "step": 29208 }, { "epoch": 0.57, "learning_rate": 4.055405676429297e-05, "loss": 0.001, "step": 29210 }, { "epoch": 0.57, "learning_rate": 4.055340980403574e-05, "loss": 0.0011, "step": 29212 }, { "epoch": 0.57, "learning_rate": 4.055276284377851e-05, "loss": 0.0001, "step": 29214 }, { "epoch": 0.57, "learning_rate": 4.055211588352128e-05, "loss": 0.0012, "step": 29216 }, { "epoch": 0.57, "learning_rate": 4.055146892326405e-05, "loss": 0.005, "step": 29218 }, { "epoch": 0.57, "learning_rate": 4.055082196300681e-05, "loss": 0.0, "step": 29220 }, { "epoch": 0.57, "learning_rate": 4.0550175002749586e-05, "loss": 0.0004, "step": 29222 }, { "epoch": 0.57, "learning_rate": 4.0549528042492355e-05, "loss": 0.0, "step": 29224 }, { "epoch": 0.57, "learning_rate": 4.054888108223512e-05, "loss": 0.0, "step": 29226 }, { "epoch": 0.57, "learning_rate": 4.054823412197789e-05, "loss": 0.0001, "step": 29228 }, { "epoch": 0.57, "learning_rate": 4.0547587161720655e-05, "loss": 0.0006, "step": 29230 }, { "epoch": 0.57, "learning_rate": 4.0546940201463425e-05, "loss": 0.0001, "step": 29232 }, { "epoch": 0.57, "learning_rate": 4.0546293241206194e-05, "loss": 0.0004, "step": 29234 }, { "epoch": 0.57, "learning_rate": 4.054564628094896e-05, "loss": 0.0037, "step": 29236 }, { "epoch": 0.57, "learning_rate": 4.054499932069173e-05, "loss": 0.008, "step": 29238 }, { "epoch": 0.57, "learning_rate": 4.05443523604345e-05, "loss": 0.0001, "step": 29240 }, { "epoch": 0.57, "learning_rate": 4.054370540017727e-05, "loss": 0.0005, "step": 29242 }, { "epoch": 0.57, "learning_rate": 4.054305843992004e-05, "loss": 0.0, "step": 29244 }, { "epoch": 0.57, "learning_rate": 4.054241147966281e-05, "loss": 0.0092, "step": 29246 }, { "epoch": 0.57, "learning_rate": 4.054176451940557e-05, "loss": 0.0003, "step": 29248 }, { "epoch": 0.57, "learning_rate": 4.0541117559148346e-05, "loss": 0.0017, "step": 29250 }, { "epoch": 0.57, "learning_rate": 4.054079407901973e-05, "loss": 0.3033, "step": 29252 }, { "epoch": 0.57, "learning_rate": 4.054014711876249e-05, "loss": 0.005, "step": 29254 }, { "epoch": 0.57, "learning_rate": 4.053950015850527e-05, "loss": 0.0211, "step": 29256 }, { "epoch": 0.57, "learning_rate": 4.053885319824803e-05, "loss": 0.0012, "step": 29258 }, { "epoch": 0.57, "learning_rate": 4.05382062379908e-05, "loss": 0.0018, "step": 29260 }, { "epoch": 0.57, "learning_rate": 4.053755927773357e-05, "loss": 0.0072, "step": 29262 }, { "epoch": 0.57, "learning_rate": 4.053691231747634e-05, "loss": 0.0, "step": 29264 }, { "epoch": 0.57, "learning_rate": 4.053626535721911e-05, "loss": 0.0003, "step": 29266 }, { "epoch": 0.57, "learning_rate": 4.053561839696188e-05, "loss": 0.0001, "step": 29268 }, { "epoch": 0.57, "learning_rate": 4.0534971436704646e-05, "loss": 0.0008, "step": 29270 }, { "epoch": 0.57, "learning_rate": 4.0534324476447415e-05, "loss": 0.0026, "step": 29272 }, { "epoch": 0.57, "learning_rate": 4.0533677516190184e-05, "loss": 0.0008, "step": 29274 }, { "epoch": 0.57, "learning_rate": 4.0533030555932946e-05, "loss": 0.0002, "step": 29276 }, { "epoch": 0.57, "learning_rate": 4.053238359567572e-05, "loss": 0.0009, "step": 29278 }, { "epoch": 0.57, "learning_rate": 4.053173663541849e-05, "loss": 0.003, "step": 29280 }, { "epoch": 0.57, "learning_rate": 4.0531089675161254e-05, "loss": 0.0001, "step": 29282 }, { "epoch": 0.57, "learning_rate": 4.053044271490403e-05, "loss": 0.0001, "step": 29284 }, { "epoch": 0.57, "learning_rate": 4.052979575464679e-05, "loss": 0.0001, "step": 29286 }, { "epoch": 0.57, "learning_rate": 4.052914879438957e-05, "loss": 0.0003, "step": 29288 }, { "epoch": 0.57, "learning_rate": 4.052850183413233e-05, "loss": 0.0001, "step": 29290 }, { "epoch": 0.57, "learning_rate": 4.05278548738751e-05, "loss": 0.0002, "step": 29292 }, { "epoch": 0.57, "learning_rate": 4.052720791361787e-05, "loss": 0.0002, "step": 29294 }, { "epoch": 0.57, "learning_rate": 4.052656095336064e-05, "loss": 0.0055, "step": 29296 }, { "epoch": 0.57, "learning_rate": 4.0525913993103406e-05, "loss": 0.0001, "step": 29298 }, { "epoch": 0.57, "learning_rate": 4.0525267032846175e-05, "loss": 0.0288, "step": 29300 }, { "epoch": 0.57, "learning_rate": 4.0524620072588944e-05, "loss": 0.0, "step": 29302 }, { "epoch": 0.57, "learning_rate": 4.052397311233171e-05, "loss": 0.0027, "step": 29304 }, { "epoch": 0.57, "learning_rate": 4.052332615207448e-05, "loss": 0.0, "step": 29306 }, { "epoch": 0.57, "learning_rate": 4.0522679191817245e-05, "loss": 0.0004, "step": 29308 }, { "epoch": 0.57, "learning_rate": 4.052203223156002e-05, "loss": 0.0001, "step": 29310 }, { "epoch": 0.57, "learning_rate": 4.052138527130278e-05, "loss": 0.0, "step": 29312 }, { "epoch": 0.57, "learning_rate": 4.052073831104555e-05, "loss": 0.0001, "step": 29314 }, { "epoch": 0.57, "learning_rate": 4.052009135078833e-05, "loss": 0.0001, "step": 29316 }, { "epoch": 0.57, "learning_rate": 4.051944439053109e-05, "loss": 0.0008, "step": 29318 }, { "epoch": 0.57, "learning_rate": 4.051879743027386e-05, "loss": 0.0, "step": 29320 }, { "epoch": 0.57, "learning_rate": 4.051815047001663e-05, "loss": 0.0, "step": 29322 }, { "epoch": 0.57, "learning_rate": 4.05175035097594e-05, "loss": 0.001, "step": 29324 }, { "epoch": 0.57, "learning_rate": 4.051685654950217e-05, "loss": 0.0001, "step": 29326 }, { "epoch": 0.57, "learning_rate": 4.0516209589244936e-05, "loss": 0.0016, "step": 29328 }, { "epoch": 0.57, "learning_rate": 4.0515562628987705e-05, "loss": 0.0001, "step": 29330 }, { "epoch": 0.57, "learning_rate": 4.0514915668730474e-05, "loss": 0.0012, "step": 29332 }, { "epoch": 0.57, "learning_rate": 4.051426870847324e-05, "loss": 0.0005, "step": 29334 }, { "epoch": 0.57, "learning_rate": 4.0513621748216005e-05, "loss": 0.0004, "step": 29336 }, { "epoch": 0.57, "learning_rate": 4.051297478795878e-05, "loss": 0.0003, "step": 29338 }, { "epoch": 0.57, "learning_rate": 4.0512327827701543e-05, "loss": 0.0002, "step": 29340 }, { "epoch": 0.57, "learning_rate": 4.051168086744431e-05, "loss": 0.0001, "step": 29342 }, { "epoch": 0.57, "learning_rate": 4.051103390718708e-05, "loss": 0.0061, "step": 29344 }, { "epoch": 0.57, "learning_rate": 4.051038694692985e-05, "loss": 0.0017, "step": 29346 }, { "epoch": 0.57, "learning_rate": 4.050973998667262e-05, "loss": 0.0001, "step": 29348 }, { "epoch": 0.57, "learning_rate": 4.050909302641539e-05, "loss": 0.0, "step": 29350 }, { "epoch": 0.57, "learning_rate": 4.050844606615816e-05, "loss": 0.0002, "step": 29352 }, { "epoch": 0.57, "learning_rate": 4.050779910590093e-05, "loss": 0.0, "step": 29354 }, { "epoch": 0.57, "learning_rate": 4.0507152145643696e-05, "loss": 0.0033, "step": 29356 }, { "epoch": 0.57, "learning_rate": 4.0506505185386465e-05, "loss": 0.0073, "step": 29358 }, { "epoch": 0.57, "learning_rate": 4.0505858225129234e-05, "loss": 0.0001, "step": 29360 }, { "epoch": 0.57, "learning_rate": 4.0505211264872003e-05, "loss": 0.0087, "step": 29362 }, { "epoch": 0.57, "learning_rate": 4.0504564304614766e-05, "loss": 0.0025, "step": 29364 }, { "epoch": 0.57, "learning_rate": 4.050391734435754e-05, "loss": 0.0001, "step": 29366 }, { "epoch": 0.57, "learning_rate": 4.0503270384100304e-05, "loss": 0.0005, "step": 29368 }, { "epoch": 0.57, "learning_rate": 4.050262342384308e-05, "loss": 0.0, "step": 29370 }, { "epoch": 0.57, "learning_rate": 4.050197646358584e-05, "loss": 0.0, "step": 29372 }, { "epoch": 0.57, "learning_rate": 4.050132950332861e-05, "loss": 0.0, "step": 29374 }, { "epoch": 0.57, "learning_rate": 4.050068254307138e-05, "loss": 0.0037, "step": 29376 }, { "epoch": 0.57, "learning_rate": 4.050003558281415e-05, "loss": 0.0001, "step": 29378 }, { "epoch": 0.57, "learning_rate": 4.049938862255692e-05, "loss": 0.003, "step": 29380 }, { "epoch": 0.57, "learning_rate": 4.049874166229969e-05, "loss": 0.0024, "step": 29382 }, { "epoch": 0.57, "learning_rate": 4.0498094702042457e-05, "loss": 0.0003, "step": 29384 }, { "epoch": 0.57, "learning_rate": 4.049744774178522e-05, "loss": 0.0007, "step": 29386 }, { "epoch": 0.57, "learning_rate": 4.0496800781527995e-05, "loss": 0.0013, "step": 29388 }, { "epoch": 0.57, "learning_rate": 4.049615382127076e-05, "loss": 0.0, "step": 29390 }, { "epoch": 0.57, "learning_rate": 4.049550686101353e-05, "loss": 0.0001, "step": 29392 }, { "epoch": 0.57, "learning_rate": 4.04948599007563e-05, "loss": 0.0019, "step": 29394 }, { "epoch": 0.57, "learning_rate": 4.0494212940499064e-05, "loss": 0.0, "step": 29396 }, { "epoch": 0.57, "learning_rate": 4.049356598024184e-05, "loss": 0.0, "step": 29398 }, { "epoch": 0.57, "learning_rate": 4.04929190199846e-05, "loss": 0.0, "step": 29400 }, { "epoch": 0.57, "learning_rate": 4.049227205972737e-05, "loss": 0.011, "step": 29402 }, { "epoch": 0.57, "learning_rate": 4.049162509947014e-05, "loss": 0.0005, "step": 29404 }, { "epoch": 0.57, "learning_rate": 4.049097813921291e-05, "loss": 0.0, "step": 29406 }, { "epoch": 0.57, "learning_rate": 4.049033117895568e-05, "loss": 0.0, "step": 29408 }, { "epoch": 0.57, "learning_rate": 4.048968421869845e-05, "loss": 0.0001, "step": 29410 }, { "epoch": 0.57, "learning_rate": 4.048903725844122e-05, "loss": 0.0036, "step": 29412 }, { "epoch": 0.57, "learning_rate": 4.0488390298183986e-05, "loss": 0.0, "step": 29414 }, { "epoch": 0.57, "learning_rate": 4.0487743337926755e-05, "loss": 0.013, "step": 29416 }, { "epoch": 0.57, "learning_rate": 4.048709637766952e-05, "loss": 0.0, "step": 29418 }, { "epoch": 0.57, "learning_rate": 4.048644941741229e-05, "loss": 0.0025, "step": 29420 }, { "epoch": 0.57, "learning_rate": 4.0485802457155056e-05, "loss": 0.002, "step": 29422 }, { "epoch": 0.57, "learning_rate": 4.0485155496897825e-05, "loss": 0.0, "step": 29424 }, { "epoch": 0.57, "learning_rate": 4.04845085366406e-05, "loss": 0.0029, "step": 29426 }, { "epoch": 0.57, "learning_rate": 4.048386157638336e-05, "loss": 0.0191, "step": 29428 }, { "epoch": 0.57, "learning_rate": 4.048321461612614e-05, "loss": 0.0, "step": 29430 }, { "epoch": 0.57, "learning_rate": 4.04825676558689e-05, "loss": 0.0001, "step": 29432 }, { "epoch": 0.57, "learning_rate": 4.048192069561167e-05, "loss": 0.0, "step": 29434 }, { "epoch": 0.57, "learning_rate": 4.048127373535444e-05, "loss": 0.0013, "step": 29436 }, { "epoch": 0.57, "learning_rate": 4.048062677509721e-05, "loss": 0.0009, "step": 29438 }, { "epoch": 0.57, "learning_rate": 4.047997981483998e-05, "loss": 0.0039, "step": 29440 }, { "epoch": 0.57, "learning_rate": 4.0479332854582746e-05, "loss": 0.0097, "step": 29442 }, { "epoch": 0.57, "learning_rate": 4.0478685894325516e-05, "loss": 0.0001, "step": 29444 }, { "epoch": 0.57, "learning_rate": 4.047803893406828e-05, "loss": 0.0006, "step": 29446 }, { "epoch": 0.57, "learning_rate": 4.0477391973811054e-05, "loss": 0.0189, "step": 29448 }, { "epoch": 0.57, "learning_rate": 4.0476745013553816e-05, "loss": 0.0286, "step": 29450 }, { "epoch": 0.57, "learning_rate": 4.047609805329659e-05, "loss": 0.0, "step": 29452 }, { "epoch": 0.57, "learning_rate": 4.0475451093039354e-05, "loss": 0.0025, "step": 29454 }, { "epoch": 0.57, "learning_rate": 4.047480413278212e-05, "loss": 0.0007, "step": 29456 }, { "epoch": 0.57, "learning_rate": 4.04741571725249e-05, "loss": 0.0602, "step": 29458 }, { "epoch": 0.57, "learning_rate": 4.047351021226766e-05, "loss": 0.0004, "step": 29460 }, { "epoch": 0.57, "learning_rate": 4.047286325201043e-05, "loss": 0.0, "step": 29462 }, { "epoch": 0.57, "learning_rate": 4.04722162917532e-05, "loss": 0.0, "step": 29464 }, { "epoch": 0.57, "learning_rate": 4.047156933149597e-05, "loss": 0.0001, "step": 29466 }, { "epoch": 0.57, "learning_rate": 4.047092237123873e-05, "loss": 0.0001, "step": 29468 }, { "epoch": 0.57, "learning_rate": 4.047027541098151e-05, "loss": 0.0, "step": 29470 }, { "epoch": 0.57, "learning_rate": 4.0469628450724276e-05, "loss": 0.0001, "step": 29472 }, { "epoch": 0.57, "learning_rate": 4.0468981490467045e-05, "loss": 0.0001, "step": 29474 }, { "epoch": 0.57, "learning_rate": 4.0468334530209814e-05, "loss": 0.0, "step": 29476 }, { "epoch": 0.57, "learning_rate": 4.0467687569952576e-05, "loss": 0.0002, "step": 29478 }, { "epoch": 0.57, "learning_rate": 4.046704060969535e-05, "loss": 0.0001, "step": 29480 }, { "epoch": 0.57, "learning_rate": 4.0466393649438115e-05, "loss": 0.0008, "step": 29482 }, { "epoch": 0.57, "learning_rate": 4.0465746689180884e-05, "loss": 0.0001, "step": 29484 }, { "epoch": 0.57, "learning_rate": 4.046509972892365e-05, "loss": 0.0003, "step": 29486 }, { "epoch": 0.57, "learning_rate": 4.046445276866642e-05, "loss": 0.0035, "step": 29488 }, { "epoch": 0.57, "learning_rate": 4.046380580840919e-05, "loss": 0.0001, "step": 29490 }, { "epoch": 0.57, "learning_rate": 4.046315884815196e-05, "loss": 0.0, "step": 29492 }, { "epoch": 0.57, "learning_rate": 4.046251188789473e-05, "loss": 0.0017, "step": 29494 }, { "epoch": 0.57, "learning_rate": 4.04618649276375e-05, "loss": 0.0003, "step": 29496 }, { "epoch": 0.57, "learning_rate": 4.046121796738027e-05, "loss": 0.0002, "step": 29498 }, { "epoch": 0.57, "learning_rate": 4.046057100712303e-05, "loss": 0.0354, "step": 29500 }, { "epoch": 0.57, "learning_rate": 4.0459924046865805e-05, "loss": 0.0001, "step": 29502 }, { "epoch": 0.57, "learning_rate": 4.0459277086608575e-05, "loss": 0.0, "step": 29504 }, { "epoch": 0.57, "learning_rate": 4.045863012635134e-05, "loss": 0.0, "step": 29506 }, { "epoch": 0.57, "learning_rate": 4.045798316609411e-05, "loss": 0.0079, "step": 29508 }, { "epoch": 0.57, "learning_rate": 4.0457336205836875e-05, "loss": 0.0023, "step": 29510 }, { "epoch": 0.57, "learning_rate": 4.045668924557965e-05, "loss": 0.0001, "step": 29512 }, { "epoch": 0.57, "learning_rate": 4.045604228532241e-05, "loss": 0.0, "step": 29514 }, { "epoch": 0.57, "learning_rate": 4.045539532506518e-05, "loss": 0.0, "step": 29516 }, { "epoch": 0.57, "learning_rate": 4.045474836480795e-05, "loss": 0.0015, "step": 29518 }, { "epoch": 0.57, "learning_rate": 4.045410140455072e-05, "loss": 0.0, "step": 29520 }, { "epoch": 0.57, "learning_rate": 4.045345444429349e-05, "loss": 0.0014, "step": 29522 }, { "epoch": 0.57, "learning_rate": 4.045280748403626e-05, "loss": 0.0, "step": 29524 }, { "epoch": 0.57, "learning_rate": 4.045216052377903e-05, "loss": 0.0008, "step": 29526 }, { "epoch": 0.57, "learning_rate": 4.045151356352179e-05, "loss": 0.0001, "step": 29528 }, { "epoch": 0.57, "learning_rate": 4.0450866603264566e-05, "loss": 0.0, "step": 29530 }, { "epoch": 0.57, "learning_rate": 4.045021964300733e-05, "loss": 0.0, "step": 29532 }, { "epoch": 0.57, "learning_rate": 4.0449572682750104e-05, "loss": 0.0002, "step": 29534 }, { "epoch": 0.57, "learning_rate": 4.044892572249287e-05, "loss": 0.0, "step": 29536 }, { "epoch": 0.57, "learning_rate": 4.0448278762235635e-05, "loss": 0.0094, "step": 29538 }, { "epoch": 0.57, "learning_rate": 4.044763180197841e-05, "loss": 0.0065, "step": 29540 }, { "epoch": 0.57, "learning_rate": 4.0446984841721174e-05, "loss": 0.0002, "step": 29542 }, { "epoch": 0.57, "learning_rate": 4.044633788146394e-05, "loss": 0.0, "step": 29544 }, { "epoch": 0.57, "learning_rate": 4.044569092120671e-05, "loss": 0.0001, "step": 29546 }, { "epoch": 0.57, "learning_rate": 4.044504396094948e-05, "loss": 0.0, "step": 29548 }, { "epoch": 0.57, "learning_rate": 4.044439700069225e-05, "loss": 0.0016, "step": 29550 }, { "epoch": 0.57, "learning_rate": 4.044375004043502e-05, "loss": 0.0006, "step": 29552 }, { "epoch": 0.57, "learning_rate": 4.044310308017779e-05, "loss": 0.0, "step": 29554 }, { "epoch": 0.57, "learning_rate": 4.044245611992056e-05, "loss": 0.0003, "step": 29556 }, { "epoch": 0.57, "learning_rate": 4.0441809159663326e-05, "loss": 0.0039, "step": 29558 }, { "epoch": 0.57, "learning_rate": 4.044116219940609e-05, "loss": 0.0008, "step": 29560 }, { "epoch": 0.57, "learning_rate": 4.0440515239148864e-05, "loss": 0.0001, "step": 29562 }, { "epoch": 0.57, "learning_rate": 4.043986827889163e-05, "loss": 0.0, "step": 29564 }, { "epoch": 0.57, "learning_rate": 4.0439221318634396e-05, "loss": 0.0, "step": 29566 }, { "epoch": 0.57, "learning_rate": 4.0438574358377165e-05, "loss": 0.0, "step": 29568 }, { "epoch": 0.57, "learning_rate": 4.0437927398119934e-05, "loss": 0.0025, "step": 29570 }, { "epoch": 0.57, "learning_rate": 4.04372804378627e-05, "loss": 0.0, "step": 29572 }, { "epoch": 0.57, "learning_rate": 4.043663347760547e-05, "loss": 0.0, "step": 29574 }, { "epoch": 0.57, "learning_rate": 4.043598651734824e-05, "loss": 0.001, "step": 29576 }, { "epoch": 0.57, "learning_rate": 4.043533955709101e-05, "loss": 0.0002, "step": 29578 }, { "epoch": 0.57, "learning_rate": 4.043469259683378e-05, "loss": 0.0, "step": 29580 }, { "epoch": 0.57, "learning_rate": 4.043404563657655e-05, "loss": 0.0001, "step": 29582 }, { "epoch": 0.57, "learning_rate": 4.043339867631932e-05, "loss": 0.0001, "step": 29584 }, { "epoch": 0.57, "learning_rate": 4.043275171606209e-05, "loss": 0.0005, "step": 29586 }, { "epoch": 0.57, "learning_rate": 4.043210475580485e-05, "loss": 0.0, "step": 29588 }, { "epoch": 0.57, "learning_rate": 4.0431457795547625e-05, "loss": 0.0001, "step": 29590 }, { "epoch": 0.57, "learning_rate": 4.043081083529039e-05, "loss": 0.0, "step": 29592 }, { "epoch": 0.57, "learning_rate": 4.043016387503316e-05, "loss": 0.0015, "step": 29594 }, { "epoch": 0.57, "learning_rate": 4.0429516914775925e-05, "loss": 0.0001, "step": 29596 }, { "epoch": 0.57, "learning_rate": 4.0428869954518694e-05, "loss": 0.0003, "step": 29598 }, { "epoch": 0.57, "learning_rate": 4.0428222994261463e-05, "loss": 0.0009, "step": 29600 }, { "epoch": 0.57, "learning_rate": 4.042757603400423e-05, "loss": 0.0033, "step": 29602 }, { "epoch": 0.57, "learning_rate": 4.0426929073747e-05, "loss": 0.0073, "step": 29604 }, { "epoch": 0.57, "learning_rate": 4.042628211348977e-05, "loss": 0.0001, "step": 29606 }, { "epoch": 0.57, "learning_rate": 4.042563515323254e-05, "loss": 0.0001, "step": 29608 }, { "epoch": 0.57, "learning_rate": 4.04249881929753e-05, "loss": 0.0024, "step": 29610 }, { "epoch": 0.57, "learning_rate": 4.042434123271808e-05, "loss": 0.0067, "step": 29612 }, { "epoch": 0.57, "learning_rate": 4.042369427246084e-05, "loss": 0.005, "step": 29614 }, { "epoch": 0.57, "learning_rate": 4.0423047312203616e-05, "loss": 0.0001, "step": 29616 }, { "epoch": 0.57, "learning_rate": 4.0422400351946385e-05, "loss": 0.0003, "step": 29618 }, { "epoch": 0.57, "learning_rate": 4.042175339168915e-05, "loss": 0.0, "step": 29620 }, { "epoch": 0.57, "learning_rate": 4.042110643143192e-05, "loss": 0.001, "step": 29622 }, { "epoch": 0.57, "learning_rate": 4.0420459471174686e-05, "loss": 0.0, "step": 29624 }, { "epoch": 0.58, "learning_rate": 4.0419812510917455e-05, "loss": 0.0, "step": 29626 }, { "epoch": 0.58, "learning_rate": 4.0419165550660224e-05, "loss": 0.004, "step": 29628 }, { "epoch": 0.58, "learning_rate": 4.041851859040299e-05, "loss": 0.0006, "step": 29630 }, { "epoch": 0.58, "learning_rate": 4.041787163014576e-05, "loss": 0.0093, "step": 29632 }, { "epoch": 0.58, "learning_rate": 4.041722466988853e-05, "loss": 0.0108, "step": 29634 }, { "epoch": 0.58, "learning_rate": 4.04165777096313e-05, "loss": 0.0001, "step": 29636 }, { "epoch": 0.58, "learning_rate": 4.041593074937407e-05, "loss": 0.0001, "step": 29638 }, { "epoch": 0.58, "learning_rate": 4.041528378911684e-05, "loss": 0.0001, "step": 29640 }, { "epoch": 0.58, "learning_rate": 4.04146368288596e-05, "loss": 0.0011, "step": 29642 }, { "epoch": 0.58, "learning_rate": 4.0413989868602377e-05, "loss": 0.0001, "step": 29644 }, { "epoch": 0.58, "learning_rate": 4.041334290834514e-05, "loss": 0.0111, "step": 29646 }, { "epoch": 0.58, "learning_rate": 4.041269594808791e-05, "loss": 0.0007, "step": 29648 }, { "epoch": 0.58, "learning_rate": 4.0412048987830684e-05, "loss": 0.0007, "step": 29650 }, { "epoch": 0.58, "learning_rate": 4.0411402027573446e-05, "loss": 0.0003, "step": 29652 }, { "epoch": 0.58, "learning_rate": 4.041075506731622e-05, "loss": 0.0016, "step": 29654 }, { "epoch": 0.58, "learning_rate": 4.0410108107058984e-05, "loss": 0.0, "step": 29656 }, { "epoch": 0.58, "learning_rate": 4.040946114680175e-05, "loss": 0.0001, "step": 29658 }, { "epoch": 0.58, "learning_rate": 4.040881418654452e-05, "loss": 0.0001, "step": 29660 }, { "epoch": 0.58, "learning_rate": 4.040816722628729e-05, "loss": 0.0001, "step": 29662 }, { "epoch": 0.58, "learning_rate": 4.040752026603006e-05, "loss": 0.0, "step": 29664 }, { "epoch": 0.58, "learning_rate": 4.040687330577283e-05, "loss": 0.0001, "step": 29666 }, { "epoch": 0.58, "learning_rate": 4.04062263455156e-05, "loss": 0.0, "step": 29668 }, { "epoch": 0.58, "learning_rate": 4.040557938525836e-05, "loss": 0.0, "step": 29670 }, { "epoch": 0.58, "learning_rate": 4.040493242500114e-05, "loss": 0.0001, "step": 29672 }, { "epoch": 0.58, "learning_rate": 4.04042854647439e-05, "loss": 0.0007, "step": 29674 }, { "epoch": 0.58, "learning_rate": 4.0403638504486675e-05, "loss": 0.001, "step": 29676 }, { "epoch": 0.58, "learning_rate": 4.040299154422944e-05, "loss": 0.0, "step": 29678 }, { "epoch": 0.58, "learning_rate": 4.0402344583972207e-05, "loss": 0.0, "step": 29680 }, { "epoch": 0.58, "learning_rate": 4.040169762371498e-05, "loss": 0.0, "step": 29682 }, { "epoch": 0.58, "learning_rate": 4.0401050663457745e-05, "loss": 0.0, "step": 29684 }, { "epoch": 0.58, "learning_rate": 4.0400403703200514e-05, "loss": 0.0, "step": 29686 }, { "epoch": 0.58, "learning_rate": 4.039975674294328e-05, "loss": 0.0001, "step": 29688 }, { "epoch": 0.58, "learning_rate": 4.039910978268605e-05, "loss": 0.0, "step": 29690 }, { "epoch": 0.58, "learning_rate": 4.0398462822428814e-05, "loss": 0.0002, "step": 29692 }, { "epoch": 0.58, "learning_rate": 4.039781586217159e-05, "loss": 0.0002, "step": 29694 }, { "epoch": 0.58, "learning_rate": 4.039716890191436e-05, "loss": 0.0, "step": 29696 }, { "epoch": 0.58, "learning_rate": 4.039652194165713e-05, "loss": 0.0065, "step": 29698 }, { "epoch": 0.58, "learning_rate": 4.03958749813999e-05, "loss": 0.0001, "step": 29700 }, { "epoch": 0.58, "learning_rate": 4.039522802114266e-05, "loss": 0.0, "step": 29702 }, { "epoch": 0.58, "learning_rate": 4.0394581060885436e-05, "loss": 0.0001, "step": 29704 }, { "epoch": 0.58, "learning_rate": 4.03939341006282e-05, "loss": 0.0099, "step": 29706 }, { "epoch": 0.58, "learning_rate": 4.039328714037097e-05, "loss": 0.0011, "step": 29708 }, { "epoch": 0.58, "learning_rate": 4.0392640180113736e-05, "loss": 0.0039, "step": 29710 }, { "epoch": 0.58, "learning_rate": 4.0391993219856505e-05, "loss": 0.0, "step": 29712 }, { "epoch": 0.58, "learning_rate": 4.0391346259599274e-05, "loss": 0.0004, "step": 29714 }, { "epoch": 0.58, "learning_rate": 4.039069929934204e-05, "loss": 0.0, "step": 29716 }, { "epoch": 0.58, "learning_rate": 4.039005233908481e-05, "loss": 0.0014, "step": 29718 }, { "epoch": 0.58, "learning_rate": 4.038940537882758e-05, "loss": 0.0, "step": 29720 }, { "epoch": 0.58, "learning_rate": 4.038875841857035e-05, "loss": 0.0043, "step": 29722 }, { "epoch": 0.58, "learning_rate": 4.038811145831311e-05, "loss": 0.0, "step": 29724 }, { "epoch": 0.58, "learning_rate": 4.038746449805589e-05, "loss": 0.0024, "step": 29726 }, { "epoch": 0.58, "learning_rate": 4.038681753779866e-05, "loss": 0.0, "step": 29728 }, { "epoch": 0.58, "learning_rate": 4.038617057754142e-05, "loss": 0.0036, "step": 29730 }, { "epoch": 0.58, "learning_rate": 4.0385523617284196e-05, "loss": 0.0019, "step": 29732 }, { "epoch": 0.58, "learning_rate": 4.038487665702696e-05, "loss": 0.0012, "step": 29734 }, { "epoch": 0.58, "learning_rate": 4.0384229696769734e-05, "loss": 0.0037, "step": 29736 }, { "epoch": 0.58, "learning_rate": 4.0383582736512496e-05, "loss": 0.0019, "step": 29738 }, { "epoch": 0.58, "learning_rate": 4.0382935776255265e-05, "loss": 0.0049, "step": 29740 }, { "epoch": 0.58, "learning_rate": 4.0382288815998035e-05, "loss": 0.0019, "step": 29742 }, { "epoch": 0.58, "learning_rate": 4.0381641855740804e-05, "loss": 0.001, "step": 29744 }, { "epoch": 0.58, "learning_rate": 4.038099489548357e-05, "loss": 0.0001, "step": 29746 }, { "epoch": 0.58, "learning_rate": 4.038034793522634e-05, "loss": 0.0, "step": 29748 }, { "epoch": 0.58, "learning_rate": 4.037970097496911e-05, "loss": 0.0001, "step": 29750 }, { "epoch": 0.58, "learning_rate": 4.037905401471187e-05, "loss": 0.0002, "step": 29752 }, { "epoch": 0.58, "learning_rate": 4.037840705445465e-05, "loss": 0.0044, "step": 29754 }, { "epoch": 0.58, "learning_rate": 4.037776009419741e-05, "loss": 0.0054, "step": 29756 }, { "epoch": 0.58, "learning_rate": 4.037711313394019e-05, "loss": 0.0001, "step": 29758 }, { "epoch": 0.58, "learning_rate": 4.0376466173682956e-05, "loss": 0.0007, "step": 29760 }, { "epoch": 0.58, "learning_rate": 4.037581921342572e-05, "loss": 0.0002, "step": 29762 }, { "epoch": 0.58, "learning_rate": 4.0375172253168494e-05, "loss": 0.0, "step": 29764 }, { "epoch": 0.58, "learning_rate": 4.037452529291126e-05, "loss": 0.0001, "step": 29766 }, { "epoch": 0.58, "learning_rate": 4.0373878332654026e-05, "loss": 0.0027, "step": 29768 }, { "epoch": 0.58, "learning_rate": 4.0373231372396795e-05, "loss": 0.0002, "step": 29770 }, { "epoch": 0.58, "learning_rate": 4.0372584412139564e-05, "loss": 0.0, "step": 29772 }, { "epoch": 0.58, "learning_rate": 4.037193745188233e-05, "loss": 0.0001, "step": 29774 }, { "epoch": 0.58, "learning_rate": 4.03712904916251e-05, "loss": 0.0001, "step": 29776 }, { "epoch": 0.58, "learning_rate": 4.037064353136787e-05, "loss": 0.0, "step": 29778 }, { "epoch": 0.58, "learning_rate": 4.036999657111064e-05, "loss": 0.0001, "step": 29780 }, { "epoch": 0.58, "learning_rate": 4.036934961085341e-05, "loss": 0.0185, "step": 29782 }, { "epoch": 0.58, "learning_rate": 4.036870265059617e-05, "loss": 0.0049, "step": 29784 }, { "epoch": 0.58, "learning_rate": 4.036805569033895e-05, "loss": 0.0059, "step": 29786 }, { "epoch": 0.58, "learning_rate": 4.036740873008171e-05, "loss": 0.0049, "step": 29788 }, { "epoch": 0.58, "learning_rate": 4.036676176982448e-05, "loss": 0.0, "step": 29790 }, { "epoch": 0.58, "learning_rate": 4.036611480956725e-05, "loss": 0.0, "step": 29792 }, { "epoch": 0.58, "learning_rate": 4.036546784931002e-05, "loss": 0.0002, "step": 29794 }, { "epoch": 0.58, "learning_rate": 4.036482088905279e-05, "loss": 0.0, "step": 29796 }, { "epoch": 0.58, "learning_rate": 4.0364173928795555e-05, "loss": 0.0092, "step": 29798 }, { "epoch": 0.58, "learning_rate": 4.0363526968538324e-05, "loss": 0.0001, "step": 29800 }, { "epoch": 0.58, "learning_rate": 4.0362880008281094e-05, "loss": 0.0, "step": 29802 }, { "epoch": 0.58, "learning_rate": 4.036223304802386e-05, "loss": 0.0002, "step": 29804 }, { "epoch": 0.58, "learning_rate": 4.036158608776663e-05, "loss": 0.0, "step": 29806 }, { "epoch": 0.58, "learning_rate": 4.03609391275094e-05, "loss": 0.0, "step": 29808 }, { "epoch": 0.58, "learning_rate": 4.036029216725217e-05, "loss": 0.007, "step": 29810 }, { "epoch": 0.58, "learning_rate": 4.035964520699493e-05, "loss": 0.0048, "step": 29812 }, { "epoch": 0.58, "learning_rate": 4.035899824673771e-05, "loss": 0.0, "step": 29814 }, { "epoch": 0.58, "learning_rate": 4.035835128648047e-05, "loss": 0.0006, "step": 29816 }, { "epoch": 0.58, "learning_rate": 4.0357704326223246e-05, "loss": 0.0157, "step": 29818 }, { "epoch": 0.58, "learning_rate": 4.035705736596601e-05, "loss": 0.0002, "step": 29820 }, { "epoch": 0.58, "learning_rate": 4.035641040570878e-05, "loss": 0.0, "step": 29822 }, { "epoch": 0.58, "learning_rate": 4.035576344545155e-05, "loss": 0.002, "step": 29824 }, { "epoch": 0.58, "learning_rate": 4.0355116485194316e-05, "loss": 0.001, "step": 29826 }, { "epoch": 0.58, "learning_rate": 4.0354469524937085e-05, "loss": 0.0009, "step": 29828 }, { "epoch": 0.58, "learning_rate": 4.0353822564679854e-05, "loss": 0.0, "step": 29830 }, { "epoch": 0.58, "learning_rate": 4.035317560442262e-05, "loss": 0.0027, "step": 29832 }, { "epoch": 0.58, "learning_rate": 4.0352528644165385e-05, "loss": 0.0, "step": 29834 }, { "epoch": 0.58, "learning_rate": 4.035188168390816e-05, "loss": 0.0, "step": 29836 }, { "epoch": 0.58, "learning_rate": 4.035123472365093e-05, "loss": 0.0003, "step": 29838 }, { "epoch": 0.58, "learning_rate": 4.03505877633937e-05, "loss": 0.0, "step": 29840 }, { "epoch": 0.58, "learning_rate": 4.034994080313647e-05, "loss": 0.0004, "step": 29842 }, { "epoch": 0.58, "learning_rate": 4.034929384287923e-05, "loss": 0.0, "step": 29844 }, { "epoch": 0.58, "learning_rate": 4.034864688262201e-05, "loss": 0.0, "step": 29846 }, { "epoch": 0.58, "learning_rate": 4.034799992236477e-05, "loss": 0.0001, "step": 29848 }, { "epoch": 0.58, "learning_rate": 4.034735296210754e-05, "loss": 0.0029, "step": 29850 }, { "epoch": 0.58, "learning_rate": 4.034670600185031e-05, "loss": 0.0001, "step": 29852 }, { "epoch": 0.58, "learning_rate": 4.0346059041593076e-05, "loss": 0.0001, "step": 29854 }, { "epoch": 0.58, "learning_rate": 4.0345412081335845e-05, "loss": 0.0006, "step": 29856 }, { "epoch": 0.58, "learning_rate": 4.0344765121078614e-05, "loss": 0.0001, "step": 29858 }, { "epoch": 0.58, "learning_rate": 4.0344118160821383e-05, "loss": 0.0, "step": 29860 }, { "epoch": 0.58, "learning_rate": 4.034347120056415e-05, "loss": 0.0016, "step": 29862 }, { "epoch": 0.58, "learning_rate": 4.034282424030692e-05, "loss": 0.0, "step": 29864 }, { "epoch": 0.58, "learning_rate": 4.0342177280049684e-05, "loss": 0.0001, "step": 29866 }, { "epoch": 0.58, "learning_rate": 4.034153031979246e-05, "loss": 0.0004, "step": 29868 }, { "epoch": 0.58, "learning_rate": 4.034088335953522e-05, "loss": 0.0, "step": 29870 }, { "epoch": 0.58, "learning_rate": 4.034023639927799e-05, "loss": 0.0, "step": 29872 }, { "epoch": 0.58, "learning_rate": 4.033958943902077e-05, "loss": 0.0, "step": 29874 }, { "epoch": 0.58, "learning_rate": 4.033894247876353e-05, "loss": 0.02, "step": 29876 }, { "epoch": 0.58, "learning_rate": 4.0338295518506305e-05, "loss": 0.0, "step": 29878 }, { "epoch": 0.58, "learning_rate": 4.033764855824907e-05, "loss": 0.0003, "step": 29880 }, { "epoch": 0.58, "learning_rate": 4.0337001597991837e-05, "loss": 0.0001, "step": 29882 }, { "epoch": 0.58, "learning_rate": 4.0336354637734606e-05, "loss": 0.0027, "step": 29884 }, { "epoch": 0.58, "learning_rate": 4.0335707677477375e-05, "loss": 0.0055, "step": 29886 }, { "epoch": 0.58, "learning_rate": 4.0335060717220144e-05, "loss": 0.0001, "step": 29888 }, { "epoch": 0.58, "learning_rate": 4.033441375696291e-05, "loss": 0.023, "step": 29890 }, { "epoch": 0.58, "learning_rate": 4.033376679670568e-05, "loss": 0.0034, "step": 29892 }, { "epoch": 0.58, "learning_rate": 4.0333119836448444e-05, "loss": 0.0026, "step": 29894 }, { "epoch": 0.58, "learning_rate": 4.033247287619122e-05, "loss": 0.0048, "step": 29896 }, { "epoch": 0.58, "learning_rate": 4.033182591593398e-05, "loss": 0.0002, "step": 29898 }, { "epoch": 0.58, "learning_rate": 4.033117895567676e-05, "loss": 0.0, "step": 29900 }, { "epoch": 0.58, "learning_rate": 4.033053199541952e-05, "loss": 0.0133, "step": 29902 }, { "epoch": 0.58, "learning_rate": 4.032988503516229e-05, "loss": 0.0, "step": 29904 }, { "epoch": 0.58, "learning_rate": 4.0329238074905066e-05, "loss": 0.0261, "step": 29906 }, { "epoch": 0.58, "learning_rate": 4.032859111464783e-05, "loss": 0.0001, "step": 29908 }, { "epoch": 0.58, "learning_rate": 4.03279441543906e-05, "loss": 0.0008, "step": 29910 }, { "epoch": 0.58, "learning_rate": 4.0327297194133366e-05, "loss": 0.0003, "step": 29912 }, { "epoch": 0.58, "learning_rate": 4.0326650233876135e-05, "loss": 0.0132, "step": 29914 }, { "epoch": 0.58, "learning_rate": 4.03260032736189e-05, "loss": 0.001, "step": 29916 }, { "epoch": 0.58, "learning_rate": 4.032535631336167e-05, "loss": 0.0001, "step": 29918 }, { "epoch": 0.58, "learning_rate": 4.032470935310444e-05, "loss": 0.0001, "step": 29920 }, { "epoch": 0.58, "learning_rate": 4.032406239284721e-05, "loss": 0.0003, "step": 29922 }, { "epoch": 0.58, "learning_rate": 4.032341543258998e-05, "loss": 0.0044, "step": 29924 }, { "epoch": 0.58, "learning_rate": 4.032276847233274e-05, "loss": 0.0001, "step": 29926 }, { "epoch": 0.58, "learning_rate": 4.032212151207552e-05, "loss": 0.0001, "step": 29928 }, { "epoch": 0.58, "learning_rate": 4.032147455181828e-05, "loss": 0.0001, "step": 29930 }, { "epoch": 0.58, "learning_rate": 4.032082759156105e-05, "loss": 0.0003, "step": 29932 }, { "epoch": 0.58, "learning_rate": 4.032018063130382e-05, "loss": 0.0001, "step": 29934 }, { "epoch": 0.58, "learning_rate": 4.031953367104659e-05, "loss": 0.0, "step": 29936 }, { "epoch": 0.58, "learning_rate": 4.0318886710789364e-05, "loss": 0.0001, "step": 29938 }, { "epoch": 0.58, "learning_rate": 4.0318239750532126e-05, "loss": 0.0002, "step": 29940 }, { "epoch": 0.58, "learning_rate": 4.0317592790274896e-05, "loss": 0.0005, "step": 29942 }, { "epoch": 0.58, "learning_rate": 4.0316945830017665e-05, "loss": 0.0008, "step": 29944 }, { "epoch": 0.58, "learning_rate": 4.0316298869760434e-05, "loss": 0.0004, "step": 29946 }, { "epoch": 0.58, "learning_rate": 4.0315651909503196e-05, "loss": 0.0001, "step": 29948 }, { "epoch": 0.58, "learning_rate": 4.031500494924597e-05, "loss": 0.0001, "step": 29950 }, { "epoch": 0.58, "learning_rate": 4.031435798898874e-05, "loss": 0.0054, "step": 29952 }, { "epoch": 0.58, "learning_rate": 4.03137110287315e-05, "loss": 0.0035, "step": 29954 }, { "epoch": 0.58, "learning_rate": 4.031306406847428e-05, "loss": 0.0001, "step": 29956 }, { "epoch": 0.58, "learning_rate": 4.031241710821704e-05, "loss": 0.0158, "step": 29958 }, { "epoch": 0.58, "learning_rate": 4.031177014795982e-05, "loss": 0.0213, "step": 29960 }, { "epoch": 0.58, "learning_rate": 4.031112318770258e-05, "loss": 0.0006, "step": 29962 }, { "epoch": 0.58, "learning_rate": 4.031047622744535e-05, "loss": 0.0017, "step": 29964 }, { "epoch": 0.58, "learning_rate": 4.030982926718812e-05, "loss": 0.0, "step": 29966 }, { "epoch": 0.58, "learning_rate": 4.030918230693089e-05, "loss": 0.0004, "step": 29968 }, { "epoch": 0.58, "learning_rate": 4.0308535346673656e-05, "loss": 0.0002, "step": 29970 }, { "epoch": 0.58, "learning_rate": 4.0307888386416425e-05, "loss": 0.0001, "step": 29972 }, { "epoch": 0.58, "learning_rate": 4.0307241426159194e-05, "loss": 0.0001, "step": 29974 }, { "epoch": 0.58, "learning_rate": 4.0306594465901956e-05, "loss": 0.0, "step": 29976 }, { "epoch": 0.58, "learning_rate": 4.030594750564473e-05, "loss": 0.0003, "step": 29978 }, { "epoch": 0.58, "learning_rate": 4.0305300545387495e-05, "loss": 0.0025, "step": 29980 }, { "epoch": 0.58, "learning_rate": 4.030465358513027e-05, "loss": 0.0001, "step": 29982 }, { "epoch": 0.58, "learning_rate": 4.030400662487304e-05, "loss": 0.0, "step": 29984 }, { "epoch": 0.58, "learning_rate": 4.03033596646158e-05, "loss": 0.0112, "step": 29986 }, { "epoch": 0.58, "learning_rate": 4.030271270435858e-05, "loss": 0.0, "step": 29988 }, { "epoch": 0.58, "learning_rate": 4.030206574410134e-05, "loss": 0.0016, "step": 29990 }, { "epoch": 0.58, "learning_rate": 4.030141878384411e-05, "loss": 0.0003, "step": 29992 }, { "epoch": 0.58, "learning_rate": 4.030077182358688e-05, "loss": 0.0001, "step": 29994 }, { "epoch": 0.58, "learning_rate": 4.030012486332965e-05, "loss": 0.0002, "step": 29996 }, { "epoch": 0.58, "learning_rate": 4.0299477903072416e-05, "loss": 0.0016, "step": 29998 }, { "epoch": 0.58, "learning_rate": 4.0298830942815185e-05, "loss": 0.0, "step": 30000 }, { "epoch": 0.58, "learning_rate": 4.0298183982557955e-05, "loss": 0.0007, "step": 30002 }, { "epoch": 0.58, "learning_rate": 4.0297537022300724e-05, "loss": 0.0, "step": 30004 }, { "epoch": 0.58, "learning_rate": 4.029689006204349e-05, "loss": 0.0104, "step": 30006 }, { "epoch": 0.58, "learning_rate": 4.0296243101786255e-05, "loss": 0.0, "step": 30008 }, { "epoch": 0.58, "learning_rate": 4.029559614152903e-05, "loss": 0.0, "step": 30010 }, { "epoch": 0.58, "learning_rate": 4.029494918127179e-05, "loss": 0.0001, "step": 30012 }, { "epoch": 0.58, "learning_rate": 4.029430222101456e-05, "loss": 0.0004, "step": 30014 }, { "epoch": 0.58, "learning_rate": 4.029365526075734e-05, "loss": 0.0002, "step": 30016 }, { "epoch": 0.58, "learning_rate": 4.02930083005001e-05, "loss": 0.0001, "step": 30018 }, { "epoch": 0.58, "learning_rate": 4.0292361340242876e-05, "loss": 0.0006, "step": 30020 }, { "epoch": 0.58, "learning_rate": 4.029171437998564e-05, "loss": 0.0009, "step": 30022 }, { "epoch": 0.58, "learning_rate": 4.029106741972841e-05, "loss": 0.0127, "step": 30024 }, { "epoch": 0.58, "learning_rate": 4.029042045947118e-05, "loss": 0.0001, "step": 30026 }, { "epoch": 0.58, "learning_rate": 4.0289773499213946e-05, "loss": 0.0336, "step": 30028 }, { "epoch": 0.58, "learning_rate": 4.0289126538956715e-05, "loss": 0.0002, "step": 30030 }, { "epoch": 0.58, "learning_rate": 4.0288479578699484e-05, "loss": 0.0003, "step": 30032 }, { "epoch": 0.58, "learning_rate": 4.028783261844225e-05, "loss": 0.0002, "step": 30034 }, { "epoch": 0.58, "learning_rate": 4.0287185658185015e-05, "loss": 0.0054, "step": 30036 }, { "epoch": 0.58, "learning_rate": 4.028653869792779e-05, "loss": 0.0003, "step": 30038 }, { "epoch": 0.58, "learning_rate": 4.0285891737670554e-05, "loss": 0.0062, "step": 30040 }, { "epoch": 0.58, "learning_rate": 4.028524477741333e-05, "loss": 0.0, "step": 30042 }, { "epoch": 0.58, "learning_rate": 4.028459781715609e-05, "loss": 0.007, "step": 30044 }, { "epoch": 0.58, "learning_rate": 4.028395085689886e-05, "loss": 0.0002, "step": 30046 }, { "epoch": 0.58, "learning_rate": 4.028330389664163e-05, "loss": 0.0001, "step": 30048 }, { "epoch": 0.58, "learning_rate": 4.02826569363844e-05, "loss": 0.0006, "step": 30050 }, { "epoch": 0.58, "learning_rate": 4.028200997612717e-05, "loss": 0.0, "step": 30052 }, { "epoch": 0.58, "learning_rate": 4.028136301586994e-05, "loss": 0.0, "step": 30054 }, { "epoch": 0.58, "learning_rate": 4.0280716055612706e-05, "loss": 0.0, "step": 30056 }, { "epoch": 0.58, "learning_rate": 4.028006909535547e-05, "loss": 0.002, "step": 30058 }, { "epoch": 0.58, "learning_rate": 4.0279422135098244e-05, "loss": 0.0069, "step": 30060 }, { "epoch": 0.58, "learning_rate": 4.0278775174841014e-05, "loss": 0.024, "step": 30062 }, { "epoch": 0.58, "learning_rate": 4.027812821458378e-05, "loss": 0.0, "step": 30064 }, { "epoch": 0.58, "learning_rate": 4.027748125432655e-05, "loss": 0.0027, "step": 30066 }, { "epoch": 0.58, "learning_rate": 4.0276834294069314e-05, "loss": 0.0008, "step": 30068 }, { "epoch": 0.58, "learning_rate": 4.027618733381209e-05, "loss": 0.0003, "step": 30070 }, { "epoch": 0.58, "learning_rate": 4.027554037355485e-05, "loss": 0.0014, "step": 30072 }, { "epoch": 0.58, "learning_rate": 4.027489341329762e-05, "loss": 0.0, "step": 30074 }, { "epoch": 0.58, "learning_rate": 4.027424645304039e-05, "loss": 0.0004, "step": 30076 }, { "epoch": 0.58, "learning_rate": 4.027359949278316e-05, "loss": 0.0, "step": 30078 }, { "epoch": 0.58, "learning_rate": 4.027295253252593e-05, "loss": 0.0, "step": 30080 }, { "epoch": 0.58, "learning_rate": 4.02723055722687e-05, "loss": 0.0007, "step": 30082 }, { "epoch": 0.58, "learning_rate": 4.027165861201147e-05, "loss": 0.0002, "step": 30084 }, { "epoch": 0.58, "learning_rate": 4.0271011651754236e-05, "loss": 0.0006, "step": 30086 }, { "epoch": 0.58, "learning_rate": 4.0270364691497005e-05, "loss": 0.0003, "step": 30088 }, { "epoch": 0.58, "learning_rate": 4.026971773123977e-05, "loss": 0.0094, "step": 30090 }, { "epoch": 0.58, "learning_rate": 4.026907077098254e-05, "loss": 0.0028, "step": 30092 }, { "epoch": 0.58, "learning_rate": 4.0268423810725305e-05, "loss": 0.0002, "step": 30094 }, { "epoch": 0.58, "learning_rate": 4.0267776850468074e-05, "loss": 0.0074, "step": 30096 }, { "epoch": 0.58, "learning_rate": 4.026712989021085e-05, "loss": 0.0001, "step": 30098 }, { "epoch": 0.58, "learning_rate": 4.026648292995361e-05, "loss": 0.0, "step": 30100 }, { "epoch": 0.58, "learning_rate": 4.026583596969639e-05, "loss": 0.0, "step": 30102 }, { "epoch": 0.58, "learning_rate": 4.026518900943915e-05, "loss": 0.0082, "step": 30104 }, { "epoch": 0.58, "learning_rate": 4.026454204918192e-05, "loss": 0.0033, "step": 30106 }, { "epoch": 0.58, "learning_rate": 4.026389508892469e-05, "loss": 0.0063, "step": 30108 }, { "epoch": 0.58, "learning_rate": 4.026324812866746e-05, "loss": 0.0001, "step": 30110 }, { "epoch": 0.58, "learning_rate": 4.026260116841023e-05, "loss": 0.0005, "step": 30112 }, { "epoch": 0.58, "learning_rate": 4.0261954208152996e-05, "loss": 0.0, "step": 30114 }, { "epoch": 0.58, "learning_rate": 4.0261307247895765e-05, "loss": 0.027, "step": 30116 }, { "epoch": 0.58, "learning_rate": 4.026066028763853e-05, "loss": 0.0002, "step": 30118 }, { "epoch": 0.58, "learning_rate": 4.0260013327381303e-05, "loss": 0.0001, "step": 30120 }, { "epoch": 0.58, "learning_rate": 4.0259366367124066e-05, "loss": 0.0302, "step": 30122 }, { "epoch": 0.58, "learning_rate": 4.025871940686684e-05, "loss": 0.0004, "step": 30124 }, { "epoch": 0.58, "learning_rate": 4.0258072446609604e-05, "loss": 0.0001, "step": 30126 }, { "epoch": 0.58, "learning_rate": 4.025742548635237e-05, "loss": 0.0003, "step": 30128 }, { "epoch": 0.58, "learning_rate": 4.025677852609515e-05, "loss": 0.0553, "step": 30130 }, { "epoch": 0.58, "learning_rate": 4.025613156583791e-05, "loss": 0.0, "step": 30132 }, { "epoch": 0.58, "learning_rate": 4.025548460558068e-05, "loss": 0.0, "step": 30134 }, { "epoch": 0.58, "learning_rate": 4.025483764532345e-05, "loss": 0.0062, "step": 30136 }, { "epoch": 0.58, "learning_rate": 4.025419068506622e-05, "loss": 0.0001, "step": 30138 }, { "epoch": 0.58, "learning_rate": 4.025354372480899e-05, "loss": 0.0002, "step": 30140 }, { "epoch": 0.59, "learning_rate": 4.0252896764551757e-05, "loss": 0.0001, "step": 30142 }, { "epoch": 0.59, "learning_rate": 4.0252249804294526e-05, "loss": 0.0004, "step": 30144 }, { "epoch": 0.59, "learning_rate": 4.0251602844037295e-05, "loss": 0.0001, "step": 30146 }, { "epoch": 0.59, "learning_rate": 4.0250955883780064e-05, "loss": 0.0001, "step": 30148 }, { "epoch": 0.59, "learning_rate": 4.0250308923522826e-05, "loss": 0.0001, "step": 30150 }, { "epoch": 0.59, "learning_rate": 4.02496619632656e-05, "loss": 0.0011, "step": 30152 }, { "epoch": 0.59, "learning_rate": 4.0249015003008364e-05, "loss": 0.0, "step": 30154 }, { "epoch": 0.59, "learning_rate": 4.024836804275113e-05, "loss": 0.0117, "step": 30156 }, { "epoch": 0.59, "learning_rate": 4.02477210824939e-05, "loss": 0.0001, "step": 30158 }, { "epoch": 0.59, "learning_rate": 4.024707412223667e-05, "loss": 0.0001, "step": 30160 }, { "epoch": 0.59, "learning_rate": 4.024642716197945e-05, "loss": 0.0001, "step": 30162 }, { "epoch": 0.59, "learning_rate": 4.024578020172221e-05, "loss": 0.0, "step": 30164 }, { "epoch": 0.59, "learning_rate": 4.024513324146498e-05, "loss": 0.0002, "step": 30166 }, { "epoch": 0.59, "learning_rate": 4.024448628120775e-05, "loss": 0.0001, "step": 30168 }, { "epoch": 0.59, "learning_rate": 4.024383932095052e-05, "loss": 0.007, "step": 30170 }, { "epoch": 0.59, "learning_rate": 4.024319236069328e-05, "loss": 0.004, "step": 30172 }, { "epoch": 0.59, "learning_rate": 4.0242545400436055e-05, "loss": 0.0005, "step": 30174 }, { "epoch": 0.59, "learning_rate": 4.0241898440178824e-05, "loss": 0.0004, "step": 30176 }, { "epoch": 0.59, "learning_rate": 4.0241251479921587e-05, "loss": 0.0, "step": 30178 }, { "epoch": 0.59, "learning_rate": 4.024060451966436e-05, "loss": 0.0, "step": 30180 }, { "epoch": 0.59, "learning_rate": 4.0239957559407125e-05, "loss": 0.0041, "step": 30182 }, { "epoch": 0.59, "learning_rate": 4.02393105991499e-05, "loss": 0.0002, "step": 30184 }, { "epoch": 0.59, "learning_rate": 4.023866363889266e-05, "loss": 0.0001, "step": 30186 }, { "epoch": 0.59, "learning_rate": 4.023801667863543e-05, "loss": 0.0001, "step": 30188 }, { "epoch": 0.59, "learning_rate": 4.02373697183782e-05, "loss": 0.0, "step": 30190 }, { "epoch": 0.59, "learning_rate": 4.023672275812097e-05, "loss": 0.0, "step": 30192 }, { "epoch": 0.59, "learning_rate": 4.023607579786374e-05, "loss": 0.0028, "step": 30194 }, { "epoch": 0.59, "learning_rate": 4.023542883760651e-05, "loss": 0.0003, "step": 30196 }, { "epoch": 0.59, "learning_rate": 4.023478187734928e-05, "loss": 0.0014, "step": 30198 }, { "epoch": 0.59, "learning_rate": 4.023413491709204e-05, "loss": 0.0007, "step": 30200 }, { "epoch": 0.59, "learning_rate": 4.0233487956834816e-05, "loss": 0.0103, "step": 30202 }, { "epoch": 0.59, "learning_rate": 4.023284099657758e-05, "loss": 0.012, "step": 30204 }, { "epoch": 0.59, "learning_rate": 4.0232194036320354e-05, "loss": 0.0043, "step": 30206 }, { "epoch": 0.59, "learning_rate": 4.023154707606312e-05, "loss": 0.0, "step": 30208 }, { "epoch": 0.59, "learning_rate": 4.0230900115805885e-05, "loss": 0.0105, "step": 30210 }, { "epoch": 0.59, "learning_rate": 4.023025315554866e-05, "loss": 0.0004, "step": 30212 }, { "epoch": 0.59, "learning_rate": 4.022960619529142e-05, "loss": 0.0011, "step": 30214 }, { "epoch": 0.59, "learning_rate": 4.022895923503419e-05, "loss": 0.0001, "step": 30216 }, { "epoch": 0.59, "learning_rate": 4.022831227477696e-05, "loss": 0.0001, "step": 30218 }, { "epoch": 0.59, "learning_rate": 4.022766531451973e-05, "loss": 0.0118, "step": 30220 }, { "epoch": 0.59, "learning_rate": 4.02270183542625e-05, "loss": 0.0013, "step": 30222 }, { "epoch": 0.59, "learning_rate": 4.022637139400527e-05, "loss": 0.0015, "step": 30224 }, { "epoch": 0.59, "learning_rate": 4.022572443374804e-05, "loss": 0.0047, "step": 30226 }, { "epoch": 0.59, "learning_rate": 4.022507747349081e-05, "loss": 0.0001, "step": 30228 }, { "epoch": 0.59, "learning_rate": 4.0224430513233576e-05, "loss": 0.0001, "step": 30230 }, { "epoch": 0.59, "learning_rate": 4.022378355297634e-05, "loss": 0.0035, "step": 30232 }, { "epoch": 0.59, "learning_rate": 4.0223136592719114e-05, "loss": 0.0002, "step": 30234 }, { "epoch": 0.59, "learning_rate": 4.0222489632461876e-05, "loss": 0.0065, "step": 30236 }, { "epoch": 0.59, "learning_rate": 4.0221842672204646e-05, "loss": 0.0002, "step": 30238 }, { "epoch": 0.59, "learning_rate": 4.022119571194742e-05, "loss": 0.0, "step": 30240 }, { "epoch": 0.59, "learning_rate": 4.0220548751690184e-05, "loss": 0.0021, "step": 30242 }, { "epoch": 0.59, "learning_rate": 4.021990179143296e-05, "loss": 0.0001, "step": 30244 }, { "epoch": 0.59, "learning_rate": 4.021925483117572e-05, "loss": 0.0003, "step": 30246 }, { "epoch": 0.59, "learning_rate": 4.021860787091849e-05, "loss": 0.0012, "step": 30248 }, { "epoch": 0.59, "learning_rate": 4.021796091066126e-05, "loss": 0.0042, "step": 30250 }, { "epoch": 0.59, "learning_rate": 4.021731395040403e-05, "loss": 0.0061, "step": 30252 }, { "epoch": 0.59, "learning_rate": 4.02166669901468e-05, "loss": 0.0088, "step": 30254 }, { "epoch": 0.59, "learning_rate": 4.021602002988957e-05, "loss": 0.0007, "step": 30256 }, { "epoch": 0.59, "learning_rate": 4.0215373069632336e-05, "loss": 0.0024, "step": 30258 }, { "epoch": 0.59, "learning_rate": 4.02147261093751e-05, "loss": 0.0, "step": 30260 }, { "epoch": 0.59, "learning_rate": 4.0214079149117875e-05, "loss": 0.0001, "step": 30262 }, { "epoch": 0.59, "learning_rate": 4.021343218886064e-05, "loss": 0.0001, "step": 30264 }, { "epoch": 0.59, "learning_rate": 4.021278522860341e-05, "loss": 0.0211, "step": 30266 }, { "epoch": 0.59, "learning_rate": 4.0212138268346175e-05, "loss": 0.0132, "step": 30268 }, { "epoch": 0.59, "learning_rate": 4.0211491308088944e-05, "loss": 0.0, "step": 30270 }, { "epoch": 0.59, "learning_rate": 4.021084434783171e-05, "loss": 0.0, "step": 30272 }, { "epoch": 0.59, "learning_rate": 4.021019738757448e-05, "loss": 0.0001, "step": 30274 }, { "epoch": 0.59, "learning_rate": 4.020955042731725e-05, "loss": 0.0, "step": 30276 }, { "epoch": 0.59, "learning_rate": 4.020890346706002e-05, "loss": 0.0, "step": 30278 }, { "epoch": 0.59, "learning_rate": 4.020825650680279e-05, "loss": 0.0034, "step": 30280 }, { "epoch": 0.59, "learning_rate": 4.020760954654555e-05, "loss": 0.0001, "step": 30282 }, { "epoch": 0.59, "learning_rate": 4.020696258628833e-05, "loss": 0.0028, "step": 30284 }, { "epoch": 0.59, "learning_rate": 4.02063156260311e-05, "loss": 0.0155, "step": 30286 }, { "epoch": 0.59, "learning_rate": 4.0205668665773866e-05, "loss": 0.0001, "step": 30288 }, { "epoch": 0.59, "learning_rate": 4.0205021705516635e-05, "loss": 0.0, "step": 30290 }, { "epoch": 0.59, "learning_rate": 4.02043747452594e-05, "loss": 0.0058, "step": 30292 }, { "epoch": 0.59, "learning_rate": 4.020372778500217e-05, "loss": 0.0, "step": 30294 }, { "epoch": 0.59, "learning_rate": 4.0203080824744935e-05, "loss": 0.0001, "step": 30296 }, { "epoch": 0.59, "learning_rate": 4.0202433864487704e-05, "loss": 0.0048, "step": 30298 }, { "epoch": 0.59, "learning_rate": 4.0201786904230474e-05, "loss": 0.0001, "step": 30300 }, { "epoch": 0.59, "learning_rate": 4.020113994397324e-05, "loss": 0.0001, "step": 30302 }, { "epoch": 0.59, "learning_rate": 4.020049298371601e-05, "loss": 0.0112, "step": 30304 }, { "epoch": 0.59, "learning_rate": 4.019984602345878e-05, "loss": 0.0001, "step": 30306 }, { "epoch": 0.59, "learning_rate": 4.019919906320155e-05, "loss": 0.0, "step": 30308 }, { "epoch": 0.59, "learning_rate": 4.019855210294432e-05, "loss": 0.0, "step": 30310 }, { "epoch": 0.59, "learning_rate": 4.019790514268709e-05, "loss": 0.0005, "step": 30312 }, { "epoch": 0.59, "learning_rate": 4.019725818242985e-05, "loss": 0.0, "step": 30314 }, { "epoch": 0.59, "learning_rate": 4.0196611222172626e-05, "loss": 0.0004, "step": 30316 }, { "epoch": 0.59, "learning_rate": 4.0195964261915395e-05, "loss": 0.0001, "step": 30318 }, { "epoch": 0.59, "learning_rate": 4.019531730165816e-05, "loss": 0.009, "step": 30320 }, { "epoch": 0.59, "learning_rate": 4.0194670341400934e-05, "loss": 0.0, "step": 30322 }, { "epoch": 0.59, "learning_rate": 4.0194023381143696e-05, "loss": 0.0, "step": 30324 }, { "epoch": 0.59, "learning_rate": 4.019337642088647e-05, "loss": 0.0001, "step": 30326 }, { "epoch": 0.59, "learning_rate": 4.0192729460629234e-05, "loss": 0.0019, "step": 30328 }, { "epoch": 0.59, "learning_rate": 4.0192082500372e-05, "loss": 0.0013, "step": 30330 }, { "epoch": 0.59, "learning_rate": 4.019143554011477e-05, "loss": 0.0002, "step": 30332 }, { "epoch": 0.59, "learning_rate": 4.019078857985754e-05, "loss": 0.0013, "step": 30334 }, { "epoch": 0.59, "learning_rate": 4.019014161960031e-05, "loss": 0.0001, "step": 30336 }, { "epoch": 0.59, "learning_rate": 4.018949465934308e-05, "loss": 0.0, "step": 30338 }, { "epoch": 0.59, "learning_rate": 4.018884769908585e-05, "loss": 0.0, "step": 30340 }, { "epoch": 0.59, "learning_rate": 4.018820073882861e-05, "loss": 0.0, "step": 30342 }, { "epoch": 0.59, "learning_rate": 4.018755377857139e-05, "loss": 0.0001, "step": 30344 }, { "epoch": 0.59, "learning_rate": 4.018690681831415e-05, "loss": 0.0036, "step": 30346 }, { "epoch": 0.59, "learning_rate": 4.0186259858056925e-05, "loss": 0.0008, "step": 30348 }, { "epoch": 0.59, "learning_rate": 4.018561289779969e-05, "loss": 0.0002, "step": 30350 }, { "epoch": 0.59, "learning_rate": 4.0184965937542456e-05, "loss": 0.0001, "step": 30352 }, { "epoch": 0.59, "learning_rate": 4.018431897728523e-05, "loss": 0.0, "step": 30354 }, { "epoch": 0.59, "learning_rate": 4.0183672017027994e-05, "loss": 0.0064, "step": 30356 }, { "epoch": 0.59, "learning_rate": 4.0183025056770763e-05, "loss": 0.0, "step": 30358 }, { "epoch": 0.59, "learning_rate": 4.018237809651353e-05, "loss": 0.0006, "step": 30360 }, { "epoch": 0.59, "learning_rate": 4.01817311362563e-05, "loss": 0.0, "step": 30362 }, { "epoch": 0.59, "learning_rate": 4.018108417599907e-05, "loss": 0.001, "step": 30364 }, { "epoch": 0.59, "learning_rate": 4.018043721574184e-05, "loss": 0.0001, "step": 30366 }, { "epoch": 0.59, "learning_rate": 4.017979025548461e-05, "loss": 0.0005, "step": 30368 }, { "epoch": 0.59, "learning_rate": 4.017914329522738e-05, "loss": 0.0, "step": 30370 }, { "epoch": 0.59, "learning_rate": 4.017849633497015e-05, "loss": 0.0, "step": 30372 }, { "epoch": 0.59, "learning_rate": 4.017784937471291e-05, "loss": 0.0079, "step": 30374 }, { "epoch": 0.59, "learning_rate": 4.0177202414455685e-05, "loss": 0.0, "step": 30376 }, { "epoch": 0.59, "learning_rate": 4.017655545419845e-05, "loss": 0.0084, "step": 30378 }, { "epoch": 0.59, "learning_rate": 4.017590849394122e-05, "loss": 0.0012, "step": 30380 }, { "epoch": 0.59, "learning_rate": 4.0175261533683986e-05, "loss": 0.0003, "step": 30382 }, { "epoch": 0.59, "learning_rate": 4.0174614573426755e-05, "loss": 0.0, "step": 30384 }, { "epoch": 0.59, "learning_rate": 4.017396761316953e-05, "loss": 0.0002, "step": 30386 }, { "epoch": 0.59, "learning_rate": 4.017332065291229e-05, "loss": 0.0, "step": 30388 }, { "epoch": 0.59, "learning_rate": 4.017267369265506e-05, "loss": 0.0001, "step": 30390 }, { "epoch": 0.59, "learning_rate": 4.017202673239783e-05, "loss": 0.0011, "step": 30392 }, { "epoch": 0.59, "learning_rate": 4.01713797721406e-05, "loss": 0.0, "step": 30394 }, { "epoch": 0.59, "learning_rate": 4.017073281188336e-05, "loss": 0.0, "step": 30396 }, { "epoch": 0.59, "learning_rate": 4.017008585162614e-05, "loss": 0.0, "step": 30398 }, { "epoch": 0.59, "learning_rate": 4.016943889136891e-05, "loss": 0.0, "step": 30400 }, { "epoch": 0.59, "learning_rate": 4.016879193111167e-05, "loss": 0.0009, "step": 30402 }, { "epoch": 0.59, "learning_rate": 4.0168144970854446e-05, "loss": 0.0204, "step": 30404 }, { "epoch": 0.59, "learning_rate": 4.016749801059721e-05, "loss": 0.0008, "step": 30406 }, { "epoch": 0.59, "learning_rate": 4.0166851050339984e-05, "loss": 0.0001, "step": 30408 }, { "epoch": 0.59, "learning_rate": 4.0166204090082746e-05, "loss": 0.0001, "step": 30410 }, { "epoch": 0.59, "learning_rate": 4.0165557129825515e-05, "loss": 0.0002, "step": 30412 }, { "epoch": 0.59, "learning_rate": 4.0164910169568284e-05, "loss": 0.0041, "step": 30414 }, { "epoch": 0.59, "learning_rate": 4.016426320931105e-05, "loss": 0.0012, "step": 30416 }, { "epoch": 0.59, "learning_rate": 4.016361624905382e-05, "loss": 0.0, "step": 30418 }, { "epoch": 0.59, "learning_rate": 4.016296928879659e-05, "loss": 0.0001, "step": 30420 }, { "epoch": 0.59, "learning_rate": 4.016232232853936e-05, "loss": 0.0, "step": 30422 }, { "epoch": 0.59, "learning_rate": 4.016167536828212e-05, "loss": 0.0159, "step": 30424 }, { "epoch": 0.59, "learning_rate": 4.01610284080249e-05, "loss": 0.0, "step": 30426 }, { "epoch": 0.59, "learning_rate": 4.016038144776766e-05, "loss": 0.0062, "step": 30428 }, { "epoch": 0.59, "learning_rate": 4.015973448751044e-05, "loss": 0.01, "step": 30430 }, { "epoch": 0.59, "learning_rate": 4.0159087527253206e-05, "loss": 0.0001, "step": 30432 }, { "epoch": 0.59, "learning_rate": 4.015844056699597e-05, "loss": 0.0002, "step": 30434 }, { "epoch": 0.59, "learning_rate": 4.0157793606738744e-05, "loss": 0.0107, "step": 30436 }, { "epoch": 0.59, "learning_rate": 4.0157146646481507e-05, "loss": 0.0009, "step": 30438 }, { "epoch": 0.59, "learning_rate": 4.0156499686224276e-05, "loss": 0.0, "step": 30440 }, { "epoch": 0.59, "learning_rate": 4.0155852725967045e-05, "loss": 0.002, "step": 30442 }, { "epoch": 0.59, "learning_rate": 4.0155205765709814e-05, "loss": 0.0003, "step": 30444 }, { "epoch": 0.59, "learning_rate": 4.015455880545258e-05, "loss": 0.0009, "step": 30446 }, { "epoch": 0.59, "learning_rate": 4.015391184519535e-05, "loss": 0.0004, "step": 30448 }, { "epoch": 0.59, "learning_rate": 4.015326488493812e-05, "loss": 0.0002, "step": 30450 }, { "epoch": 0.59, "learning_rate": 4.015261792468089e-05, "loss": 0.0519, "step": 30452 }, { "epoch": 0.59, "learning_rate": 4.015197096442366e-05, "loss": 0.0027, "step": 30454 }, { "epoch": 0.59, "learning_rate": 4.015132400416642e-05, "loss": 0.0001, "step": 30456 }, { "epoch": 0.59, "learning_rate": 4.01506770439092e-05, "loss": 0.0, "step": 30458 }, { "epoch": 0.59, "learning_rate": 4.015003008365196e-05, "loss": 0.0001, "step": 30460 }, { "epoch": 0.59, "learning_rate": 4.014938312339473e-05, "loss": 0.0116, "step": 30462 }, { "epoch": 0.59, "learning_rate": 4.0148736163137505e-05, "loss": 0.0, "step": 30464 }, { "epoch": 0.59, "learning_rate": 4.014808920288027e-05, "loss": 0.0, "step": 30466 }, { "epoch": 0.59, "learning_rate": 4.014744224262304e-05, "loss": 0.0092, "step": 30468 }, { "epoch": 0.59, "learning_rate": 4.0146795282365805e-05, "loss": 0.0099, "step": 30470 }, { "epoch": 0.59, "learning_rate": 4.0146148322108574e-05, "loss": 0.015, "step": 30472 }, { "epoch": 0.59, "learning_rate": 4.014550136185134e-05, "loss": 0.0, "step": 30474 }, { "epoch": 0.59, "learning_rate": 4.014485440159411e-05, "loss": 0.0002, "step": 30476 }, { "epoch": 0.59, "learning_rate": 4.014420744133688e-05, "loss": 0.0, "step": 30478 }, { "epoch": 0.59, "learning_rate": 4.014356048107965e-05, "loss": 0.0, "step": 30480 }, { "epoch": 0.59, "learning_rate": 4.014291352082242e-05, "loss": 0.0, "step": 30482 }, { "epoch": 0.59, "learning_rate": 4.014226656056518e-05, "loss": 0.0001, "step": 30484 }, { "epoch": 0.59, "learning_rate": 4.014161960030796e-05, "loss": 0.0, "step": 30486 }, { "epoch": 0.59, "learning_rate": 4.014097264005072e-05, "loss": 0.0001, "step": 30488 }, { "epoch": 0.59, "learning_rate": 4.0140325679793496e-05, "loss": 0.0116, "step": 30490 }, { "epoch": 0.59, "learning_rate": 4.013967871953626e-05, "loss": 0.0, "step": 30492 }, { "epoch": 0.59, "learning_rate": 4.013903175927903e-05, "loss": 0.0, "step": 30494 }, { "epoch": 0.59, "learning_rate": 4.01383847990218e-05, "loss": 0.0007, "step": 30496 }, { "epoch": 0.59, "learning_rate": 4.0137737838764565e-05, "loss": 0.0147, "step": 30498 }, { "epoch": 0.59, "learning_rate": 4.0137090878507335e-05, "loss": 0.0001, "step": 30500 }, { "epoch": 0.59, "learning_rate": 4.0136443918250104e-05, "loss": 0.0002, "step": 30502 }, { "epoch": 0.59, "learning_rate": 4.013579695799287e-05, "loss": 0.0012, "step": 30504 }, { "epoch": 0.59, "learning_rate": 4.013514999773564e-05, "loss": 0.0002, "step": 30506 }, { "epoch": 0.59, "learning_rate": 4.013450303747841e-05, "loss": 0.0002, "step": 30508 }, { "epoch": 0.59, "learning_rate": 4.013385607722118e-05, "loss": 0.0001, "step": 30510 }, { "epoch": 0.59, "learning_rate": 4.013320911696395e-05, "loss": 0.0001, "step": 30512 }, { "epoch": 0.59, "learning_rate": 4.013256215670672e-05, "loss": 0.0001, "step": 30514 }, { "epoch": 0.59, "learning_rate": 4.013191519644948e-05, "loss": 0.0072, "step": 30516 }, { "epoch": 0.59, "learning_rate": 4.0131268236192256e-05, "loss": 0.0, "step": 30518 }, { "epoch": 0.59, "learning_rate": 4.013062127593502e-05, "loss": 0.0, "step": 30520 }, { "epoch": 0.59, "learning_rate": 4.012997431567779e-05, "loss": 0.0, "step": 30522 }, { "epoch": 0.59, "learning_rate": 4.012932735542056e-05, "loss": 0.0008, "step": 30524 }, { "epoch": 0.59, "learning_rate": 4.0128680395163326e-05, "loss": 0.0062, "step": 30526 }, { "epoch": 0.59, "learning_rate": 4.0128033434906095e-05, "loss": 0.0009, "step": 30528 }, { "epoch": 0.59, "learning_rate": 4.0127386474648864e-05, "loss": 0.0, "step": 30530 }, { "epoch": 0.59, "learning_rate": 4.012673951439163e-05, "loss": 0.0001, "step": 30532 }, { "epoch": 0.59, "learning_rate": 4.01260925541344e-05, "loss": 0.0001, "step": 30534 }, { "epoch": 0.59, "learning_rate": 4.012544559387717e-05, "loss": 0.0003, "step": 30536 }, { "epoch": 0.59, "learning_rate": 4.0124798633619934e-05, "loss": 0.0, "step": 30538 }, { "epoch": 0.59, "learning_rate": 4.012415167336271e-05, "loss": 0.0002, "step": 30540 }, { "epoch": 0.59, "learning_rate": 4.012350471310548e-05, "loss": 0.0, "step": 30542 }, { "epoch": 0.59, "learning_rate": 4.012285775284824e-05, "loss": 0.0109, "step": 30544 }, { "epoch": 0.59, "learning_rate": 4.012221079259102e-05, "loss": 0.0005, "step": 30546 }, { "epoch": 0.59, "learning_rate": 4.012156383233378e-05, "loss": 0.0001, "step": 30548 }, { "epoch": 0.59, "learning_rate": 4.0120916872076555e-05, "loss": 0.0009, "step": 30550 }, { "epoch": 0.59, "learning_rate": 4.012026991181932e-05, "loss": 0.0, "step": 30552 }, { "epoch": 0.59, "learning_rate": 4.0119622951562086e-05, "loss": 0.0, "step": 30554 }, { "epoch": 0.59, "learning_rate": 4.0118975991304855e-05, "loss": 0.0038, "step": 30556 }, { "epoch": 0.59, "learning_rate": 4.0118329031047624e-05, "loss": 0.0001, "step": 30558 }, { "epoch": 0.59, "learning_rate": 4.0117682070790394e-05, "loss": 0.0036, "step": 30560 }, { "epoch": 0.59, "learning_rate": 4.011703511053316e-05, "loss": 0.0, "step": 30562 }, { "epoch": 0.59, "learning_rate": 4.011638815027593e-05, "loss": 0.0, "step": 30564 }, { "epoch": 0.59, "learning_rate": 4.0115741190018694e-05, "loss": 0.0, "step": 30566 }, { "epoch": 0.59, "learning_rate": 4.011509422976147e-05, "loss": 0.0, "step": 30568 }, { "epoch": 0.59, "learning_rate": 4.011444726950423e-05, "loss": 0.0088, "step": 30570 }, { "epoch": 0.59, "learning_rate": 4.011380030924701e-05, "loss": 0.0069, "step": 30572 }, { "epoch": 0.59, "learning_rate": 4.011315334898977e-05, "loss": 0.0035, "step": 30574 }, { "epoch": 0.59, "learning_rate": 4.011250638873254e-05, "loss": 0.0001, "step": 30576 }, { "epoch": 0.59, "learning_rate": 4.0111859428475315e-05, "loss": 0.0131, "step": 30578 }, { "epoch": 0.59, "learning_rate": 4.011121246821808e-05, "loss": 0.0, "step": 30580 }, { "epoch": 0.59, "learning_rate": 4.011056550796085e-05, "loss": 0.0, "step": 30582 }, { "epoch": 0.59, "learning_rate": 4.0109918547703616e-05, "loss": 0.0, "step": 30584 }, { "epoch": 0.59, "learning_rate": 4.0109271587446385e-05, "loss": 0.004, "step": 30586 }, { "epoch": 0.59, "learning_rate": 4.0108624627189154e-05, "loss": 0.0, "step": 30588 }, { "epoch": 0.59, "learning_rate": 4.010797766693192e-05, "loss": 0.003, "step": 30590 }, { "epoch": 0.59, "learning_rate": 4.010733070667469e-05, "loss": 0.0176, "step": 30592 }, { "epoch": 0.59, "learning_rate": 4.010668374641746e-05, "loss": 0.0051, "step": 30594 }, { "epoch": 0.59, "learning_rate": 4.010603678616023e-05, "loss": 0.0, "step": 30596 }, { "epoch": 0.59, "learning_rate": 4.010538982590299e-05, "loss": 0.0052, "step": 30598 }, { "epoch": 0.59, "learning_rate": 4.010474286564577e-05, "loss": 0.0141, "step": 30600 }, { "epoch": 0.59, "learning_rate": 4.010409590538853e-05, "loss": 0.0002, "step": 30602 }, { "epoch": 0.59, "learning_rate": 4.01034489451313e-05, "loss": 0.0001, "step": 30604 }, { "epoch": 0.59, "learning_rate": 4.010280198487407e-05, "loss": 0.0002, "step": 30606 }, { "epoch": 0.59, "learning_rate": 4.010215502461684e-05, "loss": 0.0005, "step": 30608 }, { "epoch": 0.59, "learning_rate": 4.0101508064359614e-05, "loss": 0.0002, "step": 30610 }, { "epoch": 0.59, "learning_rate": 4.0100861104102376e-05, "loss": 0.0005, "step": 30612 }, { "epoch": 0.59, "learning_rate": 4.0100214143845145e-05, "loss": 0.001, "step": 30614 }, { "epoch": 0.59, "learning_rate": 4.0099567183587914e-05, "loss": 0.0001, "step": 30616 }, { "epoch": 0.59, "learning_rate": 4.0098920223330683e-05, "loss": 0.0002, "step": 30618 }, { "epoch": 0.59, "learning_rate": 4.009827326307345e-05, "loss": 0.0, "step": 30620 }, { "epoch": 0.59, "learning_rate": 4.009762630281622e-05, "loss": 0.0003, "step": 30622 }, { "epoch": 0.59, "learning_rate": 4.009697934255899e-05, "loss": 0.0, "step": 30624 }, { "epoch": 0.59, "learning_rate": 4.009633238230175e-05, "loss": 0.0088, "step": 30626 }, { "epoch": 0.59, "learning_rate": 4.009568542204453e-05, "loss": 0.0045, "step": 30628 }, { "epoch": 0.59, "learning_rate": 4.009503846178729e-05, "loss": 0.0, "step": 30630 }, { "epoch": 0.59, "learning_rate": 4.009439150153007e-05, "loss": 0.0144, "step": 30632 }, { "epoch": 0.59, "learning_rate": 4.009374454127283e-05, "loss": 0.0, "step": 30634 }, { "epoch": 0.59, "learning_rate": 4.00930975810156e-05, "loss": 0.0208, "step": 30636 }, { "epoch": 0.59, "learning_rate": 4.009245062075837e-05, "loss": 0.0001, "step": 30638 }, { "epoch": 0.59, "learning_rate": 4.0091803660501137e-05, "loss": 0.0001, "step": 30640 }, { "epoch": 0.59, "learning_rate": 4.0091156700243906e-05, "loss": 0.0001, "step": 30642 }, { "epoch": 0.59, "learning_rate": 4.0090509739986675e-05, "loss": 0.0004, "step": 30644 }, { "epoch": 0.59, "learning_rate": 4.0089862779729444e-05, "loss": 0.0003, "step": 30646 }, { "epoch": 0.59, "learning_rate": 4.008921581947221e-05, "loss": 0.0001, "step": 30648 }, { "epoch": 0.59, "learning_rate": 4.008856885921498e-05, "loss": 0.0, "step": 30650 }, { "epoch": 0.59, "learning_rate": 4.0087921898957744e-05, "loss": 0.0004, "step": 30652 }, { "epoch": 0.59, "learning_rate": 4.008727493870052e-05, "loss": 0.0001, "step": 30654 }, { "epoch": 0.59, "learning_rate": 4.008662797844329e-05, "loss": 0.001, "step": 30656 }, { "epoch": 0.6, "learning_rate": 4.008598101818605e-05, "loss": 0.0, "step": 30658 }, { "epoch": 0.6, "learning_rate": 4.008533405792883e-05, "loss": 0.0, "step": 30660 }, { "epoch": 0.6, "learning_rate": 4.008468709767159e-05, "loss": 0.0048, "step": 30662 }, { "epoch": 0.6, "learning_rate": 4.008404013741436e-05, "loss": 0.0003, "step": 30664 }, { "epoch": 0.6, "learning_rate": 4.008339317715713e-05, "loss": 0.0058, "step": 30666 }, { "epoch": 0.6, "learning_rate": 4.00827462168999e-05, "loss": 0.0, "step": 30668 }, { "epoch": 0.6, "learning_rate": 4.0082099256642666e-05, "loss": 0.0043, "step": 30670 }, { "epoch": 0.6, "learning_rate": 4.0081452296385435e-05, "loss": 0.0, "step": 30672 }, { "epoch": 0.6, "learning_rate": 4.0080805336128204e-05, "loss": 0.0015, "step": 30674 }, { "epoch": 0.6, "learning_rate": 4.008015837587097e-05, "loss": 0.0013, "step": 30676 }, { "epoch": 0.6, "learning_rate": 4.007951141561374e-05, "loss": 0.0001, "step": 30678 }, { "epoch": 0.6, "learning_rate": 4.0078864455356505e-05, "loss": 0.0, "step": 30680 }, { "epoch": 0.6, "learning_rate": 4.007821749509928e-05, "loss": 0.002, "step": 30682 }, { "epoch": 0.6, "learning_rate": 4.007757053484204e-05, "loss": 0.0001, "step": 30684 }, { "epoch": 0.6, "learning_rate": 4.007692357458481e-05, "loss": 0.0029, "step": 30686 }, { "epoch": 0.6, "learning_rate": 4.007627661432759e-05, "loss": 0.0001, "step": 30688 }, { "epoch": 0.6, "learning_rate": 4.007562965407035e-05, "loss": 0.0001, "step": 30690 }, { "epoch": 0.6, "learning_rate": 4.0074982693813126e-05, "loss": 0.0003, "step": 30692 }, { "epoch": 0.6, "learning_rate": 4.007433573355589e-05, "loss": 0.0013, "step": 30694 }, { "epoch": 0.6, "learning_rate": 4.007368877329866e-05, "loss": 0.0001, "step": 30696 }, { "epoch": 0.6, "learning_rate": 4.0073041813041427e-05, "loss": 0.0001, "step": 30698 }, { "epoch": 0.6, "learning_rate": 4.0072394852784196e-05, "loss": 0.0, "step": 30700 }, { "epoch": 0.6, "learning_rate": 4.0071747892526965e-05, "loss": 0.0, "step": 30702 }, { "epoch": 0.6, "learning_rate": 4.0071100932269734e-05, "loss": 0.0289, "step": 30704 }, { "epoch": 0.6, "learning_rate": 4.00704539720125e-05, "loss": 0.0043, "step": 30706 }, { "epoch": 0.6, "learning_rate": 4.0069807011755265e-05, "loss": 0.0, "step": 30708 }, { "epoch": 0.6, "learning_rate": 4.006916005149804e-05, "loss": 0.0001, "step": 30710 }, { "epoch": 0.6, "learning_rate": 4.00685130912408e-05, "loss": 0.0133, "step": 30712 }, { "epoch": 0.6, "learning_rate": 4.006786613098358e-05, "loss": 0.0, "step": 30714 }, { "epoch": 0.6, "learning_rate": 4.006721917072634e-05, "loss": 0.0356, "step": 30716 }, { "epoch": 0.6, "learning_rate": 4.006657221046911e-05, "loss": 0.0002, "step": 30718 }, { "epoch": 0.6, "learning_rate": 4.0065925250211886e-05, "loss": 0.0001, "step": 30720 }, { "epoch": 0.6, "learning_rate": 4.006527828995465e-05, "loss": 0.0001, "step": 30722 }, { "epoch": 0.6, "learning_rate": 4.006463132969742e-05, "loss": 0.0118, "step": 30724 }, { "epoch": 0.6, "learning_rate": 4.006398436944019e-05, "loss": 0.0, "step": 30726 }, { "epoch": 0.6, "learning_rate": 4.0063337409182956e-05, "loss": 0.0148, "step": 30728 }, { "epoch": 0.6, "learning_rate": 4.0062690448925725e-05, "loss": 0.0001, "step": 30730 }, { "epoch": 0.6, "learning_rate": 4.0062043488668494e-05, "loss": 0.0, "step": 30732 }, { "epoch": 0.6, "learning_rate": 4.006139652841126e-05, "loss": 0.0001, "step": 30734 }, { "epoch": 0.6, "learning_rate": 4.006074956815403e-05, "loss": 0.0, "step": 30736 }, { "epoch": 0.6, "learning_rate": 4.00601026078968e-05, "loss": 0.0002, "step": 30738 }, { "epoch": 0.6, "learning_rate": 4.0059455647639564e-05, "loss": 0.0008, "step": 30740 }, { "epoch": 0.6, "learning_rate": 4.005880868738234e-05, "loss": 0.0001, "step": 30742 }, { "epoch": 0.6, "learning_rate": 4.00581617271251e-05, "loss": 0.0024, "step": 30744 }, { "epoch": 0.6, "learning_rate": 4.005751476686787e-05, "loss": 0.0001, "step": 30746 }, { "epoch": 0.6, "learning_rate": 4.005686780661064e-05, "loss": 0.0022, "step": 30748 }, { "epoch": 0.6, "learning_rate": 4.005622084635341e-05, "loss": 0.0, "step": 30750 }, { "epoch": 0.6, "learning_rate": 4.005557388609618e-05, "loss": 0.0, "step": 30752 }, { "epoch": 0.6, "learning_rate": 4.005492692583895e-05, "loss": 0.0046, "step": 30754 }, { "epoch": 0.6, "learning_rate": 4.0054279965581716e-05, "loss": 0.0001, "step": 30756 }, { "epoch": 0.6, "learning_rate": 4.0053633005324485e-05, "loss": 0.0, "step": 30758 }, { "epoch": 0.6, "learning_rate": 4.0052986045067255e-05, "loss": 0.0057, "step": 30760 }, { "epoch": 0.6, "learning_rate": 4.005233908481002e-05, "loss": 0.0025, "step": 30762 }, { "epoch": 0.6, "learning_rate": 4.005169212455279e-05, "loss": 0.0008, "step": 30764 }, { "epoch": 0.6, "learning_rate": 4.005104516429556e-05, "loss": 0.0002, "step": 30766 }, { "epoch": 0.6, "learning_rate": 4.0050398204038324e-05, "loss": 0.0002, "step": 30768 }, { "epoch": 0.6, "learning_rate": 4.00497512437811e-05, "loss": 0.0013, "step": 30770 }, { "epoch": 0.6, "learning_rate": 4.004910428352386e-05, "loss": 0.029, "step": 30772 }, { "epoch": 0.6, "learning_rate": 4.004845732326664e-05, "loss": 0.0001, "step": 30774 }, { "epoch": 0.6, "learning_rate": 4.00478103630094e-05, "loss": 0.0, "step": 30776 }, { "epoch": 0.6, "learning_rate": 4.004716340275217e-05, "loss": 0.0, "step": 30778 }, { "epoch": 0.6, "learning_rate": 4.004651644249494e-05, "loss": 0.0003, "step": 30780 }, { "epoch": 0.6, "learning_rate": 4.004586948223771e-05, "loss": 0.0043, "step": 30782 }, { "epoch": 0.6, "learning_rate": 4.004522252198048e-05, "loss": 0.0048, "step": 30784 }, { "epoch": 0.6, "learning_rate": 4.0044575561723246e-05, "loss": 0.0002, "step": 30786 }, { "epoch": 0.6, "learning_rate": 4.0043928601466015e-05, "loss": 0.0, "step": 30788 }, { "epoch": 0.6, "learning_rate": 4.0043281641208784e-05, "loss": 0.0004, "step": 30790 }, { "epoch": 0.6, "learning_rate": 4.004263468095155e-05, "loss": 0.0111, "step": 30792 }, { "epoch": 0.6, "learning_rate": 4.0041987720694315e-05, "loss": 0.0, "step": 30794 }, { "epoch": 0.6, "learning_rate": 4.004134076043709e-05, "loss": 0.0001, "step": 30796 }, { "epoch": 0.6, "learning_rate": 4.004069380017986e-05, "loss": 0.0001, "step": 30798 }, { "epoch": 0.6, "learning_rate": 4.004004683992262e-05, "loss": 0.0001, "step": 30800 }, { "epoch": 0.6, "learning_rate": 4.00393998796654e-05, "loss": 0.0001, "step": 30802 }, { "epoch": 0.6, "learning_rate": 4.003875291940816e-05, "loss": 0.0147, "step": 30804 }, { "epoch": 0.6, "learning_rate": 4.003810595915093e-05, "loss": 0.0002, "step": 30806 }, { "epoch": 0.6, "learning_rate": 4.00374589988937e-05, "loss": 0.0001, "step": 30808 }, { "epoch": 0.6, "learning_rate": 4.003681203863647e-05, "loss": 0.0001, "step": 30810 }, { "epoch": 0.6, "learning_rate": 4.003616507837924e-05, "loss": 0.0004, "step": 30812 }, { "epoch": 0.6, "learning_rate": 4.0035518118122006e-05, "loss": 0.0165, "step": 30814 }, { "epoch": 0.6, "learning_rate": 4.0034871157864775e-05, "loss": 0.0001, "step": 30816 }, { "epoch": 0.6, "learning_rate": 4.0034224197607544e-05, "loss": 0.0, "step": 30818 }, { "epoch": 0.6, "learning_rate": 4.0033577237350314e-05, "loss": 0.0, "step": 30820 }, { "epoch": 0.6, "learning_rate": 4.0032930277093076e-05, "loss": 0.0009, "step": 30822 }, { "epoch": 0.6, "learning_rate": 4.003228331683585e-05, "loss": 0.0054, "step": 30824 }, { "epoch": 0.6, "learning_rate": 4.0031636356578614e-05, "loss": 0.0003, "step": 30826 }, { "epoch": 0.6, "learning_rate": 4.003098939632138e-05, "loss": 0.0003, "step": 30828 }, { "epoch": 0.6, "learning_rate": 4.003034243606415e-05, "loss": 0.0144, "step": 30830 }, { "epoch": 0.6, "learning_rate": 4.002969547580692e-05, "loss": 0.0003, "step": 30832 }, { "epoch": 0.6, "learning_rate": 4.00290485155497e-05, "loss": 0.0046, "step": 30834 }, { "epoch": 0.6, "learning_rate": 4.002840155529246e-05, "loss": 0.0, "step": 30836 }, { "epoch": 0.6, "learning_rate": 4.002775459503523e-05, "loss": 0.0001, "step": 30838 }, { "epoch": 0.6, "learning_rate": 4.0027107634778e-05, "loss": 0.0, "step": 30840 }, { "epoch": 0.6, "learning_rate": 4.002646067452077e-05, "loss": 0.0014, "step": 30842 }, { "epoch": 0.6, "learning_rate": 4.0025813714263536e-05, "loss": 0.0003, "step": 30844 }, { "epoch": 0.6, "learning_rate": 4.0025166754006305e-05, "loss": 0.0002, "step": 30846 }, { "epoch": 0.6, "learning_rate": 4.0024519793749074e-05, "loss": 0.0027, "step": 30848 }, { "epoch": 0.6, "learning_rate": 4.0023872833491836e-05, "loss": 0.0001, "step": 30850 }, { "epoch": 0.6, "learning_rate": 4.002322587323461e-05, "loss": 0.0021, "step": 30852 }, { "epoch": 0.6, "learning_rate": 4.0022578912977374e-05, "loss": 0.0, "step": 30854 }, { "epoch": 0.6, "learning_rate": 4.002193195272015e-05, "loss": 0.0002, "step": 30856 }, { "epoch": 0.6, "learning_rate": 4.002128499246291e-05, "loss": 0.0001, "step": 30858 }, { "epoch": 0.6, "learning_rate": 4.002063803220568e-05, "loss": 0.0001, "step": 30860 }, { "epoch": 0.6, "learning_rate": 4.001999107194845e-05, "loss": 0.0021, "step": 30862 }, { "epoch": 0.6, "learning_rate": 4.001934411169122e-05, "loss": 0.0, "step": 30864 }, { "epoch": 0.6, "learning_rate": 4.001869715143399e-05, "loss": 0.0395, "step": 30866 }, { "epoch": 0.6, "learning_rate": 4.001805019117676e-05, "loss": 0.0013, "step": 30868 }, { "epoch": 0.6, "learning_rate": 4.001740323091953e-05, "loss": 0.0022, "step": 30870 }, { "epoch": 0.6, "learning_rate": 4.0016756270662296e-05, "loss": 0.0171, "step": 30872 }, { "epoch": 0.6, "learning_rate": 4.0016109310405065e-05, "loss": 0.0, "step": 30874 }, { "epoch": 0.6, "learning_rate": 4.001546235014783e-05, "loss": 0.0007, "step": 30876 }, { "epoch": 0.6, "learning_rate": 4.0014815389890603e-05, "loss": 0.0001, "step": 30878 }, { "epoch": 0.6, "learning_rate": 4.001416842963337e-05, "loss": 0.0001, "step": 30880 }, { "epoch": 0.6, "learning_rate": 4.0013521469376135e-05, "loss": 0.0, "step": 30882 }, { "epoch": 0.6, "learning_rate": 4.001287450911891e-05, "loss": 0.0006, "step": 30884 }, { "epoch": 0.6, "learning_rate": 4.001222754886167e-05, "loss": 0.0, "step": 30886 }, { "epoch": 0.6, "learning_rate": 4.001158058860444e-05, "loss": 0.0003, "step": 30888 }, { "epoch": 0.6, "learning_rate": 4.001093362834721e-05, "loss": 0.0003, "step": 30890 }, { "epoch": 0.6, "learning_rate": 4.001028666808998e-05, "loss": 0.0001, "step": 30892 }, { "epoch": 0.6, "learning_rate": 4.000963970783275e-05, "loss": 0.0016, "step": 30894 }, { "epoch": 0.6, "learning_rate": 4.000899274757552e-05, "loss": 0.004, "step": 30896 }, { "epoch": 0.6, "learning_rate": 4.000834578731829e-05, "loss": 0.0075, "step": 30898 }, { "epoch": 0.6, "learning_rate": 4.0007698827061057e-05, "loss": 0.0001, "step": 30900 }, { "epoch": 0.6, "learning_rate": 4.0007051866803826e-05, "loss": 0.0014, "step": 30902 }, { "epoch": 0.6, "learning_rate": 4.000640490654659e-05, "loss": 0.0, "step": 30904 }, { "epoch": 0.6, "learning_rate": 4.0005757946289364e-05, "loss": 0.0047, "step": 30906 }, { "epoch": 0.6, "learning_rate": 4.0005110986032126e-05, "loss": 0.0, "step": 30908 }, { "epoch": 0.6, "learning_rate": 4.0004464025774895e-05, "loss": 0.0, "step": 30910 }, { "epoch": 0.6, "learning_rate": 4.000381706551767e-05, "loss": 0.0002, "step": 30912 }, { "epoch": 0.6, "learning_rate": 4.0003170105260433e-05, "loss": 0.0091, "step": 30914 }, { "epoch": 0.6, "learning_rate": 4.000252314500321e-05, "loss": 0.0002, "step": 30916 }, { "epoch": 0.6, "learning_rate": 4.000187618474597e-05, "loss": 0.0, "step": 30918 }, { "epoch": 0.6, "learning_rate": 4.000122922448874e-05, "loss": 0.0, "step": 30920 }, { "epoch": 0.6, "learning_rate": 4.000058226423151e-05, "loss": 0.0001, "step": 30922 }, { "epoch": 0.6, "learning_rate": 3.999993530397428e-05, "loss": 0.0005, "step": 30924 }, { "epoch": 0.6, "learning_rate": 3.999928834371705e-05, "loss": 0.0, "step": 30926 }, { "epoch": 0.6, "learning_rate": 3.999864138345982e-05, "loss": 0.0224, "step": 30928 }, { "epoch": 0.6, "learning_rate": 3.9997994423202586e-05, "loss": 0.0001, "step": 30930 }, { "epoch": 0.6, "learning_rate": 3.9997347462945355e-05, "loss": 0.0014, "step": 30932 }, { "epoch": 0.6, "learning_rate": 3.9996700502688124e-05, "loss": 0.0032, "step": 30934 }, { "epoch": 0.6, "learning_rate": 3.9996053542430887e-05, "loss": 0.0001, "step": 30936 }, { "epoch": 0.6, "learning_rate": 3.999540658217366e-05, "loss": 0.004, "step": 30938 }, { "epoch": 0.6, "learning_rate": 3.9994759621916425e-05, "loss": 0.0001, "step": 30940 }, { "epoch": 0.6, "learning_rate": 3.9994112661659194e-05, "loss": 0.0, "step": 30942 }, { "epoch": 0.6, "learning_rate": 3.999346570140197e-05, "loss": 0.0, "step": 30944 }, { "epoch": 0.6, "learning_rate": 3.999281874114473e-05, "loss": 0.0001, "step": 30946 }, { "epoch": 0.6, "learning_rate": 3.99921717808875e-05, "loss": 0.098, "step": 30948 }, { "epoch": 0.6, "learning_rate": 3.999152482063027e-05, "loss": 0.0005, "step": 30950 }, { "epoch": 0.6, "learning_rate": 3.999087786037304e-05, "loss": 0.0, "step": 30952 }, { "epoch": 0.6, "learning_rate": 3.999023090011581e-05, "loss": 0.0003, "step": 30954 }, { "epoch": 0.6, "learning_rate": 3.998958393985858e-05, "loss": 0.0, "step": 30956 }, { "epoch": 0.6, "learning_rate": 3.9988936979601346e-05, "loss": 0.0, "step": 30958 }, { "epoch": 0.6, "learning_rate": 3.9988290019344116e-05, "loss": 0.0037, "step": 30960 }, { "epoch": 0.6, "learning_rate": 3.9987643059086885e-05, "loss": 0.0001, "step": 30962 }, { "epoch": 0.6, "learning_rate": 3.998699609882965e-05, "loss": 0.0188, "step": 30964 }, { "epoch": 0.6, "learning_rate": 3.998634913857242e-05, "loss": 0.0006, "step": 30966 }, { "epoch": 0.6, "learning_rate": 3.9985702178315185e-05, "loss": 0.0001, "step": 30968 }, { "epoch": 0.6, "learning_rate": 3.9985055218057954e-05, "loss": 0.0002, "step": 30970 }, { "epoch": 0.6, "learning_rate": 3.998440825780072e-05, "loss": 0.0001, "step": 30972 }, { "epoch": 0.6, "learning_rate": 3.998376129754349e-05, "loss": 0.0, "step": 30974 }, { "epoch": 0.6, "learning_rate": 3.998311433728627e-05, "loss": 0.0001, "step": 30976 }, { "epoch": 0.6, "learning_rate": 3.998246737702903e-05, "loss": 0.0046, "step": 30978 }, { "epoch": 0.6, "learning_rate": 3.99818204167718e-05, "loss": 0.0003, "step": 30980 }, { "epoch": 0.6, "learning_rate": 3.998117345651457e-05, "loss": 0.0001, "step": 30982 }, { "epoch": 0.6, "learning_rate": 3.998052649625734e-05, "loss": 0.0, "step": 30984 }, { "epoch": 0.6, "learning_rate": 3.99798795360001e-05, "loss": 0.0032, "step": 30986 }, { "epoch": 0.6, "learning_rate": 3.9979232575742876e-05, "loss": 0.008, "step": 30988 }, { "epoch": 0.6, "learning_rate": 3.9978585615485645e-05, "loss": 0.0001, "step": 30990 }, { "epoch": 0.6, "learning_rate": 3.997793865522841e-05, "loss": 0.0019, "step": 30992 }, { "epoch": 0.6, "learning_rate": 3.997729169497118e-05, "loss": 0.0009, "step": 30994 }, { "epoch": 0.6, "learning_rate": 3.9976644734713946e-05, "loss": 0.0006, "step": 30996 }, { "epoch": 0.6, "learning_rate": 3.997599777445672e-05, "loss": 0.0001, "step": 30998 }, { "epoch": 0.6, "learning_rate": 3.9975350814199484e-05, "loss": 0.0, "step": 31000 }, { "epoch": 0.6, "learning_rate": 3.997470385394225e-05, "loss": 0.0, "step": 31002 }, { "epoch": 0.6, "learning_rate": 3.997405689368502e-05, "loss": 0.005, "step": 31004 }, { "epoch": 0.6, "learning_rate": 3.997340993342779e-05, "loss": 0.0, "step": 31006 }, { "epoch": 0.6, "learning_rate": 3.997276297317056e-05, "loss": 0.0001, "step": 31008 }, { "epoch": 0.6, "learning_rate": 3.997211601291333e-05, "loss": 0.0004, "step": 31010 }, { "epoch": 0.6, "learning_rate": 3.99714690526561e-05, "loss": 0.0002, "step": 31012 }, { "epoch": 0.6, "learning_rate": 3.997082209239887e-05, "loss": 0.0001, "step": 31014 }, { "epoch": 0.6, "learning_rate": 3.9970175132141636e-05, "loss": 0.0, "step": 31016 }, { "epoch": 0.6, "learning_rate": 3.99695281718844e-05, "loss": 0.0002, "step": 31018 }, { "epoch": 0.6, "learning_rate": 3.9968881211627175e-05, "loss": 0.0019, "step": 31020 }, { "epoch": 0.6, "learning_rate": 3.9968234251369944e-05, "loss": 0.0001, "step": 31022 }, { "epoch": 0.6, "learning_rate": 3.9967587291112706e-05, "loss": 0.0, "step": 31024 }, { "epoch": 0.6, "learning_rate": 3.996694033085548e-05, "loss": 0.0001, "step": 31026 }, { "epoch": 0.6, "learning_rate": 3.9966293370598244e-05, "loss": 0.0008, "step": 31028 }, { "epoch": 0.6, "learning_rate": 3.996564641034101e-05, "loss": 0.0056, "step": 31030 }, { "epoch": 0.6, "learning_rate": 3.996499945008378e-05, "loss": 0.0056, "step": 31032 }, { "epoch": 0.6, "learning_rate": 3.996435248982655e-05, "loss": 0.0, "step": 31034 }, { "epoch": 0.6, "learning_rate": 3.996370552956932e-05, "loss": 0.0007, "step": 31036 }, { "epoch": 0.6, "learning_rate": 3.996305856931209e-05, "loss": 0.0014, "step": 31038 }, { "epoch": 0.6, "learning_rate": 3.996241160905486e-05, "loss": 0.004, "step": 31040 }, { "epoch": 0.6, "learning_rate": 3.996176464879763e-05, "loss": 0.0001, "step": 31042 }, { "epoch": 0.6, "learning_rate": 3.99611176885404e-05, "loss": 0.0012, "step": 31044 }, { "epoch": 0.6, "learning_rate": 3.996047072828316e-05, "loss": 0.0, "step": 31046 }, { "epoch": 0.6, "learning_rate": 3.9959823768025935e-05, "loss": 0.0, "step": 31048 }, { "epoch": 0.6, "learning_rate": 3.99591768077687e-05, "loss": 0.0002, "step": 31050 }, { "epoch": 0.6, "learning_rate": 3.9958529847511466e-05, "loss": 0.001, "step": 31052 }, { "epoch": 0.6, "learning_rate": 3.9957882887254235e-05, "loss": 0.0007, "step": 31054 }, { "epoch": 0.6, "learning_rate": 3.9957235926997005e-05, "loss": 0.0017, "step": 31056 }, { "epoch": 0.6, "learning_rate": 3.995658896673978e-05, "loss": 0.0001, "step": 31058 }, { "epoch": 0.6, "learning_rate": 3.995594200648254e-05, "loss": 0.0098, "step": 31060 }, { "epoch": 0.6, "learning_rate": 3.995529504622531e-05, "loss": 0.0, "step": 31062 }, { "epoch": 0.6, "learning_rate": 3.995464808596808e-05, "loss": 0.0002, "step": 31064 }, { "epoch": 0.6, "learning_rate": 3.995400112571085e-05, "loss": 0.01, "step": 31066 }, { "epoch": 0.6, "learning_rate": 3.995335416545362e-05, "loss": 0.0, "step": 31068 }, { "epoch": 0.6, "learning_rate": 3.995270720519639e-05, "loss": 0.0, "step": 31070 }, { "epoch": 0.6, "learning_rate": 3.995206024493916e-05, "loss": 0.0326, "step": 31072 }, { "epoch": 0.6, "learning_rate": 3.9951413284681926e-05, "loss": 0.001, "step": 31074 }, { "epoch": 0.6, "learning_rate": 3.9950766324424695e-05, "loss": 0.0037, "step": 31076 }, { "epoch": 0.6, "learning_rate": 3.995011936416746e-05, "loss": 0.0001, "step": 31078 }, { "epoch": 0.6, "learning_rate": 3.9949472403910234e-05, "loss": 0.001, "step": 31080 }, { "epoch": 0.6, "learning_rate": 3.9948825443652996e-05, "loss": 0.0141, "step": 31082 }, { "epoch": 0.6, "learning_rate": 3.9948178483395765e-05, "loss": 0.0077, "step": 31084 }, { "epoch": 0.6, "learning_rate": 3.9947531523138534e-05, "loss": 0.0001, "step": 31086 }, { "epoch": 0.6, "learning_rate": 3.99468845628813e-05, "loss": 0.002, "step": 31088 }, { "epoch": 0.6, "learning_rate": 3.994623760262407e-05, "loss": 0.0, "step": 31090 }, { "epoch": 0.6, "learning_rate": 3.994559064236684e-05, "loss": 0.0003, "step": 31092 }, { "epoch": 0.6, "learning_rate": 3.994494368210961e-05, "loss": 0.0021, "step": 31094 }, { "epoch": 0.6, "learning_rate": 3.994429672185238e-05, "loss": 0.0001, "step": 31096 }, { "epoch": 0.6, "learning_rate": 3.994364976159515e-05, "loss": 0.0002, "step": 31098 }, { "epoch": 0.6, "learning_rate": 3.994300280133792e-05, "loss": 0.0001, "step": 31100 }, { "epoch": 0.6, "learning_rate": 3.994235584108069e-05, "loss": 0.0, "step": 31102 }, { "epoch": 0.6, "learning_rate": 3.9941708880823456e-05, "loss": 0.0043, "step": 31104 }, { "epoch": 0.6, "learning_rate": 3.994106192056622e-05, "loss": 0.0011, "step": 31106 }, { "epoch": 0.6, "learning_rate": 3.9940414960308994e-05, "loss": 0.0332, "step": 31108 }, { "epoch": 0.6, "learning_rate": 3.9939768000051756e-05, "loss": 0.0004, "step": 31110 }, { "epoch": 0.6, "learning_rate": 3.9939121039794525e-05, "loss": 0.0049, "step": 31112 }, { "epoch": 0.6, "learning_rate": 3.9938474079537294e-05, "loss": 0.0001, "step": 31114 }, { "epoch": 0.6, "learning_rate": 3.9937827119280063e-05, "loss": 0.006, "step": 31116 }, { "epoch": 0.6, "learning_rate": 3.993718015902283e-05, "loss": 0.0, "step": 31118 }, { "epoch": 0.6, "learning_rate": 3.99365331987656e-05, "loss": 0.0012, "step": 31120 }, { "epoch": 0.6, "learning_rate": 3.993588623850837e-05, "loss": 0.0, "step": 31122 }, { "epoch": 0.6, "learning_rate": 3.993523927825114e-05, "loss": 0.0001, "step": 31124 }, { "epoch": 0.6, "learning_rate": 3.993459231799391e-05, "loss": 0.007, "step": 31126 }, { "epoch": 0.6, "learning_rate": 3.993394535773667e-05, "loss": 0.0, "step": 31128 }, { "epoch": 0.6, "learning_rate": 3.993329839747945e-05, "loss": 0.0003, "step": 31130 }, { "epoch": 0.6, "learning_rate": 3.993265143722221e-05, "loss": 0.0014, "step": 31132 }, { "epoch": 0.6, "learning_rate": 3.993200447696498e-05, "loss": 0.0001, "step": 31134 }, { "epoch": 0.6, "learning_rate": 3.9931357516707754e-05, "loss": 0.0001, "step": 31136 }, { "epoch": 0.6, "learning_rate": 3.993071055645052e-05, "loss": 0.0001, "step": 31138 }, { "epoch": 0.6, "learning_rate": 3.993006359619329e-05, "loss": 0.0004, "step": 31140 }, { "epoch": 0.6, "learning_rate": 3.9929416635936055e-05, "loss": 0.0, "step": 31142 }, { "epoch": 0.6, "learning_rate": 3.9928769675678824e-05, "loss": 0.0001, "step": 31144 }, { "epoch": 0.6, "learning_rate": 3.992812271542159e-05, "loss": 0.0001, "step": 31146 }, { "epoch": 0.6, "learning_rate": 3.992747575516436e-05, "loss": 0.0007, "step": 31148 }, { "epoch": 0.6, "learning_rate": 3.992682879490713e-05, "loss": 0.0004, "step": 31150 }, { "epoch": 0.6, "learning_rate": 3.99261818346499e-05, "loss": 0.0004, "step": 31152 }, { "epoch": 0.6, "learning_rate": 3.992553487439267e-05, "loss": 0.0141, "step": 31154 }, { "epoch": 0.6, "learning_rate": 3.992488791413544e-05, "loss": 0.0, "step": 31156 }, { "epoch": 0.6, "learning_rate": 3.992424095387821e-05, "loss": 0.0, "step": 31158 }, { "epoch": 0.6, "learning_rate": 3.992359399362097e-05, "loss": 0.0, "step": 31160 }, { "epoch": 0.6, "learning_rate": 3.9922947033363746e-05, "loss": 0.0001, "step": 31162 }, { "epoch": 0.6, "learning_rate": 3.992230007310651e-05, "loss": 0.0001, "step": 31164 }, { "epoch": 0.6, "learning_rate": 3.992165311284928e-05, "loss": 0.0002, "step": 31166 }, { "epoch": 0.6, "learning_rate": 3.992100615259205e-05, "loss": 0.0, "step": 31168 }, { "epoch": 0.6, "learning_rate": 3.9920359192334815e-05, "loss": 0.0, "step": 31170 }, { "epoch": 0.61, "learning_rate": 3.9919712232077584e-05, "loss": 0.0002, "step": 31172 }, { "epoch": 0.61, "learning_rate": 3.991906527182035e-05, "loss": 0.0, "step": 31174 }, { "epoch": 0.61, "learning_rate": 3.991841831156312e-05, "loss": 0.0006, "step": 31176 }, { "epoch": 0.61, "learning_rate": 3.991777135130589e-05, "loss": 0.0001, "step": 31178 }, { "epoch": 0.61, "learning_rate": 3.991712439104866e-05, "loss": 0.0017, "step": 31180 }, { "epoch": 0.61, "learning_rate": 3.991647743079143e-05, "loss": 0.0001, "step": 31182 }, { "epoch": 0.61, "learning_rate": 3.99158304705342e-05, "loss": 0.0, "step": 31184 }, { "epoch": 0.61, "learning_rate": 3.991518351027697e-05, "loss": 0.0, "step": 31186 }, { "epoch": 0.61, "learning_rate": 3.991453655001973e-05, "loss": 0.0004, "step": 31188 }, { "epoch": 0.61, "learning_rate": 3.9913889589762506e-05, "loss": 0.0034, "step": 31190 }, { "epoch": 0.61, "learning_rate": 3.991324262950527e-05, "loss": 0.0001, "step": 31192 }, { "epoch": 0.61, "learning_rate": 3.991259566924804e-05, "loss": 0.002, "step": 31194 }, { "epoch": 0.61, "learning_rate": 3.9911948708990807e-05, "loss": 0.0013, "step": 31196 }, { "epoch": 0.61, "learning_rate": 3.9911301748733576e-05, "loss": 0.0042, "step": 31198 }, { "epoch": 0.61, "learning_rate": 3.991065478847635e-05, "loss": 0.0, "step": 31200 }, { "epoch": 0.61, "learning_rate": 3.9910007828219114e-05, "loss": 0.0001, "step": 31202 }, { "epoch": 0.61, "learning_rate": 3.990936086796188e-05, "loss": 0.0001, "step": 31204 }, { "epoch": 0.61, "learning_rate": 3.990871390770465e-05, "loss": 0.0001, "step": 31206 }, { "epoch": 0.61, "learning_rate": 3.990806694744742e-05, "loss": 0.0043, "step": 31208 }, { "epoch": 0.61, "learning_rate": 3.990741998719018e-05, "loss": 0.0, "step": 31210 }, { "epoch": 0.61, "learning_rate": 3.990677302693296e-05, "loss": 0.0026, "step": 31212 }, { "epoch": 0.61, "learning_rate": 3.990612606667573e-05, "loss": 0.0001, "step": 31214 }, { "epoch": 0.61, "learning_rate": 3.990547910641849e-05, "loss": 0.0003, "step": 31216 }, { "epoch": 0.61, "learning_rate": 3.9904832146161266e-05, "loss": 0.0065, "step": 31218 }, { "epoch": 0.61, "learning_rate": 3.990418518590403e-05, "loss": 0.0, "step": 31220 }, { "epoch": 0.61, "learning_rate": 3.9903538225646805e-05, "loss": 0.0, "step": 31222 }, { "epoch": 0.61, "learning_rate": 3.990289126538957e-05, "loss": 0.0002, "step": 31224 }, { "epoch": 0.61, "learning_rate": 3.9902244305132336e-05, "loss": 0.0, "step": 31226 }, { "epoch": 0.61, "learning_rate": 3.9901597344875105e-05, "loss": 0.0004, "step": 31228 }, { "epoch": 0.61, "learning_rate": 3.9900950384617874e-05, "loss": 0.0, "step": 31230 }, { "epoch": 0.61, "learning_rate": 3.990030342436064e-05, "loss": 0.0001, "step": 31232 }, { "epoch": 0.61, "learning_rate": 3.989965646410341e-05, "loss": 0.0, "step": 31234 }, { "epoch": 0.61, "learning_rate": 3.989900950384618e-05, "loss": 0.0026, "step": 31236 }, { "epoch": 0.61, "learning_rate": 3.989836254358895e-05, "loss": 0.0, "step": 31238 }, { "epoch": 0.61, "learning_rate": 3.989771558333172e-05, "loss": 0.0003, "step": 31240 }, { "epoch": 0.61, "learning_rate": 3.989706862307448e-05, "loss": 0.0098, "step": 31242 }, { "epoch": 0.61, "learning_rate": 3.989642166281726e-05, "loss": 0.0, "step": 31244 }, { "epoch": 0.61, "learning_rate": 3.989577470256003e-05, "loss": 0.0, "step": 31246 }, { "epoch": 0.61, "learning_rate": 3.989512774230279e-05, "loss": 0.0078, "step": 31248 }, { "epoch": 0.61, "learning_rate": 3.9894480782045565e-05, "loss": 0.0002, "step": 31250 }, { "epoch": 0.61, "learning_rate": 3.989383382178833e-05, "loss": 0.0, "step": 31252 }, { "epoch": 0.61, "learning_rate": 3.9893186861531096e-05, "loss": 0.0024, "step": 31254 }, { "epoch": 0.61, "learning_rate": 3.9892539901273866e-05, "loss": 0.0002, "step": 31256 }, { "epoch": 0.61, "learning_rate": 3.9891892941016635e-05, "loss": 0.0001, "step": 31258 }, { "epoch": 0.61, "learning_rate": 3.9891245980759404e-05, "loss": 0.0, "step": 31260 }, { "epoch": 0.61, "learning_rate": 3.989059902050217e-05, "loss": 0.0019, "step": 31262 }, { "epoch": 0.61, "learning_rate": 3.988995206024494e-05, "loss": 0.0174, "step": 31264 }, { "epoch": 0.61, "learning_rate": 3.988930509998771e-05, "loss": 0.0053, "step": 31266 }, { "epoch": 0.61, "learning_rate": 3.988865813973048e-05, "loss": 0.0006, "step": 31268 }, { "epoch": 0.61, "learning_rate": 3.988801117947324e-05, "loss": 0.0001, "step": 31270 }, { "epoch": 0.61, "learning_rate": 3.988736421921602e-05, "loss": 0.0095, "step": 31272 }, { "epoch": 0.61, "learning_rate": 3.988671725895878e-05, "loss": 0.0, "step": 31274 }, { "epoch": 0.61, "learning_rate": 3.988607029870155e-05, "loss": 0.0059, "step": 31276 }, { "epoch": 0.61, "learning_rate": 3.9885423338444325e-05, "loss": 0.0008, "step": 31278 }, { "epoch": 0.61, "learning_rate": 3.988477637818709e-05, "loss": 0.0002, "step": 31280 }, { "epoch": 0.61, "learning_rate": 3.9884129417929864e-05, "loss": 0.0002, "step": 31282 }, { "epoch": 0.61, "learning_rate": 3.9883482457672626e-05, "loss": 0.0011, "step": 31284 }, { "epoch": 0.61, "learning_rate": 3.9882835497415395e-05, "loss": 0.0002, "step": 31286 }, { "epoch": 0.61, "learning_rate": 3.9882188537158164e-05, "loss": 0.0008, "step": 31288 }, { "epoch": 0.61, "learning_rate": 3.988154157690093e-05, "loss": 0.0003, "step": 31290 }, { "epoch": 0.61, "learning_rate": 3.98808946166437e-05, "loss": 0.0117, "step": 31292 }, { "epoch": 0.61, "learning_rate": 3.988024765638647e-05, "loss": 0.0002, "step": 31294 }, { "epoch": 0.61, "learning_rate": 3.987960069612924e-05, "loss": 0.0001, "step": 31296 }, { "epoch": 0.61, "learning_rate": 3.987895373587201e-05, "loss": 0.0009, "step": 31298 }, { "epoch": 0.61, "learning_rate": 3.987830677561478e-05, "loss": 0.0, "step": 31300 }, { "epoch": 0.61, "learning_rate": 3.987765981535754e-05, "loss": 0.0, "step": 31302 }, { "epoch": 0.61, "learning_rate": 3.987701285510032e-05, "loss": 0.0001, "step": 31304 }, { "epoch": 0.61, "learning_rate": 3.987636589484308e-05, "loss": 0.0002, "step": 31306 }, { "epoch": 0.61, "learning_rate": 3.987571893458585e-05, "loss": 0.0, "step": 31308 }, { "epoch": 0.61, "learning_rate": 3.987507197432862e-05, "loss": 0.0024, "step": 31310 }, { "epoch": 0.61, "learning_rate": 3.9874425014071386e-05, "loss": 0.0, "step": 31312 }, { "epoch": 0.61, "learning_rate": 3.9873778053814155e-05, "loss": 0.0, "step": 31314 }, { "epoch": 0.61, "learning_rate": 3.9873131093556924e-05, "loss": 0.0064, "step": 31316 }, { "epoch": 0.61, "learning_rate": 3.9872484133299694e-05, "loss": 0.0048, "step": 31318 }, { "epoch": 0.61, "learning_rate": 3.987183717304246e-05, "loss": 0.0078, "step": 31320 }, { "epoch": 0.61, "learning_rate": 3.987119021278523e-05, "loss": 0.0, "step": 31322 }, { "epoch": 0.61, "learning_rate": 3.9870543252528e-05, "loss": 0.0082, "step": 31324 }, { "epoch": 0.61, "learning_rate": 3.986989629227077e-05, "loss": 0.0001, "step": 31326 }, { "epoch": 0.61, "learning_rate": 3.986924933201354e-05, "loss": 0.0028, "step": 31328 }, { "epoch": 0.61, "learning_rate": 3.98686023717563e-05, "loss": 0.0002, "step": 31330 }, { "epoch": 0.61, "learning_rate": 3.986795541149908e-05, "loss": 0.0001, "step": 31332 }, { "epoch": 0.61, "learning_rate": 3.986730845124184e-05, "loss": 0.0, "step": 31334 }, { "epoch": 0.61, "learning_rate": 3.986666149098461e-05, "loss": 0.0001, "step": 31336 }, { "epoch": 0.61, "learning_rate": 3.986601453072738e-05, "loss": 0.0002, "step": 31338 }, { "epoch": 0.61, "learning_rate": 3.986536757047015e-05, "loss": 0.0109, "step": 31340 }, { "epoch": 0.61, "learning_rate": 3.9864720610212916e-05, "loss": 0.0046, "step": 31342 }, { "epoch": 0.61, "learning_rate": 3.9864073649955685e-05, "loss": 0.0002, "step": 31344 }, { "epoch": 0.61, "learning_rate": 3.9863426689698454e-05, "loss": 0.0, "step": 31346 }, { "epoch": 0.61, "learning_rate": 3.986277972944122e-05, "loss": 0.0011, "step": 31348 }, { "epoch": 0.61, "learning_rate": 3.986213276918399e-05, "loss": 0.0027, "step": 31350 }, { "epoch": 0.61, "learning_rate": 3.9861485808926754e-05, "loss": 0.0001, "step": 31352 }, { "epoch": 0.61, "learning_rate": 3.986083884866953e-05, "loss": 0.0, "step": 31354 }, { "epoch": 0.61, "learning_rate": 3.986019188841229e-05, "loss": 0.0, "step": 31356 }, { "epoch": 0.61, "learning_rate": 3.985954492815506e-05, "loss": 0.0, "step": 31358 }, { "epoch": 0.61, "learning_rate": 3.985889796789784e-05, "loss": 0.0, "step": 31360 }, { "epoch": 0.61, "learning_rate": 3.98582510076406e-05, "loss": 0.0001, "step": 31362 }, { "epoch": 0.61, "learning_rate": 3.9857604047383376e-05, "loss": 0.0238, "step": 31364 }, { "epoch": 0.61, "learning_rate": 3.985695708712614e-05, "loss": 0.0016, "step": 31366 }, { "epoch": 0.61, "learning_rate": 3.985631012686891e-05, "loss": 0.0001, "step": 31368 }, { "epoch": 0.61, "learning_rate": 3.9855663166611676e-05, "loss": 0.0, "step": 31370 }, { "epoch": 0.61, "learning_rate": 3.9855016206354445e-05, "loss": 0.0008, "step": 31372 }, { "epoch": 0.61, "learning_rate": 3.9854369246097214e-05, "loss": 0.0, "step": 31374 }, { "epoch": 0.61, "learning_rate": 3.9853722285839983e-05, "loss": 0.0071, "step": 31376 }, { "epoch": 0.61, "learning_rate": 3.985307532558275e-05, "loss": 0.0001, "step": 31378 }, { "epoch": 0.61, "learning_rate": 3.985242836532552e-05, "loss": 0.0001, "step": 31380 }, { "epoch": 0.61, "learning_rate": 3.985178140506829e-05, "loss": 0.0343, "step": 31382 }, { "epoch": 0.61, "learning_rate": 3.985113444481105e-05, "loss": 0.0064, "step": 31384 }, { "epoch": 0.61, "learning_rate": 3.985048748455383e-05, "loss": 0.0002, "step": 31386 }, { "epoch": 0.61, "learning_rate": 3.984984052429659e-05, "loss": 0.0015, "step": 31388 }, { "epoch": 0.61, "learning_rate": 3.984919356403936e-05, "loss": 0.009, "step": 31390 }, { "epoch": 0.61, "learning_rate": 3.9848546603782136e-05, "loss": 0.0031, "step": 31392 }, { "epoch": 0.61, "learning_rate": 3.98478996435249e-05, "loss": 0.0002, "step": 31394 }, { "epoch": 0.61, "learning_rate": 3.984725268326767e-05, "loss": 0.0, "step": 31396 }, { "epoch": 0.61, "learning_rate": 3.984660572301044e-05, "loss": 0.0, "step": 31398 }, { "epoch": 0.61, "learning_rate": 3.9845958762753206e-05, "loss": 0.0, "step": 31400 }, { "epoch": 0.61, "learning_rate": 3.9845311802495975e-05, "loss": 0.0003, "step": 31402 }, { "epoch": 0.61, "learning_rate": 3.9844664842238744e-05, "loss": 0.0001, "step": 31404 }, { "epoch": 0.61, "learning_rate": 3.984401788198151e-05, "loss": 0.0, "step": 31406 }, { "epoch": 0.61, "learning_rate": 3.984337092172428e-05, "loss": 0.0201, "step": 31408 }, { "epoch": 0.61, "learning_rate": 3.984272396146705e-05, "loss": 0.0078, "step": 31410 }, { "epoch": 0.61, "learning_rate": 3.9842077001209813e-05, "loss": 0.0002, "step": 31412 }, { "epoch": 0.61, "learning_rate": 3.984143004095259e-05, "loss": 0.0011, "step": 31414 }, { "epoch": 0.61, "learning_rate": 3.984078308069535e-05, "loss": 0.0006, "step": 31416 }, { "epoch": 0.61, "learning_rate": 3.984013612043812e-05, "loss": 0.0039, "step": 31418 }, { "epoch": 0.61, "learning_rate": 3.983948916018089e-05, "loss": 0.0002, "step": 31420 }, { "epoch": 0.61, "learning_rate": 3.983884219992366e-05, "loss": 0.0, "step": 31422 }, { "epoch": 0.61, "learning_rate": 3.9838195239666435e-05, "loss": 0.0005, "step": 31424 }, { "epoch": 0.61, "learning_rate": 3.98375482794092e-05, "loss": 0.0009, "step": 31426 }, { "epoch": 0.61, "learning_rate": 3.9836901319151966e-05, "loss": 0.0007, "step": 31428 }, { "epoch": 0.61, "learning_rate": 3.9836254358894735e-05, "loss": 0.0001, "step": 31430 }, { "epoch": 0.61, "learning_rate": 3.9835607398637504e-05, "loss": 0.0015, "step": 31432 }, { "epoch": 0.61, "learning_rate": 3.9834960438380267e-05, "loss": 0.0, "step": 31434 }, { "epoch": 0.61, "learning_rate": 3.983431347812304e-05, "loss": 0.0002, "step": 31436 }, { "epoch": 0.61, "learning_rate": 3.983366651786581e-05, "loss": 0.0001, "step": 31438 }, { "epoch": 0.61, "learning_rate": 3.983301955760858e-05, "loss": 0.0, "step": 31440 }, { "epoch": 0.61, "learning_rate": 3.983237259735135e-05, "loss": 0.0001, "step": 31442 }, { "epoch": 0.61, "learning_rate": 3.983172563709411e-05, "loss": 0.0008, "step": 31444 }, { "epoch": 0.61, "learning_rate": 3.983107867683689e-05, "loss": 0.0, "step": 31446 }, { "epoch": 0.61, "learning_rate": 3.983043171657965e-05, "loss": 0.0001, "step": 31448 }, { "epoch": 0.61, "learning_rate": 3.982978475632242e-05, "loss": 0.0007, "step": 31450 }, { "epoch": 0.61, "learning_rate": 3.982913779606519e-05, "loss": 0.0, "step": 31452 }, { "epoch": 0.61, "learning_rate": 3.982849083580796e-05, "loss": 0.0, "step": 31454 }, { "epoch": 0.61, "learning_rate": 3.9827843875550727e-05, "loss": 0.0033, "step": 31456 }, { "epoch": 0.61, "learning_rate": 3.9827196915293496e-05, "loss": 0.0, "step": 31458 }, { "epoch": 0.61, "learning_rate": 3.9826549955036265e-05, "loss": 0.0, "step": 31460 }, { "epoch": 0.61, "learning_rate": 3.9825902994779034e-05, "loss": 0.0, "step": 31462 }, { "epoch": 0.61, "learning_rate": 3.98252560345218e-05, "loss": 0.0001, "step": 31464 }, { "epoch": 0.61, "learning_rate": 3.9824609074264565e-05, "loss": 0.0002, "step": 31466 }, { "epoch": 0.61, "learning_rate": 3.982396211400734e-05, "loss": 0.0, "step": 31468 }, { "epoch": 0.61, "learning_rate": 3.982331515375011e-05, "loss": 0.0004, "step": 31470 }, { "epoch": 0.61, "learning_rate": 3.982266819349287e-05, "loss": 0.0017, "step": 31472 }, { "epoch": 0.61, "learning_rate": 3.982202123323565e-05, "loss": 0.0, "step": 31474 }, { "epoch": 0.61, "learning_rate": 3.982137427297841e-05, "loss": 0.0, "step": 31476 }, { "epoch": 0.61, "learning_rate": 3.982072731272118e-05, "loss": 0.0064, "step": 31478 }, { "epoch": 0.61, "learning_rate": 3.982008035246395e-05, "loss": 0.0, "step": 31480 }, { "epoch": 0.61, "learning_rate": 3.981943339220672e-05, "loss": 0.0002, "step": 31482 }, { "epoch": 0.61, "learning_rate": 3.981878643194949e-05, "loss": 0.0013, "step": 31484 }, { "epoch": 0.61, "learning_rate": 3.9818139471692256e-05, "loss": 0.0126, "step": 31486 }, { "epoch": 0.61, "learning_rate": 3.9817492511435025e-05, "loss": 0.0, "step": 31488 }, { "epoch": 0.61, "learning_rate": 3.9816845551177794e-05, "loss": 0.0007, "step": 31490 }, { "epoch": 0.61, "learning_rate": 3.981619859092056e-05, "loss": 0.0, "step": 31492 }, { "epoch": 0.61, "learning_rate": 3.9815551630663326e-05, "loss": 0.0033, "step": 31494 }, { "epoch": 0.61, "learning_rate": 3.98149046704061e-05, "loss": 0.0003, "step": 31496 }, { "epoch": 0.61, "learning_rate": 3.9814257710148864e-05, "loss": 0.0001, "step": 31498 }, { "epoch": 0.61, "learning_rate": 3.981361074989163e-05, "loss": 0.0001, "step": 31500 }, { "epoch": 0.61, "learning_rate": 3.981296378963441e-05, "loss": 0.0001, "step": 31502 }, { "epoch": 0.61, "learning_rate": 3.981231682937717e-05, "loss": 0.0, "step": 31504 }, { "epoch": 0.61, "learning_rate": 3.981166986911995e-05, "loss": 0.0127, "step": 31506 }, { "epoch": 0.61, "learning_rate": 3.981102290886271e-05, "loss": 0.0001, "step": 31508 }, { "epoch": 0.61, "learning_rate": 3.981037594860548e-05, "loss": 0.0002, "step": 31510 }, { "epoch": 0.61, "learning_rate": 3.980972898834825e-05, "loss": 0.0034, "step": 31512 }, { "epoch": 0.61, "learning_rate": 3.9809082028091016e-05, "loss": 0.0, "step": 31514 }, { "epoch": 0.61, "learning_rate": 3.9808435067833785e-05, "loss": 0.0116, "step": 31516 }, { "epoch": 0.61, "learning_rate": 3.9807788107576555e-05, "loss": 0.0, "step": 31518 }, { "epoch": 0.61, "learning_rate": 3.9807141147319324e-05, "loss": 0.0002, "step": 31520 }, { "epoch": 0.61, "learning_rate": 3.980649418706209e-05, "loss": 0.0002, "step": 31522 }, { "epoch": 0.61, "learning_rate": 3.980584722680486e-05, "loss": 0.0007, "step": 31524 }, { "epoch": 0.61, "learning_rate": 3.9805200266547624e-05, "loss": 0.0, "step": 31526 }, { "epoch": 0.61, "learning_rate": 3.98045533062904e-05, "loss": 0.0, "step": 31528 }, { "epoch": 0.61, "learning_rate": 3.980390634603316e-05, "loss": 0.0001, "step": 31530 }, { "epoch": 0.61, "learning_rate": 3.980325938577593e-05, "loss": 0.0, "step": 31532 }, { "epoch": 0.61, "learning_rate": 3.98026124255187e-05, "loss": 0.0001, "step": 31534 }, { "epoch": 0.61, "learning_rate": 3.980196546526147e-05, "loss": 0.0, "step": 31536 }, { "epoch": 0.61, "learning_rate": 3.980131850500424e-05, "loss": 0.0003, "step": 31538 }, { "epoch": 0.61, "learning_rate": 3.980067154474701e-05, "loss": 0.0001, "step": 31540 }, { "epoch": 0.61, "learning_rate": 3.980002458448978e-05, "loss": 0.0001, "step": 31542 }, { "epoch": 0.61, "learning_rate": 3.9799377624232546e-05, "loss": 0.0, "step": 31544 }, { "epoch": 0.61, "learning_rate": 3.9798730663975315e-05, "loss": 0.0, "step": 31546 }, { "epoch": 0.61, "learning_rate": 3.9798083703718084e-05, "loss": 0.0, "step": 31548 }, { "epoch": 0.61, "learning_rate": 3.979743674346085e-05, "loss": 0.0, "step": 31550 }, { "epoch": 0.61, "learning_rate": 3.979711326333224e-05, "loss": 0.2036, "step": 31552 }, { "epoch": 0.61, "learning_rate": 3.9796466303075e-05, "loss": 0.0003, "step": 31554 }, { "epoch": 0.61, "learning_rate": 3.9795819342817776e-05, "loss": 0.0, "step": 31556 }, { "epoch": 0.61, "learning_rate": 3.9795172382560545e-05, "loss": 0.0002, "step": 31558 }, { "epoch": 0.61, "learning_rate": 3.979452542230331e-05, "loss": 0.0137, "step": 31560 }, { "epoch": 0.61, "learning_rate": 3.979387846204608e-05, "loss": 0.0038, "step": 31562 }, { "epoch": 0.61, "learning_rate": 3.9793231501788845e-05, "loss": 0.0002, "step": 31564 }, { "epoch": 0.61, "learning_rate": 3.9792584541531615e-05, "loss": 0.0096, "step": 31566 }, { "epoch": 0.61, "learning_rate": 3.9791937581274384e-05, "loss": 0.0058, "step": 31568 }, { "epoch": 0.61, "learning_rate": 3.979129062101715e-05, "loss": 0.0008, "step": 31570 }, { "epoch": 0.61, "learning_rate": 3.979064366075992e-05, "loss": 0.0, "step": 31572 }, { "epoch": 0.61, "learning_rate": 3.978999670050269e-05, "loss": 0.0001, "step": 31574 }, { "epoch": 0.61, "learning_rate": 3.978934974024546e-05, "loss": 0.0001, "step": 31576 }, { "epoch": 0.61, "learning_rate": 3.978870277998823e-05, "loss": 0.0, "step": 31578 }, { "epoch": 0.61, "learning_rate": 3.9788055819731e-05, "loss": 0.0019, "step": 31580 }, { "epoch": 0.61, "learning_rate": 3.978740885947376e-05, "loss": 0.0122, "step": 31582 }, { "epoch": 0.61, "learning_rate": 3.9786761899216536e-05, "loss": 0.0, "step": 31584 }, { "epoch": 0.61, "learning_rate": 3.97861149389593e-05, "loss": 0.0, "step": 31586 }, { "epoch": 0.61, "learning_rate": 3.978546797870207e-05, "loss": 0.0003, "step": 31588 }, { "epoch": 0.61, "learning_rate": 3.978482101844484e-05, "loss": 0.001, "step": 31590 }, { "epoch": 0.61, "learning_rate": 3.9784174058187606e-05, "loss": 0.0, "step": 31592 }, { "epoch": 0.61, "learning_rate": 3.978352709793038e-05, "loss": 0.0002, "step": 31594 }, { "epoch": 0.61, "learning_rate": 3.9782880137673144e-05, "loss": 0.0, "step": 31596 }, { "epoch": 0.61, "learning_rate": 3.978223317741591e-05, "loss": 0.0127, "step": 31598 }, { "epoch": 0.61, "learning_rate": 3.978158621715868e-05, "loss": 0.0052, "step": 31600 }, { "epoch": 0.61, "learning_rate": 3.978093925690145e-05, "loss": 0.0001, "step": 31602 }, { "epoch": 0.61, "learning_rate": 3.978029229664422e-05, "loss": 0.0, "step": 31604 }, { "epoch": 0.61, "learning_rate": 3.977964533638699e-05, "loss": 0.0062, "step": 31606 }, { "epoch": 0.61, "learning_rate": 3.977899837612976e-05, "loss": 0.0001, "step": 31608 }, { "epoch": 0.61, "learning_rate": 3.977835141587252e-05, "loss": 0.0089, "step": 31610 }, { "epoch": 0.61, "learning_rate": 3.97777044556153e-05, "loss": 0.0003, "step": 31612 }, { "epoch": 0.61, "learning_rate": 3.977705749535806e-05, "loss": 0.0, "step": 31614 }, { "epoch": 0.61, "learning_rate": 3.9776410535100835e-05, "loss": 0.0002, "step": 31616 }, { "epoch": 0.61, "learning_rate": 3.97757635748436e-05, "loss": 0.0056, "step": 31618 }, { "epoch": 0.61, "learning_rate": 3.9775116614586366e-05, "loss": 0.0, "step": 31620 }, { "epoch": 0.61, "learning_rate": 3.9774469654329135e-05, "loss": 0.0003, "step": 31622 }, { "epoch": 0.61, "learning_rate": 3.9773822694071904e-05, "loss": 0.0206, "step": 31624 }, { "epoch": 0.61, "learning_rate": 3.9773175733814673e-05, "loss": 0.0, "step": 31626 }, { "epoch": 0.61, "learning_rate": 3.977252877355744e-05, "loss": 0.0, "step": 31628 }, { "epoch": 0.61, "learning_rate": 3.977188181330021e-05, "loss": 0.0072, "step": 31630 }, { "epoch": 0.61, "learning_rate": 3.9771234853042974e-05, "loss": 0.0, "step": 31632 }, { "epoch": 0.61, "learning_rate": 3.977058789278575e-05, "loss": 0.0, "step": 31634 }, { "epoch": 0.61, "learning_rate": 3.976994093252851e-05, "loss": 0.0001, "step": 31636 }, { "epoch": 0.61, "learning_rate": 3.976929397227129e-05, "loss": 0.0098, "step": 31638 }, { "epoch": 0.61, "learning_rate": 3.976864701201406e-05, "loss": 0.0001, "step": 31640 }, { "epoch": 0.61, "learning_rate": 3.976800005175682e-05, "loss": 0.0001, "step": 31642 }, { "epoch": 0.61, "learning_rate": 3.9767353091499595e-05, "loss": 0.0007, "step": 31644 }, { "epoch": 0.61, "learning_rate": 3.976670613124236e-05, "loss": 0.0001, "step": 31646 }, { "epoch": 0.61, "learning_rate": 3.976605917098513e-05, "loss": 0.0002, "step": 31648 }, { "epoch": 0.61, "learning_rate": 3.9765412210727896e-05, "loss": 0.016, "step": 31650 }, { "epoch": 0.61, "learning_rate": 3.9764765250470665e-05, "loss": 0.0225, "step": 31652 }, { "epoch": 0.61, "learning_rate": 3.9764118290213434e-05, "loss": 0.0025, "step": 31654 }, { "epoch": 0.61, "learning_rate": 3.97634713299562e-05, "loss": 0.0, "step": 31656 }, { "epoch": 0.61, "learning_rate": 3.976282436969897e-05, "loss": 0.001, "step": 31658 }, { "epoch": 0.61, "learning_rate": 3.976217740944174e-05, "loss": 0.0167, "step": 31660 }, { "epoch": 0.61, "learning_rate": 3.976153044918451e-05, "loss": 0.0234, "step": 31662 }, { "epoch": 0.61, "learning_rate": 3.976088348892727e-05, "loss": 0.0, "step": 31664 }, { "epoch": 0.61, "learning_rate": 3.976023652867005e-05, "loss": 0.0, "step": 31666 }, { "epoch": 0.61, "learning_rate": 3.975958956841281e-05, "loss": 0.0, "step": 31668 }, { "epoch": 0.61, "learning_rate": 3.975894260815558e-05, "loss": 0.0001, "step": 31670 }, { "epoch": 0.61, "learning_rate": 3.9758295647898356e-05, "loss": 0.0002, "step": 31672 }, { "epoch": 0.61, "learning_rate": 3.975764868764112e-05, "loss": 0.0016, "step": 31674 }, { "epoch": 0.61, "learning_rate": 3.9757001727383894e-05, "loss": 0.0002, "step": 31676 }, { "epoch": 0.61, "learning_rate": 3.9756354767126656e-05, "loss": 0.0, "step": 31678 }, { "epoch": 0.61, "learning_rate": 3.9755707806869425e-05, "loss": 0.0002, "step": 31680 }, { "epoch": 0.61, "learning_rate": 3.9755060846612194e-05, "loss": 0.0, "step": 31682 }, { "epoch": 0.61, "learning_rate": 3.975441388635496e-05, "loss": 0.0053, "step": 31684 }, { "epoch": 0.61, "learning_rate": 3.975376692609773e-05, "loss": 0.0004, "step": 31686 }, { "epoch": 0.62, "learning_rate": 3.97531199658405e-05, "loss": 0.0002, "step": 31688 }, { "epoch": 0.62, "learning_rate": 3.975247300558327e-05, "loss": 0.0034, "step": 31690 }, { "epoch": 0.62, "learning_rate": 3.975182604532603e-05, "loss": 0.0002, "step": 31692 }, { "epoch": 0.62, "learning_rate": 3.975117908506881e-05, "loss": 0.0181, "step": 31694 }, { "epoch": 0.62, "learning_rate": 3.975053212481157e-05, "loss": 0.0001, "step": 31696 }, { "epoch": 0.62, "learning_rate": 3.974988516455435e-05, "loss": 0.0, "step": 31698 }, { "epoch": 0.62, "learning_rate": 3.974923820429711e-05, "loss": 0.0079, "step": 31700 }, { "epoch": 0.62, "learning_rate": 3.974859124403988e-05, "loss": 0.0002, "step": 31702 }, { "epoch": 0.62, "learning_rate": 3.9747944283782654e-05, "loss": 0.0007, "step": 31704 }, { "epoch": 0.62, "learning_rate": 3.9747297323525417e-05, "loss": 0.0, "step": 31706 }, { "epoch": 0.62, "learning_rate": 3.9746650363268186e-05, "loss": 0.0, "step": 31708 }, { "epoch": 0.62, "learning_rate": 3.9746003403010955e-05, "loss": 0.0001, "step": 31710 }, { "epoch": 0.62, "learning_rate": 3.9745356442753724e-05, "loss": 0.0022, "step": 31712 }, { "epoch": 0.62, "learning_rate": 3.9744709482496486e-05, "loss": 0.0, "step": 31714 }, { "epoch": 0.62, "learning_rate": 3.974406252223926e-05, "loss": 0.0002, "step": 31716 }, { "epoch": 0.62, "learning_rate": 3.974341556198203e-05, "loss": 0.0, "step": 31718 }, { "epoch": 0.62, "learning_rate": 3.97427686017248e-05, "loss": 0.0006, "step": 31720 }, { "epoch": 0.62, "learning_rate": 3.974212164146757e-05, "loss": 0.0001, "step": 31722 }, { "epoch": 0.62, "learning_rate": 3.974147468121033e-05, "loss": 0.0, "step": 31724 }, { "epoch": 0.62, "learning_rate": 3.974082772095311e-05, "loss": 0.0012, "step": 31726 }, { "epoch": 0.62, "learning_rate": 3.974018076069587e-05, "loss": 0.0067, "step": 31728 }, { "epoch": 0.62, "learning_rate": 3.973953380043864e-05, "loss": 0.001, "step": 31730 }, { "epoch": 0.62, "learning_rate": 3.973888684018141e-05, "loss": 0.0, "step": 31732 }, { "epoch": 0.62, "learning_rate": 3.973823987992418e-05, "loss": 0.0139, "step": 31734 }, { "epoch": 0.62, "learning_rate": 3.973759291966695e-05, "loss": 0.0036, "step": 31736 }, { "epoch": 0.62, "learning_rate": 3.9736945959409715e-05, "loss": 0.0001, "step": 31738 }, { "epoch": 0.62, "learning_rate": 3.9736298999152484e-05, "loss": 0.0008, "step": 31740 }, { "epoch": 0.62, "learning_rate": 3.973565203889525e-05, "loss": 0.0003, "step": 31742 }, { "epoch": 0.62, "learning_rate": 3.973500507863802e-05, "loss": 0.0002, "step": 31744 }, { "epoch": 0.62, "learning_rate": 3.9734358118380785e-05, "loss": 0.0005, "step": 31746 }, { "epoch": 0.62, "learning_rate": 3.973371115812356e-05, "loss": 0.0002, "step": 31748 }, { "epoch": 0.62, "learning_rate": 3.973306419786633e-05, "loss": 0.0, "step": 31750 }, { "epoch": 0.62, "learning_rate": 3.973241723760909e-05, "loss": 0.0003, "step": 31752 }, { "epoch": 0.62, "learning_rate": 3.973177027735187e-05, "loss": 0.0, "step": 31754 }, { "epoch": 0.62, "learning_rate": 3.973112331709463e-05, "loss": 0.0, "step": 31756 }, { "epoch": 0.62, "learning_rate": 3.9730476356837406e-05, "loss": 0.0009, "step": 31758 }, { "epoch": 0.62, "learning_rate": 3.972982939658017e-05, "loss": 0.0239, "step": 31760 }, { "epoch": 0.62, "learning_rate": 3.972918243632294e-05, "loss": 0.0026, "step": 31762 }, { "epoch": 0.62, "learning_rate": 3.9728535476065706e-05, "loss": 0.0, "step": 31764 }, { "epoch": 0.62, "learning_rate": 3.9727888515808476e-05, "loss": 0.0051, "step": 31766 }, { "epoch": 0.62, "learning_rate": 3.9727241555551245e-05, "loss": 0.0, "step": 31768 }, { "epoch": 0.62, "learning_rate": 3.9726594595294014e-05, "loss": 0.0001, "step": 31770 }, { "epoch": 0.62, "learning_rate": 3.972594763503678e-05, "loss": 0.0053, "step": 31772 }, { "epoch": 0.62, "learning_rate": 3.9725300674779545e-05, "loss": 0.0001, "step": 31774 }, { "epoch": 0.62, "learning_rate": 3.972465371452232e-05, "loss": 0.0006, "step": 31776 }, { "epoch": 0.62, "learning_rate": 3.972400675426508e-05, "loss": 0.0004, "step": 31778 }, { "epoch": 0.62, "learning_rate": 3.972335979400786e-05, "loss": 0.0054, "step": 31780 }, { "epoch": 0.62, "learning_rate": 3.972271283375063e-05, "loss": 0.0004, "step": 31782 }, { "epoch": 0.62, "learning_rate": 3.972206587349339e-05, "loss": 0.0, "step": 31784 }, { "epoch": 0.62, "learning_rate": 3.9721418913236166e-05, "loss": 0.0001, "step": 31786 }, { "epoch": 0.62, "learning_rate": 3.972077195297893e-05, "loss": 0.0001, "step": 31788 }, { "epoch": 0.62, "learning_rate": 3.97201249927217e-05, "loss": 0.0001, "step": 31790 }, { "epoch": 0.62, "learning_rate": 3.971947803246447e-05, "loss": 0.0001, "step": 31792 }, { "epoch": 0.62, "learning_rate": 3.9718831072207236e-05, "loss": 0.0001, "step": 31794 }, { "epoch": 0.62, "learning_rate": 3.9718184111950005e-05, "loss": 0.0001, "step": 31796 }, { "epoch": 0.62, "learning_rate": 3.9717537151692774e-05, "loss": 0.0, "step": 31798 }, { "epoch": 0.62, "learning_rate": 3.971689019143554e-05, "loss": 0.0002, "step": 31800 }, { "epoch": 0.62, "learning_rate": 3.971624323117831e-05, "loss": 0.0002, "step": 31802 }, { "epoch": 0.62, "learning_rate": 3.971559627092108e-05, "loss": 0.0023, "step": 31804 }, { "epoch": 0.62, "learning_rate": 3.9714949310663844e-05, "loss": 0.0001, "step": 31806 }, { "epoch": 0.62, "learning_rate": 3.971430235040662e-05, "loss": 0.0, "step": 31808 }, { "epoch": 0.62, "learning_rate": 3.971365539014938e-05, "loss": 0.0001, "step": 31810 }, { "epoch": 0.62, "learning_rate": 3.971300842989215e-05, "loss": 0.0001, "step": 31812 }, { "epoch": 0.62, "learning_rate": 3.971236146963492e-05, "loss": 0.0, "step": 31814 }, { "epoch": 0.62, "learning_rate": 3.971171450937769e-05, "loss": 0.0001, "step": 31816 }, { "epoch": 0.62, "learning_rate": 3.9711067549120465e-05, "loss": 0.0001, "step": 31818 }, { "epoch": 0.62, "learning_rate": 3.971042058886323e-05, "loss": 0.0041, "step": 31820 }, { "epoch": 0.62, "learning_rate": 3.9709773628605996e-05, "loss": 0.0004, "step": 31822 }, { "epoch": 0.62, "learning_rate": 3.9709126668348765e-05, "loss": 0.0001, "step": 31824 }, { "epoch": 0.62, "learning_rate": 3.9708479708091534e-05, "loss": 0.0, "step": 31826 }, { "epoch": 0.62, "learning_rate": 3.9707832747834304e-05, "loss": 0.0006, "step": 31828 }, { "epoch": 0.62, "learning_rate": 3.970718578757707e-05, "loss": 0.0004, "step": 31830 }, { "epoch": 0.62, "learning_rate": 3.970653882731984e-05, "loss": 0.0, "step": 31832 }, { "epoch": 0.62, "learning_rate": 3.9705891867062604e-05, "loss": 0.0004, "step": 31834 }, { "epoch": 0.62, "learning_rate": 3.970524490680538e-05, "loss": 0.0018, "step": 31836 }, { "epoch": 0.62, "learning_rate": 3.970459794654814e-05, "loss": 0.0003, "step": 31838 }, { "epoch": 0.62, "learning_rate": 3.970395098629092e-05, "loss": 0.0159, "step": 31840 }, { "epoch": 0.62, "learning_rate": 3.970330402603368e-05, "loss": 0.0036, "step": 31842 }, { "epoch": 0.62, "learning_rate": 3.970265706577645e-05, "loss": 0.0004, "step": 31844 }, { "epoch": 0.62, "learning_rate": 3.970201010551922e-05, "loss": 0.0105, "step": 31846 }, { "epoch": 0.62, "learning_rate": 3.970136314526199e-05, "loss": 0.0014, "step": 31848 }, { "epoch": 0.62, "learning_rate": 3.970071618500476e-05, "loss": 0.0156, "step": 31850 }, { "epoch": 0.62, "learning_rate": 3.9700069224747526e-05, "loss": 0.0, "step": 31852 }, { "epoch": 0.62, "learning_rate": 3.9699422264490295e-05, "loss": 0.0, "step": 31854 }, { "epoch": 0.62, "learning_rate": 3.969877530423306e-05, "loss": 0.0001, "step": 31856 }, { "epoch": 0.62, "learning_rate": 3.969812834397583e-05, "loss": 0.0, "step": 31858 }, { "epoch": 0.62, "learning_rate": 3.96974813837186e-05, "loss": 0.0, "step": 31860 }, { "epoch": 0.62, "learning_rate": 3.969683442346137e-05, "loss": 0.0119, "step": 31862 }, { "epoch": 0.62, "learning_rate": 3.969618746320414e-05, "loss": 0.0, "step": 31864 }, { "epoch": 0.62, "learning_rate": 3.96955405029469e-05, "loss": 0.0, "step": 31866 }, { "epoch": 0.62, "learning_rate": 3.969489354268968e-05, "loss": 0.0002, "step": 31868 }, { "epoch": 0.62, "learning_rate": 3.969424658243244e-05, "loss": 0.0342, "step": 31870 }, { "epoch": 0.62, "learning_rate": 3.969359962217521e-05, "loss": 0.0, "step": 31872 }, { "epoch": 0.62, "learning_rate": 3.969295266191798e-05, "loss": 0.0002, "step": 31874 }, { "epoch": 0.62, "learning_rate": 3.969230570166075e-05, "loss": 0.0005, "step": 31876 }, { "epoch": 0.62, "learning_rate": 3.969165874140352e-05, "loss": 0.0, "step": 31878 }, { "epoch": 0.62, "learning_rate": 3.9691011781146286e-05, "loss": 0.0087, "step": 31880 }, { "epoch": 0.62, "learning_rate": 3.9690364820889055e-05, "loss": 0.0003, "step": 31882 }, { "epoch": 0.62, "learning_rate": 3.9689717860631824e-05, "loss": 0.0, "step": 31884 }, { "epoch": 0.62, "learning_rate": 3.9689070900374593e-05, "loss": 0.001, "step": 31886 }, { "epoch": 0.62, "learning_rate": 3.9688423940117356e-05, "loss": 0.0042, "step": 31888 }, { "epoch": 0.62, "learning_rate": 3.968777697986013e-05, "loss": 0.0002, "step": 31890 }, { "epoch": 0.62, "learning_rate": 3.9687130019602894e-05, "loss": 0.0001, "step": 31892 }, { "epoch": 0.62, "learning_rate": 3.968648305934566e-05, "loss": 0.0053, "step": 31894 }, { "epoch": 0.62, "learning_rate": 3.968583609908844e-05, "loss": 0.0001, "step": 31896 }, { "epoch": 0.62, "learning_rate": 3.96851891388312e-05, "loss": 0.0007, "step": 31898 }, { "epoch": 0.62, "learning_rate": 3.968454217857398e-05, "loss": 0.0174, "step": 31900 }, { "epoch": 0.62, "learning_rate": 3.968389521831674e-05, "loss": 0.0, "step": 31902 }, { "epoch": 0.62, "learning_rate": 3.968324825805951e-05, "loss": 0.0004, "step": 31904 }, { "epoch": 0.62, "learning_rate": 3.968260129780228e-05, "loss": 0.0003, "step": 31906 }, { "epoch": 0.62, "learning_rate": 3.968195433754505e-05, "loss": 0.0007, "step": 31908 }, { "epoch": 0.62, "learning_rate": 3.9681307377287816e-05, "loss": 0.0, "step": 31910 }, { "epoch": 0.62, "learning_rate": 3.9680660417030585e-05, "loss": 0.0127, "step": 31912 }, { "epoch": 0.62, "learning_rate": 3.9680013456773354e-05, "loss": 0.0244, "step": 31914 }, { "epoch": 0.62, "learning_rate": 3.9679366496516116e-05, "loss": 0.0001, "step": 31916 }, { "epoch": 0.62, "learning_rate": 3.967871953625889e-05, "loss": 0.0162, "step": 31918 }, { "epoch": 0.62, "learning_rate": 3.9678072576001654e-05, "loss": 0.0081, "step": 31920 }, { "epoch": 0.62, "learning_rate": 3.967742561574443e-05, "loss": 0.0007, "step": 31922 }, { "epoch": 0.62, "learning_rate": 3.967677865548719e-05, "loss": 0.0043, "step": 31924 }, { "epoch": 0.62, "learning_rate": 3.967613169522996e-05, "loss": 0.0065, "step": 31926 }, { "epoch": 0.62, "learning_rate": 3.967548473497274e-05, "loss": 0.0, "step": 31928 }, { "epoch": 0.62, "learning_rate": 3.96748377747155e-05, "loss": 0.0139, "step": 31930 }, { "epoch": 0.62, "learning_rate": 3.967419081445827e-05, "loss": 0.0003, "step": 31932 }, { "epoch": 0.62, "learning_rate": 3.967354385420104e-05, "loss": 0.0024, "step": 31934 }, { "epoch": 0.62, "learning_rate": 3.967289689394381e-05, "loss": 0.0001, "step": 31936 }, { "epoch": 0.62, "learning_rate": 3.9672249933686576e-05, "loss": 0.0, "step": 31938 }, { "epoch": 0.62, "learning_rate": 3.9671602973429345e-05, "loss": 0.0002, "step": 31940 }, { "epoch": 0.62, "learning_rate": 3.9670956013172114e-05, "loss": 0.0004, "step": 31942 }, { "epoch": 0.62, "learning_rate": 3.967030905291488e-05, "loss": 0.0002, "step": 31944 }, { "epoch": 0.62, "learning_rate": 3.966966209265765e-05, "loss": 0.0, "step": 31946 }, { "epoch": 0.62, "learning_rate": 3.9669015132400415e-05, "loss": 0.0002, "step": 31948 }, { "epoch": 0.62, "learning_rate": 3.966836817214319e-05, "loss": 0.0005, "step": 31950 }, { "epoch": 0.62, "learning_rate": 3.966772121188595e-05, "loss": 0.01, "step": 31952 }, { "epoch": 0.62, "learning_rate": 3.966707425162872e-05, "loss": 0.0, "step": 31954 }, { "epoch": 0.62, "learning_rate": 3.966642729137149e-05, "loss": 0.0141, "step": 31956 }, { "epoch": 0.62, "learning_rate": 3.966578033111426e-05, "loss": 0.0028, "step": 31958 }, { "epoch": 0.62, "learning_rate": 3.9665133370857036e-05, "loss": 0.0002, "step": 31960 }, { "epoch": 0.62, "learning_rate": 3.96644864105998e-05, "loss": 0.0002, "step": 31962 }, { "epoch": 0.62, "learning_rate": 3.966383945034257e-05, "loss": 0.0001, "step": 31964 }, { "epoch": 0.62, "learning_rate": 3.9663192490085337e-05, "loss": 0.0, "step": 31966 }, { "epoch": 0.62, "learning_rate": 3.9662545529828106e-05, "loss": 0.0001, "step": 31968 }, { "epoch": 0.62, "learning_rate": 3.966189856957087e-05, "loss": 0.0037, "step": 31970 }, { "epoch": 0.62, "learning_rate": 3.9661251609313644e-05, "loss": 0.0001, "step": 31972 }, { "epoch": 0.62, "learning_rate": 3.966060464905641e-05, "loss": 0.0032, "step": 31974 }, { "epoch": 0.62, "learning_rate": 3.9659957688799175e-05, "loss": 0.0017, "step": 31976 }, { "epoch": 0.62, "learning_rate": 3.965931072854195e-05, "loss": 0.0, "step": 31978 }, { "epoch": 0.62, "learning_rate": 3.965866376828471e-05, "loss": 0.0, "step": 31980 }, { "epoch": 0.62, "learning_rate": 3.965801680802749e-05, "loss": 0.0, "step": 31982 }, { "epoch": 0.62, "learning_rate": 3.965736984777025e-05, "loss": 0.0011, "step": 31984 }, { "epoch": 0.62, "learning_rate": 3.965672288751302e-05, "loss": 0.0137, "step": 31986 }, { "epoch": 0.62, "learning_rate": 3.965607592725579e-05, "loss": 0.0084, "step": 31988 }, { "epoch": 0.62, "learning_rate": 3.965542896699856e-05, "loss": 0.0, "step": 31990 }, { "epoch": 0.62, "learning_rate": 3.965478200674133e-05, "loss": 0.0, "step": 31992 }, { "epoch": 0.62, "learning_rate": 3.96541350464841e-05, "loss": 0.0, "step": 31994 }, { "epoch": 0.62, "learning_rate": 3.9653488086226866e-05, "loss": 0.0, "step": 31996 }, { "epoch": 0.62, "learning_rate": 3.965284112596963e-05, "loss": 0.003, "step": 31998 }, { "epoch": 0.62, "learning_rate": 3.9652194165712404e-05, "loss": 0.0008, "step": 32000 }, { "epoch": 0.62, "learning_rate": 3.9651547205455166e-05, "loss": 0.0072, "step": 32002 }, { "epoch": 0.62, "learning_rate": 3.965090024519794e-05, "loss": 0.0001, "step": 32004 }, { "epoch": 0.62, "learning_rate": 3.965025328494071e-05, "loss": 0.0027, "step": 32006 }, { "epoch": 0.62, "learning_rate": 3.9649606324683474e-05, "loss": 0.0, "step": 32008 }, { "epoch": 0.62, "learning_rate": 3.964895936442625e-05, "loss": 0.0667, "step": 32010 }, { "epoch": 0.62, "learning_rate": 3.964831240416901e-05, "loss": 0.0006, "step": 32012 }, { "epoch": 0.62, "learning_rate": 3.964766544391178e-05, "loss": 0.0, "step": 32014 }, { "epoch": 0.62, "learning_rate": 3.964701848365455e-05, "loss": 0.0017, "step": 32016 }, { "epoch": 0.62, "learning_rate": 3.964637152339732e-05, "loss": 0.0001, "step": 32018 }, { "epoch": 0.62, "learning_rate": 3.964572456314009e-05, "loss": 0.0, "step": 32020 }, { "epoch": 0.62, "learning_rate": 3.964507760288286e-05, "loss": 0.0001, "step": 32022 }, { "epoch": 0.62, "learning_rate": 3.9644430642625626e-05, "loss": 0.0057, "step": 32024 }, { "epoch": 0.62, "learning_rate": 3.9643783682368395e-05, "loss": 0.0001, "step": 32026 }, { "epoch": 0.62, "learning_rate": 3.9643136722111165e-05, "loss": 0.0012, "step": 32028 }, { "epoch": 0.62, "learning_rate": 3.964248976185393e-05, "loss": 0.0001, "step": 32030 }, { "epoch": 0.62, "learning_rate": 3.96418428015967e-05, "loss": 0.0001, "step": 32032 }, { "epoch": 0.62, "learning_rate": 3.9641195841339465e-05, "loss": 0.0025, "step": 32034 }, { "epoch": 0.62, "learning_rate": 3.9640548881082234e-05, "loss": 0.0033, "step": 32036 }, { "epoch": 0.62, "learning_rate": 3.963990192082501e-05, "loss": 0.0318, "step": 32038 }, { "epoch": 0.62, "learning_rate": 3.963925496056777e-05, "loss": 0.0001, "step": 32040 }, { "epoch": 0.62, "learning_rate": 3.963860800031055e-05, "loss": 0.0001, "step": 32042 }, { "epoch": 0.62, "learning_rate": 3.963796104005331e-05, "loss": 0.0242, "step": 32044 }, { "epoch": 0.62, "learning_rate": 3.963731407979608e-05, "loss": 0.0005, "step": 32046 }, { "epoch": 0.62, "learning_rate": 3.963666711953885e-05, "loss": 0.0, "step": 32048 }, { "epoch": 0.62, "learning_rate": 3.963602015928162e-05, "loss": 0.0001, "step": 32050 }, { "epoch": 0.62, "learning_rate": 3.963537319902439e-05, "loss": 0.0098, "step": 32052 }, { "epoch": 0.62, "learning_rate": 3.9634726238767156e-05, "loss": 0.0047, "step": 32054 }, { "epoch": 0.62, "learning_rate": 3.9634079278509925e-05, "loss": 0.0, "step": 32056 }, { "epoch": 0.62, "learning_rate": 3.963343231825269e-05, "loss": 0.0, "step": 32058 }, { "epoch": 0.62, "learning_rate": 3.963278535799546e-05, "loss": 0.0, "step": 32060 }, { "epoch": 0.62, "learning_rate": 3.9632138397738225e-05, "loss": 0.0001, "step": 32062 }, { "epoch": 0.62, "learning_rate": 3.9631491437481e-05, "loss": 0.0, "step": 32064 }, { "epoch": 0.62, "learning_rate": 3.9630844477223764e-05, "loss": 0.0001, "step": 32066 }, { "epoch": 0.62, "learning_rate": 3.963019751696653e-05, "loss": 0.0, "step": 32068 }, { "epoch": 0.62, "learning_rate": 3.96295505567093e-05, "loss": 0.0001, "step": 32070 }, { "epoch": 0.62, "learning_rate": 3.962890359645207e-05, "loss": 0.0, "step": 32072 }, { "epoch": 0.62, "learning_rate": 3.962825663619484e-05, "loss": 0.0, "step": 32074 }, { "epoch": 0.62, "learning_rate": 3.962760967593761e-05, "loss": 0.0, "step": 32076 }, { "epoch": 0.62, "learning_rate": 3.962696271568038e-05, "loss": 0.0, "step": 32078 }, { "epoch": 0.62, "learning_rate": 3.962631575542315e-05, "loss": 0.0003, "step": 32080 }, { "epoch": 0.62, "learning_rate": 3.9625668795165916e-05, "loss": 0.0038, "step": 32082 }, { "epoch": 0.62, "learning_rate": 3.9625021834908685e-05, "loss": 0.0, "step": 32084 }, { "epoch": 0.62, "learning_rate": 3.9624374874651454e-05, "loss": 0.0, "step": 32086 }, { "epoch": 0.62, "learning_rate": 3.9623727914394224e-05, "loss": 0.0087, "step": 32088 }, { "epoch": 0.62, "learning_rate": 3.9623080954136986e-05, "loss": 0.0048, "step": 32090 }, { "epoch": 0.62, "learning_rate": 3.962243399387976e-05, "loss": 0.0013, "step": 32092 }, { "epoch": 0.62, "learning_rate": 3.9621787033622524e-05, "loss": 0.0, "step": 32094 }, { "epoch": 0.62, "learning_rate": 3.962114007336529e-05, "loss": 0.0111, "step": 32096 }, { "epoch": 0.62, "learning_rate": 3.962049311310806e-05, "loss": 0.0001, "step": 32098 }, { "epoch": 0.62, "learning_rate": 3.961984615285083e-05, "loss": 0.0, "step": 32100 }, { "epoch": 0.62, "learning_rate": 3.96191991925936e-05, "loss": 0.0002, "step": 32102 }, { "epoch": 0.62, "learning_rate": 3.961855223233637e-05, "loss": 0.0001, "step": 32104 }, { "epoch": 0.62, "learning_rate": 3.961790527207914e-05, "loss": 0.0, "step": 32106 }, { "epoch": 0.62, "learning_rate": 3.961725831182191e-05, "loss": 0.0083, "step": 32108 }, { "epoch": 0.62, "learning_rate": 3.961661135156468e-05, "loss": 0.0047, "step": 32110 }, { "epoch": 0.62, "learning_rate": 3.961596439130744e-05, "loss": 0.0067, "step": 32112 }, { "epoch": 0.62, "learning_rate": 3.9615317431050215e-05, "loss": 0.0, "step": 32114 }, { "epoch": 0.62, "learning_rate": 3.961467047079298e-05, "loss": 0.0001, "step": 32116 }, { "epoch": 0.62, "learning_rate": 3.9614023510535746e-05, "loss": 0.0, "step": 32118 }, { "epoch": 0.62, "learning_rate": 3.961337655027852e-05, "loss": 0.0001, "step": 32120 }, { "epoch": 0.62, "learning_rate": 3.9612729590021284e-05, "loss": 0.0, "step": 32122 }, { "epoch": 0.62, "learning_rate": 3.961208262976406e-05, "loss": 0.0001, "step": 32124 }, { "epoch": 0.62, "learning_rate": 3.961143566950682e-05, "loss": 0.0066, "step": 32126 }, { "epoch": 0.62, "learning_rate": 3.961078870924959e-05, "loss": 0.0, "step": 32128 }, { "epoch": 0.62, "learning_rate": 3.961014174899236e-05, "loss": 0.0, "step": 32130 }, { "epoch": 0.62, "learning_rate": 3.960949478873513e-05, "loss": 0.0002, "step": 32132 }, { "epoch": 0.62, "learning_rate": 3.96088478284779e-05, "loss": 0.0033, "step": 32134 }, { "epoch": 0.62, "learning_rate": 3.960820086822067e-05, "loss": 0.0, "step": 32136 }, { "epoch": 0.62, "learning_rate": 3.960755390796344e-05, "loss": 0.0082, "step": 32138 }, { "epoch": 0.62, "learning_rate": 3.96069069477062e-05, "loss": 0.0002, "step": 32140 }, { "epoch": 0.62, "learning_rate": 3.9606259987448975e-05, "loss": 0.0, "step": 32142 }, { "epoch": 0.62, "learning_rate": 3.960561302719174e-05, "loss": 0.0001, "step": 32144 }, { "epoch": 0.62, "learning_rate": 3.9604966066934513e-05, "loss": 0.0001, "step": 32146 }, { "epoch": 0.62, "learning_rate": 3.9604319106677276e-05, "loss": 0.0003, "step": 32148 }, { "epoch": 0.62, "learning_rate": 3.9603672146420045e-05, "loss": 0.0034, "step": 32150 }, { "epoch": 0.62, "learning_rate": 3.960302518616282e-05, "loss": 0.0, "step": 32152 }, { "epoch": 0.62, "learning_rate": 3.960237822590558e-05, "loss": 0.0, "step": 32154 }, { "epoch": 0.62, "learning_rate": 3.960173126564835e-05, "loss": 0.0258, "step": 32156 }, { "epoch": 0.62, "learning_rate": 3.960108430539112e-05, "loss": 0.0001, "step": 32158 }, { "epoch": 0.62, "learning_rate": 3.960043734513389e-05, "loss": 0.0004, "step": 32160 }, { "epoch": 0.62, "learning_rate": 3.959979038487666e-05, "loss": 0.0002, "step": 32162 }, { "epoch": 0.62, "learning_rate": 3.959914342461943e-05, "loss": 0.0024, "step": 32164 }, { "epoch": 0.62, "learning_rate": 3.95984964643622e-05, "loss": 0.0013, "step": 32166 }, { "epoch": 0.62, "learning_rate": 3.9597849504104967e-05, "loss": 0.0027, "step": 32168 }, { "epoch": 0.62, "learning_rate": 3.9597202543847736e-05, "loss": 0.002, "step": 32170 }, { "epoch": 0.62, "learning_rate": 3.95965555835905e-05, "loss": 0.0, "step": 32172 }, { "epoch": 0.62, "learning_rate": 3.9595908623333274e-05, "loss": 0.0, "step": 32174 }, { "epoch": 0.62, "learning_rate": 3.9595261663076036e-05, "loss": 0.0002, "step": 32176 }, { "epoch": 0.62, "learning_rate": 3.9594614702818805e-05, "loss": 0.007, "step": 32178 }, { "epoch": 0.62, "learning_rate": 3.9593967742561574e-05, "loss": 0.0016, "step": 32180 }, { "epoch": 0.62, "learning_rate": 3.9593320782304343e-05, "loss": 0.0001, "step": 32182 }, { "epoch": 0.62, "learning_rate": 3.959267382204712e-05, "loss": 0.0, "step": 32184 }, { "epoch": 0.62, "learning_rate": 3.959202686178988e-05, "loss": 0.0, "step": 32186 }, { "epoch": 0.62, "learning_rate": 3.959137990153265e-05, "loss": 0.0, "step": 32188 }, { "epoch": 0.62, "learning_rate": 3.959073294127542e-05, "loss": 0.0001, "step": 32190 }, { "epoch": 0.62, "learning_rate": 3.959008598101819e-05, "loss": 0.0, "step": 32192 }, { "epoch": 0.62, "learning_rate": 3.958943902076095e-05, "loss": 0.0, "step": 32194 }, { "epoch": 0.62, "learning_rate": 3.958879206050373e-05, "loss": 0.0089, "step": 32196 }, { "epoch": 0.62, "learning_rate": 3.9588145100246496e-05, "loss": 0.0, "step": 32198 }, { "epoch": 0.62, "learning_rate": 3.958749813998926e-05, "loss": 0.0001, "step": 32200 }, { "epoch": 0.63, "learning_rate": 3.9586851179732034e-05, "loss": 0.0008, "step": 32202 }, { "epoch": 0.63, "learning_rate": 3.9586204219474797e-05, "loss": 0.0001, "step": 32204 }, { "epoch": 0.63, "learning_rate": 3.958555725921757e-05, "loss": 0.0006, "step": 32206 }, { "epoch": 0.63, "learning_rate": 3.9584910298960335e-05, "loss": 0.0001, "step": 32208 }, { "epoch": 0.63, "learning_rate": 3.9584263338703104e-05, "loss": 0.0001, "step": 32210 }, { "epoch": 0.63, "learning_rate": 3.958361637844587e-05, "loss": 0.0003, "step": 32212 }, { "epoch": 0.63, "learning_rate": 3.958296941818864e-05, "loss": 0.0006, "step": 32214 }, { "epoch": 0.63, "learning_rate": 3.958232245793141e-05, "loss": 0.0028, "step": 32216 }, { "epoch": 0.63, "learning_rate": 3.958167549767418e-05, "loss": 0.0031, "step": 32218 }, { "epoch": 0.63, "learning_rate": 3.958102853741695e-05, "loss": 0.0, "step": 32220 }, { "epoch": 0.63, "learning_rate": 3.958038157715972e-05, "loss": 0.0, "step": 32222 }, { "epoch": 0.63, "learning_rate": 3.957973461690249e-05, "loss": 0.0007, "step": 32224 }, { "epoch": 0.63, "learning_rate": 3.957908765664525e-05, "loss": 0.0001, "step": 32226 }, { "epoch": 0.63, "learning_rate": 3.9578440696388026e-05, "loss": 0.0002, "step": 32228 }, { "epoch": 0.63, "learning_rate": 3.9577793736130795e-05, "loss": 0.0054, "step": 32230 }, { "epoch": 0.63, "learning_rate": 3.957714677587356e-05, "loss": 0.0043, "step": 32232 }, { "epoch": 0.63, "learning_rate": 3.957649981561633e-05, "loss": 0.0018, "step": 32234 }, { "epoch": 0.63, "learning_rate": 3.9575852855359095e-05, "loss": 0.0001, "step": 32236 }, { "epoch": 0.63, "learning_rate": 3.9575205895101864e-05, "loss": 0.006, "step": 32238 }, { "epoch": 0.63, "learning_rate": 3.957455893484463e-05, "loss": 0.0001, "step": 32240 }, { "epoch": 0.63, "learning_rate": 3.95739119745874e-05, "loss": 0.0026, "step": 32242 }, { "epoch": 0.63, "learning_rate": 3.957326501433017e-05, "loss": 0.0001, "step": 32244 }, { "epoch": 0.63, "learning_rate": 3.957261805407294e-05, "loss": 0.0001, "step": 32246 }, { "epoch": 0.63, "learning_rate": 3.957197109381571e-05, "loss": 0.0001, "step": 32248 }, { "epoch": 0.63, "learning_rate": 3.957132413355848e-05, "loss": 0.006, "step": 32250 }, { "epoch": 0.63, "learning_rate": 3.957067717330125e-05, "loss": 0.0003, "step": 32252 }, { "epoch": 0.63, "learning_rate": 3.957003021304401e-05, "loss": 0.014, "step": 32254 }, { "epoch": 0.63, "learning_rate": 3.9569383252786786e-05, "loss": 0.0001, "step": 32256 }, { "epoch": 0.63, "learning_rate": 3.956873629252955e-05, "loss": 0.0164, "step": 32258 }, { "epoch": 0.63, "learning_rate": 3.956808933227232e-05, "loss": 0.0032, "step": 32260 }, { "epoch": 0.63, "learning_rate": 3.956744237201509e-05, "loss": 0.0003, "step": 32262 }, { "epoch": 0.63, "learning_rate": 3.9566795411757856e-05, "loss": 0.0, "step": 32264 }, { "epoch": 0.63, "learning_rate": 3.956614845150063e-05, "loss": 0.0008, "step": 32266 }, { "epoch": 0.63, "learning_rate": 3.9565501491243394e-05, "loss": 0.0, "step": 32268 }, { "epoch": 0.63, "learning_rate": 3.956485453098616e-05, "loss": 0.0008, "step": 32270 }, { "epoch": 0.63, "learning_rate": 3.956420757072893e-05, "loss": 0.0, "step": 32272 }, { "epoch": 0.63, "learning_rate": 3.95635606104717e-05, "loss": 0.0, "step": 32274 }, { "epoch": 0.63, "learning_rate": 3.956291365021447e-05, "loss": 0.0001, "step": 32276 }, { "epoch": 0.63, "learning_rate": 3.956226668995724e-05, "loss": 0.0, "step": 32278 }, { "epoch": 0.63, "learning_rate": 3.956161972970001e-05, "loss": 0.0, "step": 32280 }, { "epoch": 0.63, "learning_rate": 3.956097276944277e-05, "loss": 0.0002, "step": 32282 }, { "epoch": 0.63, "learning_rate": 3.9560325809185546e-05, "loss": 0.0, "step": 32284 }, { "epoch": 0.63, "learning_rate": 3.955967884892831e-05, "loss": 0.0001, "step": 32286 }, { "epoch": 0.63, "learning_rate": 3.9559031888671085e-05, "loss": 0.0006, "step": 32288 }, { "epoch": 0.63, "learning_rate": 3.955838492841385e-05, "loss": 0.0001, "step": 32290 }, { "epoch": 0.63, "learning_rate": 3.9557737968156616e-05, "loss": 0.0001, "step": 32292 }, { "epoch": 0.63, "learning_rate": 3.9557091007899385e-05, "loss": 0.0, "step": 32294 }, { "epoch": 0.63, "learning_rate": 3.9556444047642154e-05, "loss": 0.0081, "step": 32296 }, { "epoch": 0.63, "learning_rate": 3.955579708738492e-05, "loss": 0.0003, "step": 32298 }, { "epoch": 0.63, "learning_rate": 3.955515012712769e-05, "loss": 0.0005, "step": 32300 }, { "epoch": 0.63, "learning_rate": 3.955450316687046e-05, "loss": 0.0022, "step": 32302 }, { "epoch": 0.63, "learning_rate": 3.955385620661323e-05, "loss": 0.0002, "step": 32304 }, { "epoch": 0.63, "learning_rate": 3.9553209246356e-05, "loss": 0.0001, "step": 32306 }, { "epoch": 0.63, "learning_rate": 3.955256228609877e-05, "loss": 0.0, "step": 32308 }, { "epoch": 0.63, "learning_rate": 3.955191532584154e-05, "loss": 0.0, "step": 32310 }, { "epoch": 0.63, "learning_rate": 3.955126836558431e-05, "loss": 0.0014, "step": 32312 }, { "epoch": 0.63, "learning_rate": 3.955062140532707e-05, "loss": 0.0, "step": 32314 }, { "epoch": 0.63, "learning_rate": 3.9549974445069845e-05, "loss": 0.0, "step": 32316 }, { "epoch": 0.63, "learning_rate": 3.954932748481261e-05, "loss": 0.0074, "step": 32318 }, { "epoch": 0.63, "learning_rate": 3.9548680524555376e-05, "loss": 0.0, "step": 32320 }, { "epoch": 0.63, "learning_rate": 3.9548033564298145e-05, "loss": 0.0, "step": 32322 }, { "epoch": 0.63, "learning_rate": 3.9547386604040915e-05, "loss": 0.0001, "step": 32324 }, { "epoch": 0.63, "learning_rate": 3.9546739643783684e-05, "loss": 0.0002, "step": 32326 }, { "epoch": 0.63, "learning_rate": 3.954609268352645e-05, "loss": 0.0, "step": 32328 }, { "epoch": 0.63, "learning_rate": 3.954544572326922e-05, "loss": 0.0, "step": 32330 }, { "epoch": 0.63, "learning_rate": 3.954479876301199e-05, "loss": 0.0079, "step": 32332 }, { "epoch": 0.63, "learning_rate": 3.954415180275476e-05, "loss": 0.0, "step": 32334 }, { "epoch": 0.63, "learning_rate": 3.954350484249752e-05, "loss": 0.0001, "step": 32336 }, { "epoch": 0.63, "learning_rate": 3.95428578822403e-05, "loss": 0.0, "step": 32338 }, { "epoch": 0.63, "learning_rate": 3.954221092198307e-05, "loss": 0.0002, "step": 32340 }, { "epoch": 0.63, "learning_rate": 3.954156396172583e-05, "loss": 0.007, "step": 32342 }, { "epoch": 0.63, "learning_rate": 3.9540917001468605e-05, "loss": 0.0067, "step": 32344 }, { "epoch": 0.63, "learning_rate": 3.954027004121137e-05, "loss": 0.0001, "step": 32346 }, { "epoch": 0.63, "learning_rate": 3.9539623080954144e-05, "loss": 0.0002, "step": 32348 }, { "epoch": 0.63, "learning_rate": 3.9538976120696906e-05, "loss": 0.0001, "step": 32350 }, { "epoch": 0.63, "learning_rate": 3.9538329160439675e-05, "loss": 0.0167, "step": 32352 }, { "epoch": 0.63, "learning_rate": 3.9537682200182444e-05, "loss": 0.0064, "step": 32354 }, { "epoch": 0.63, "learning_rate": 3.953703523992521e-05, "loss": 0.0004, "step": 32356 }, { "epoch": 0.63, "learning_rate": 3.953638827966798e-05, "loss": 0.0069, "step": 32358 }, { "epoch": 0.63, "learning_rate": 3.953574131941075e-05, "loss": 0.0014, "step": 32360 }, { "epoch": 0.63, "learning_rate": 3.953509435915352e-05, "loss": 0.007, "step": 32362 }, { "epoch": 0.63, "learning_rate": 3.953444739889629e-05, "loss": 0.026, "step": 32364 }, { "epoch": 0.63, "learning_rate": 3.953380043863906e-05, "loss": 0.0001, "step": 32366 }, { "epoch": 0.63, "learning_rate": 3.953315347838182e-05, "loss": 0.017, "step": 32368 }, { "epoch": 0.63, "learning_rate": 3.95325065181246e-05, "loss": 0.0001, "step": 32370 }, { "epoch": 0.63, "learning_rate": 3.953185955786736e-05, "loss": 0.0003, "step": 32372 }, { "epoch": 0.63, "learning_rate": 3.953121259761013e-05, "loss": 0.0011, "step": 32374 }, { "epoch": 0.63, "learning_rate": 3.9530565637352904e-05, "loss": 0.0, "step": 32376 }, { "epoch": 0.63, "learning_rate": 3.9529918677095666e-05, "loss": 0.0, "step": 32378 }, { "epoch": 0.63, "learning_rate": 3.9529271716838435e-05, "loss": 0.0, "step": 32380 }, { "epoch": 0.63, "learning_rate": 3.9528624756581204e-05, "loss": 0.0002, "step": 32382 }, { "epoch": 0.63, "learning_rate": 3.9527977796323974e-05, "loss": 0.0003, "step": 32384 }, { "epoch": 0.63, "learning_rate": 3.952733083606674e-05, "loss": 0.0039, "step": 32386 }, { "epoch": 0.63, "learning_rate": 3.952668387580951e-05, "loss": 0.0, "step": 32388 }, { "epoch": 0.63, "learning_rate": 3.952603691555228e-05, "loss": 0.0, "step": 32390 }, { "epoch": 0.63, "learning_rate": 3.952538995529505e-05, "loss": 0.0005, "step": 32392 }, { "epoch": 0.63, "learning_rate": 3.952474299503782e-05, "loss": 0.0002, "step": 32394 }, { "epoch": 0.63, "learning_rate": 3.952409603478058e-05, "loss": 0.0018, "step": 32396 }, { "epoch": 0.63, "learning_rate": 3.952344907452336e-05, "loss": 0.0001, "step": 32398 }, { "epoch": 0.63, "learning_rate": 3.952280211426612e-05, "loss": 0.0, "step": 32400 }, { "epoch": 0.63, "learning_rate": 3.952215515400889e-05, "loss": 0.0, "step": 32402 }, { "epoch": 0.63, "learning_rate": 3.952150819375166e-05, "loss": 0.0, "step": 32404 }, { "epoch": 0.63, "learning_rate": 3.952086123349443e-05, "loss": 0.0001, "step": 32406 }, { "epoch": 0.63, "learning_rate": 3.95202142732372e-05, "loss": 0.0, "step": 32408 }, { "epoch": 0.63, "learning_rate": 3.9519567312979965e-05, "loss": 0.0001, "step": 32410 }, { "epoch": 0.63, "learning_rate": 3.9518920352722734e-05, "loss": 0.0, "step": 32412 }, { "epoch": 0.63, "learning_rate": 3.95182733924655e-05, "loss": 0.0005, "step": 32414 }, { "epoch": 0.63, "learning_rate": 3.951762643220827e-05, "loss": 0.0085, "step": 32416 }, { "epoch": 0.63, "learning_rate": 3.951697947195104e-05, "loss": 0.0, "step": 32418 }, { "epoch": 0.63, "learning_rate": 3.951633251169381e-05, "loss": 0.0, "step": 32420 }, { "epoch": 0.63, "learning_rate": 3.951568555143658e-05, "loss": 0.0, "step": 32422 }, { "epoch": 0.63, "learning_rate": 3.951503859117934e-05, "loss": 0.0024, "step": 32424 }, { "epoch": 0.63, "learning_rate": 3.951439163092212e-05, "loss": 0.0003, "step": 32426 }, { "epoch": 0.63, "learning_rate": 3.951374467066488e-05, "loss": 0.0001, "step": 32428 }, { "epoch": 0.63, "learning_rate": 3.9513097710407656e-05, "loss": 0.0001, "step": 32430 }, { "epoch": 0.63, "learning_rate": 3.951245075015042e-05, "loss": 0.0001, "step": 32432 }, { "epoch": 0.63, "learning_rate": 3.951180378989319e-05, "loss": 0.0, "step": 32434 }, { "epoch": 0.63, "learning_rate": 3.9511156829635956e-05, "loss": 0.0013, "step": 32436 }, { "epoch": 0.63, "learning_rate": 3.9510509869378725e-05, "loss": 0.0001, "step": 32438 }, { "epoch": 0.63, "learning_rate": 3.9509862909121494e-05, "loss": 0.0089, "step": 32440 }, { "epoch": 0.63, "learning_rate": 3.9509215948864263e-05, "loss": 0.0, "step": 32442 }, { "epoch": 0.63, "learning_rate": 3.950856898860703e-05, "loss": 0.0, "step": 32444 }, { "epoch": 0.63, "learning_rate": 3.95079220283498e-05, "loss": 0.0001, "step": 32446 }, { "epoch": 0.63, "learning_rate": 3.950727506809257e-05, "loss": 0.0009, "step": 32448 }, { "epoch": 0.63, "learning_rate": 3.950662810783533e-05, "loss": 0.0002, "step": 32450 }, { "epoch": 0.63, "learning_rate": 3.950598114757811e-05, "loss": 0.0002, "step": 32452 }, { "epoch": 0.63, "learning_rate": 3.950533418732088e-05, "loss": 0.005, "step": 32454 }, { "epoch": 0.63, "learning_rate": 3.950468722706364e-05, "loss": 0.0039, "step": 32456 }, { "epoch": 0.63, "learning_rate": 3.9504040266806416e-05, "loss": 0.0005, "step": 32458 }, { "epoch": 0.63, "learning_rate": 3.950339330654918e-05, "loss": 0.0, "step": 32460 }, { "epoch": 0.63, "learning_rate": 3.950274634629195e-05, "loss": 0.0001, "step": 32462 }, { "epoch": 0.63, "learning_rate": 3.9502099386034717e-05, "loss": 0.0014, "step": 32464 }, { "epoch": 0.63, "learning_rate": 3.9501452425777486e-05, "loss": 0.001, "step": 32466 }, { "epoch": 0.63, "learning_rate": 3.9500805465520255e-05, "loss": 0.0001, "step": 32468 }, { "epoch": 0.63, "learning_rate": 3.9500158505263024e-05, "loss": 0.0001, "step": 32470 }, { "epoch": 0.63, "learning_rate": 3.949951154500579e-05, "loss": 0.0001, "step": 32472 }, { "epoch": 0.63, "learning_rate": 3.949886458474856e-05, "loss": 0.0, "step": 32474 }, { "epoch": 0.63, "learning_rate": 3.949821762449133e-05, "loss": 0.0001, "step": 32476 }, { "epoch": 0.63, "learning_rate": 3.949757066423409e-05, "loss": 0.0007, "step": 32478 }, { "epoch": 0.63, "learning_rate": 3.949692370397687e-05, "loss": 0.0, "step": 32480 }, { "epoch": 0.63, "learning_rate": 3.949627674371963e-05, "loss": 0.0118, "step": 32482 }, { "epoch": 0.63, "learning_rate": 3.94956297834624e-05, "loss": 0.0007, "step": 32484 }, { "epoch": 0.63, "learning_rate": 3.9494982823205176e-05, "loss": 0.0025, "step": 32486 }, { "epoch": 0.63, "learning_rate": 3.949433586294794e-05, "loss": 0.0078, "step": 32488 }, { "epoch": 0.63, "learning_rate": 3.9493688902690715e-05, "loss": 0.0013, "step": 32490 }, { "epoch": 0.63, "learning_rate": 3.949304194243348e-05, "loss": 0.0001, "step": 32492 }, { "epoch": 0.63, "learning_rate": 3.9492394982176246e-05, "loss": 0.0012, "step": 32494 }, { "epoch": 0.63, "learning_rate": 3.9491748021919015e-05, "loss": 0.0, "step": 32496 }, { "epoch": 0.63, "learning_rate": 3.9491101061661784e-05, "loss": 0.0, "step": 32498 }, { "epoch": 0.63, "learning_rate": 3.949045410140455e-05, "loss": 0.0055, "step": 32500 }, { "epoch": 0.63, "learning_rate": 3.948980714114732e-05, "loss": 0.0001, "step": 32502 }, { "epoch": 0.63, "learning_rate": 3.948916018089009e-05, "loss": 0.0, "step": 32504 }, { "epoch": 0.63, "learning_rate": 3.948851322063286e-05, "loss": 0.0006, "step": 32506 }, { "epoch": 0.63, "learning_rate": 3.948786626037563e-05, "loss": 0.0006, "step": 32508 }, { "epoch": 0.63, "learning_rate": 3.948721930011839e-05, "loss": 0.0, "step": 32510 }, { "epoch": 0.63, "learning_rate": 3.948657233986117e-05, "loss": 0.0022, "step": 32512 }, { "epoch": 0.63, "learning_rate": 3.948592537960393e-05, "loss": 0.0043, "step": 32514 }, { "epoch": 0.63, "learning_rate": 3.94852784193467e-05, "loss": 0.0002, "step": 32516 }, { "epoch": 0.63, "learning_rate": 3.9484631459089475e-05, "loss": 0.0, "step": 32518 }, { "epoch": 0.63, "learning_rate": 3.948398449883224e-05, "loss": 0.0003, "step": 32520 }, { "epoch": 0.63, "learning_rate": 3.9483337538575006e-05, "loss": 0.0065, "step": 32522 }, { "epoch": 0.63, "learning_rate": 3.9482690578317776e-05, "loss": 0.0019, "step": 32524 }, { "epoch": 0.63, "learning_rate": 3.9482043618060545e-05, "loss": 0.0051, "step": 32526 }, { "epoch": 0.63, "learning_rate": 3.9481396657803314e-05, "loss": 0.0, "step": 32528 }, { "epoch": 0.63, "learning_rate": 3.948074969754608e-05, "loss": 0.0, "step": 32530 }, { "epoch": 0.63, "learning_rate": 3.948010273728885e-05, "loss": 0.001, "step": 32532 }, { "epoch": 0.63, "learning_rate": 3.947945577703162e-05, "loss": 0.0001, "step": 32534 }, { "epoch": 0.63, "learning_rate": 3.947880881677439e-05, "loss": 0.0, "step": 32536 }, { "epoch": 0.63, "learning_rate": 3.947816185651715e-05, "loss": 0.0, "step": 32538 }, { "epoch": 0.63, "learning_rate": 3.947751489625993e-05, "loss": 0.0046, "step": 32540 }, { "epoch": 0.63, "learning_rate": 3.947686793600269e-05, "loss": 0.009, "step": 32542 }, { "epoch": 0.63, "learning_rate": 3.947622097574546e-05, "loss": 0.0, "step": 32544 }, { "epoch": 0.63, "learning_rate": 3.947557401548823e-05, "loss": 0.0001, "step": 32546 }, { "epoch": 0.63, "learning_rate": 3.9474927055231e-05, "loss": 0.0, "step": 32548 }, { "epoch": 0.63, "learning_rate": 3.947428009497377e-05, "loss": 0.0134, "step": 32550 }, { "epoch": 0.63, "learning_rate": 3.9473633134716536e-05, "loss": 0.0066, "step": 32552 }, { "epoch": 0.63, "learning_rate": 3.9472986174459305e-05, "loss": 0.0046, "step": 32554 }, { "epoch": 0.63, "learning_rate": 3.9472339214202074e-05, "loss": 0.0002, "step": 32556 }, { "epoch": 0.63, "learning_rate": 3.947169225394484e-05, "loss": 0.0, "step": 32558 }, { "epoch": 0.63, "learning_rate": 3.9471045293687605e-05, "loss": 0.0002, "step": 32560 }, { "epoch": 0.63, "learning_rate": 3.947039833343038e-05, "loss": 0.0, "step": 32562 }, { "epoch": 0.63, "learning_rate": 3.946975137317315e-05, "loss": 0.1007, "step": 32564 }, { "epoch": 0.63, "learning_rate": 3.946910441291591e-05, "loss": 0.0, "step": 32566 }, { "epoch": 0.63, "learning_rate": 3.946845745265869e-05, "loss": 0.0, "step": 32568 }, { "epoch": 0.63, "learning_rate": 3.946781049240145e-05, "loss": 0.0046, "step": 32570 }, { "epoch": 0.63, "learning_rate": 3.946716353214423e-05, "loss": 0.0122, "step": 32572 }, { "epoch": 0.63, "learning_rate": 3.946651657188699e-05, "loss": 0.0001, "step": 32574 }, { "epoch": 0.63, "learning_rate": 3.946586961162976e-05, "loss": 0.0, "step": 32576 }, { "epoch": 0.63, "learning_rate": 3.946522265137253e-05, "loss": 0.0011, "step": 32578 }, { "epoch": 0.63, "learning_rate": 3.9464575691115296e-05, "loss": 0.0002, "step": 32580 }, { "epoch": 0.63, "learning_rate": 3.9463928730858065e-05, "loss": 0.0001, "step": 32582 }, { "epoch": 0.63, "learning_rate": 3.9463281770600835e-05, "loss": 0.0001, "step": 32584 }, { "epoch": 0.63, "learning_rate": 3.9462634810343604e-05, "loss": 0.0003, "step": 32586 }, { "epoch": 0.63, "learning_rate": 3.946198785008637e-05, "loss": 0.0002, "step": 32588 }, { "epoch": 0.63, "learning_rate": 3.946134088982914e-05, "loss": 0.0011, "step": 32590 }, { "epoch": 0.63, "learning_rate": 3.9460693929571904e-05, "loss": 0.0002, "step": 32592 }, { "epoch": 0.63, "learning_rate": 3.946004696931468e-05, "loss": 0.0001, "step": 32594 }, { "epoch": 0.63, "learning_rate": 3.945940000905744e-05, "loss": 0.0001, "step": 32596 }, { "epoch": 0.63, "learning_rate": 3.945875304880021e-05, "loss": 0.0001, "step": 32598 }, { "epoch": 0.63, "learning_rate": 3.945810608854299e-05, "loss": 0.0, "step": 32600 }, { "epoch": 0.63, "learning_rate": 3.945745912828575e-05, "loss": 0.0, "step": 32602 }, { "epoch": 0.63, "learning_rate": 3.945681216802852e-05, "loss": 0.0002, "step": 32604 }, { "epoch": 0.63, "learning_rate": 3.945616520777129e-05, "loss": 0.0, "step": 32606 }, { "epoch": 0.63, "learning_rate": 3.945551824751406e-05, "loss": 0.0, "step": 32608 }, { "epoch": 0.63, "learning_rate": 3.9454871287256826e-05, "loss": 0.0002, "step": 32610 }, { "epoch": 0.63, "learning_rate": 3.9454224326999595e-05, "loss": 0.0, "step": 32612 }, { "epoch": 0.63, "learning_rate": 3.9453577366742364e-05, "loss": 0.0088, "step": 32614 }, { "epoch": 0.63, "learning_rate": 3.945293040648513e-05, "loss": 0.0007, "step": 32616 }, { "epoch": 0.63, "learning_rate": 3.94522834462279e-05, "loss": 0.0006, "step": 32618 }, { "epoch": 0.63, "learning_rate": 3.9451636485970664e-05, "loss": 0.0, "step": 32620 }, { "epoch": 0.63, "learning_rate": 3.945098952571344e-05, "loss": 0.0167, "step": 32622 }, { "epoch": 0.63, "learning_rate": 3.94503425654562e-05, "loss": 0.0031, "step": 32624 }, { "epoch": 0.63, "learning_rate": 3.944969560519897e-05, "loss": 0.0063, "step": 32626 }, { "epoch": 0.63, "learning_rate": 3.944904864494174e-05, "loss": 0.0002, "step": 32628 }, { "epoch": 0.63, "learning_rate": 3.944840168468451e-05, "loss": 0.0, "step": 32630 }, { "epoch": 0.63, "learning_rate": 3.9447754724427286e-05, "loss": 0.0, "step": 32632 }, { "epoch": 0.63, "learning_rate": 3.944710776417005e-05, "loss": 0.0001, "step": 32634 }, { "epoch": 0.63, "learning_rate": 3.944646080391282e-05, "loss": 0.0, "step": 32636 }, { "epoch": 0.63, "learning_rate": 3.9445813843655586e-05, "loss": 0.0021, "step": 32638 }, { "epoch": 0.63, "learning_rate": 3.9445166883398355e-05, "loss": 0.0, "step": 32640 }, { "epoch": 0.63, "learning_rate": 3.9444519923141124e-05, "loss": 0.0031, "step": 32642 }, { "epoch": 0.63, "learning_rate": 3.9443872962883893e-05, "loss": 0.0005, "step": 32644 }, { "epoch": 0.63, "learning_rate": 3.944322600262666e-05, "loss": 0.0008, "step": 32646 }, { "epoch": 0.63, "learning_rate": 3.944257904236943e-05, "loss": 0.0001, "step": 32648 }, { "epoch": 0.63, "learning_rate": 3.94419320821122e-05, "loss": 0.0005, "step": 32650 }, { "epoch": 0.63, "learning_rate": 3.944128512185496e-05, "loss": 0.0013, "step": 32652 }, { "epoch": 0.63, "learning_rate": 3.944063816159774e-05, "loss": 0.0, "step": 32654 }, { "epoch": 0.63, "learning_rate": 3.94399912013405e-05, "loss": 0.0001, "step": 32656 }, { "epoch": 0.63, "learning_rate": 3.943934424108327e-05, "loss": 0.0001, "step": 32658 }, { "epoch": 0.63, "learning_rate": 3.943869728082604e-05, "loss": 0.0, "step": 32660 }, { "epoch": 0.63, "learning_rate": 3.943805032056881e-05, "loss": 0.0007, "step": 32662 }, { "epoch": 0.63, "learning_rate": 3.943740336031158e-05, "loss": 0.0004, "step": 32664 }, { "epoch": 0.63, "learning_rate": 3.943675640005435e-05, "loss": 0.0, "step": 32666 }, { "epoch": 0.63, "learning_rate": 3.9436109439797116e-05, "loss": 0.0027, "step": 32668 }, { "epoch": 0.63, "learning_rate": 3.9435462479539885e-05, "loss": 0.0035, "step": 32670 }, { "epoch": 0.63, "learning_rate": 3.9434815519282654e-05, "loss": 0.0029, "step": 32672 }, { "epoch": 0.63, "learning_rate": 3.9434168559025416e-05, "loss": 0.0, "step": 32674 }, { "epoch": 0.63, "learning_rate": 3.943352159876819e-05, "loss": 0.0001, "step": 32676 }, { "epoch": 0.63, "learning_rate": 3.943287463851096e-05, "loss": 0.0009, "step": 32678 }, { "epoch": 0.63, "learning_rate": 3.9432227678253723e-05, "loss": 0.0159, "step": 32680 }, { "epoch": 0.63, "learning_rate": 3.94315807179965e-05, "loss": 0.0012, "step": 32682 }, { "epoch": 0.63, "learning_rate": 3.943093375773926e-05, "loss": 0.0065, "step": 32684 }, { "epoch": 0.63, "learning_rate": 3.943028679748203e-05, "loss": 0.0062, "step": 32686 }, { "epoch": 0.63, "learning_rate": 3.94296398372248e-05, "loss": 0.0001, "step": 32688 }, { "epoch": 0.63, "learning_rate": 3.942899287696757e-05, "loss": 0.0074, "step": 32690 }, { "epoch": 0.63, "learning_rate": 3.942834591671034e-05, "loss": 0.0, "step": 32692 }, { "epoch": 0.63, "learning_rate": 3.942769895645311e-05, "loss": 0.0048, "step": 32694 }, { "epoch": 0.63, "learning_rate": 3.9427051996195876e-05, "loss": 0.0025, "step": 32696 }, { "epoch": 0.63, "learning_rate": 3.9426405035938645e-05, "loss": 0.0003, "step": 32698 }, { "epoch": 0.63, "learning_rate": 3.9425758075681414e-05, "loss": 0.0002, "step": 32700 }, { "epoch": 0.63, "learning_rate": 3.9425111115424177e-05, "loss": 0.0004, "step": 32702 }, { "epoch": 0.63, "learning_rate": 3.942446415516695e-05, "loss": 0.0001, "step": 32704 }, { "epoch": 0.63, "learning_rate": 3.9423817194909715e-05, "loss": 0.0003, "step": 32706 }, { "epoch": 0.63, "learning_rate": 3.9423170234652484e-05, "loss": 0.0012, "step": 32708 }, { "epoch": 0.63, "learning_rate": 3.942252327439526e-05, "loss": 0.0229, "step": 32710 }, { "epoch": 0.63, "learning_rate": 3.942187631413802e-05, "loss": 0.0016, "step": 32712 }, { "epoch": 0.63, "learning_rate": 3.94212293538808e-05, "loss": 0.0, "step": 32714 }, { "epoch": 0.63, "learning_rate": 3.942058239362356e-05, "loss": 0.0007, "step": 32716 }, { "epoch": 0.64, "learning_rate": 3.941993543336633e-05, "loss": 0.0002, "step": 32718 }, { "epoch": 0.64, "learning_rate": 3.94192884731091e-05, "loss": 0.001, "step": 32720 }, { "epoch": 0.64, "learning_rate": 3.941864151285187e-05, "loss": 0.0, "step": 32722 }, { "epoch": 0.64, "learning_rate": 3.9417994552594637e-05, "loss": 0.0056, "step": 32724 }, { "epoch": 0.64, "learning_rate": 3.9417347592337406e-05, "loss": 0.0005, "step": 32726 }, { "epoch": 0.64, "learning_rate": 3.9416700632080175e-05, "loss": 0.0, "step": 32728 }, { "epoch": 0.64, "learning_rate": 3.9416053671822944e-05, "loss": 0.0077, "step": 32730 }, { "epoch": 0.64, "learning_rate": 3.941540671156571e-05, "loss": 0.0, "step": 32732 }, { "epoch": 0.64, "learning_rate": 3.9414759751308475e-05, "loss": 0.0011, "step": 32734 }, { "epoch": 0.64, "learning_rate": 3.941411279105125e-05, "loss": 0.0047, "step": 32736 }, { "epoch": 0.64, "learning_rate": 3.941346583079401e-05, "loss": 0.0004, "step": 32738 }, { "epoch": 0.64, "learning_rate": 3.941281887053678e-05, "loss": 0.0065, "step": 32740 }, { "epoch": 0.64, "learning_rate": 3.941217191027956e-05, "loss": 0.0, "step": 32742 }, { "epoch": 0.64, "learning_rate": 3.941152495002232e-05, "loss": 0.0002, "step": 32744 }, { "epoch": 0.64, "learning_rate": 3.941087798976509e-05, "loss": 0.0, "step": 32746 }, { "epoch": 0.64, "learning_rate": 3.941023102950786e-05, "loss": 0.0001, "step": 32748 }, { "epoch": 0.64, "learning_rate": 3.940958406925063e-05, "loss": 0.008, "step": 32750 }, { "epoch": 0.64, "learning_rate": 3.94089371089934e-05, "loss": 0.0, "step": 32752 }, { "epoch": 0.64, "learning_rate": 3.9408290148736166e-05, "loss": 0.0001, "step": 32754 }, { "epoch": 0.64, "learning_rate": 3.9407643188478935e-05, "loss": 0.0, "step": 32756 }, { "epoch": 0.64, "learning_rate": 3.9406996228221704e-05, "loss": 0.0002, "step": 32758 }, { "epoch": 0.64, "learning_rate": 3.940634926796447e-05, "loss": 0.0026, "step": 32760 }, { "epoch": 0.64, "learning_rate": 3.9405702307707236e-05, "loss": 0.0027, "step": 32762 }, { "epoch": 0.64, "learning_rate": 3.940505534745001e-05, "loss": 0.0, "step": 32764 }, { "epoch": 0.64, "learning_rate": 3.9404408387192774e-05, "loss": 0.0103, "step": 32766 }, { "epoch": 0.64, "learning_rate": 3.940376142693554e-05, "loss": 0.0001, "step": 32768 }, { "epoch": 0.64, "learning_rate": 3.940311446667831e-05, "loss": 0.0002, "step": 32770 }, { "epoch": 0.64, "learning_rate": 3.940246750642108e-05, "loss": 0.0001, "step": 32772 }, { "epoch": 0.64, "learning_rate": 3.940182054616385e-05, "loss": 0.0002, "step": 32774 }, { "epoch": 0.64, "learning_rate": 3.940117358590662e-05, "loss": 0.0, "step": 32776 }, { "epoch": 0.64, "learning_rate": 3.940052662564939e-05, "loss": 0.0085, "step": 32778 }, { "epoch": 0.64, "learning_rate": 3.939987966539216e-05, "loss": 0.0119, "step": 32780 }, { "epoch": 0.64, "learning_rate": 3.9399232705134926e-05, "loss": 0.0, "step": 32782 }, { "epoch": 0.64, "learning_rate": 3.939858574487769e-05, "loss": 0.0002, "step": 32784 }, { "epoch": 0.64, "learning_rate": 3.9397938784620465e-05, "loss": 0.0001, "step": 32786 }, { "epoch": 0.64, "learning_rate": 3.9397291824363234e-05, "loss": 0.0022, "step": 32788 }, { "epoch": 0.64, "learning_rate": 3.9396644864105996e-05, "loss": 0.0035, "step": 32790 }, { "epoch": 0.64, "learning_rate": 3.939599790384877e-05, "loss": 0.0003, "step": 32792 }, { "epoch": 0.64, "learning_rate": 3.9395350943591534e-05, "loss": 0.0, "step": 32794 }, { "epoch": 0.64, "learning_rate": 3.939470398333431e-05, "loss": 0.0001, "step": 32796 }, { "epoch": 0.64, "learning_rate": 3.939405702307707e-05, "loss": 0.0, "step": 32798 }, { "epoch": 0.64, "learning_rate": 3.939341006281984e-05, "loss": 0.0005, "step": 32800 }, { "epoch": 0.64, "learning_rate": 3.939276310256261e-05, "loss": 0.0002, "step": 32802 }, { "epoch": 0.64, "learning_rate": 3.939211614230538e-05, "loss": 0.0001, "step": 32804 }, { "epoch": 0.64, "learning_rate": 3.939146918204815e-05, "loss": 0.0, "step": 32806 }, { "epoch": 0.64, "learning_rate": 3.939082222179092e-05, "loss": 0.0, "step": 32808 }, { "epoch": 0.64, "learning_rate": 3.939017526153369e-05, "loss": 0.0, "step": 32810 }, { "epoch": 0.64, "learning_rate": 3.9389528301276456e-05, "loss": 0.0, "step": 32812 }, { "epoch": 0.64, "learning_rate": 3.9388881341019225e-05, "loss": 0.0002, "step": 32814 }, { "epoch": 0.64, "learning_rate": 3.938823438076199e-05, "loss": 0.0097, "step": 32816 }, { "epoch": 0.64, "learning_rate": 3.938758742050476e-05, "loss": 0.0024, "step": 32818 }, { "epoch": 0.64, "learning_rate": 3.938694046024753e-05, "loss": 0.0, "step": 32820 }, { "epoch": 0.64, "learning_rate": 3.9386293499990295e-05, "loss": 0.0, "step": 32822 }, { "epoch": 0.64, "learning_rate": 3.938564653973307e-05, "loss": 0.0, "step": 32824 }, { "epoch": 0.64, "learning_rate": 3.938499957947583e-05, "loss": 0.0004, "step": 32826 }, { "epoch": 0.64, "learning_rate": 3.93843526192186e-05, "loss": 0.0, "step": 32828 }, { "epoch": 0.64, "learning_rate": 3.938370565896137e-05, "loss": 0.0001, "step": 32830 }, { "epoch": 0.64, "learning_rate": 3.938305869870414e-05, "loss": 0.003, "step": 32832 }, { "epoch": 0.64, "learning_rate": 3.938241173844691e-05, "loss": 0.0001, "step": 32834 }, { "epoch": 0.64, "learning_rate": 3.938176477818968e-05, "loss": 0.0105, "step": 32836 }, { "epoch": 0.64, "learning_rate": 3.938111781793245e-05, "loss": 0.0, "step": 32838 }, { "epoch": 0.64, "learning_rate": 3.9380470857675216e-05, "loss": 0.0046, "step": 32840 }, { "epoch": 0.64, "learning_rate": 3.9379823897417985e-05, "loss": 0.0014, "step": 32842 }, { "epoch": 0.64, "learning_rate": 3.937917693716075e-05, "loss": 0.0, "step": 32844 }, { "epoch": 0.64, "learning_rate": 3.9378529976903524e-05, "loss": 0.0, "step": 32846 }, { "epoch": 0.64, "learning_rate": 3.9377883016646286e-05, "loss": 0.0, "step": 32848 }, { "epoch": 0.64, "learning_rate": 3.9377236056389055e-05, "loss": 0.0, "step": 32850 }, { "epoch": 0.64, "learning_rate": 3.9376589096131824e-05, "loss": 0.0001, "step": 32852 }, { "epoch": 0.64, "learning_rate": 3.937594213587459e-05, "loss": 0.0, "step": 32854 }, { "epoch": 0.64, "learning_rate": 3.937529517561737e-05, "loss": 0.0016, "step": 32856 }, { "epoch": 0.64, "learning_rate": 3.937464821536013e-05, "loss": 0.0074, "step": 32858 }, { "epoch": 0.64, "learning_rate": 3.93740012551029e-05, "loss": 0.0001, "step": 32860 }, { "epoch": 0.64, "learning_rate": 3.937335429484567e-05, "loss": 0.0, "step": 32862 }, { "epoch": 0.64, "learning_rate": 3.937270733458844e-05, "loss": 0.0074, "step": 32864 }, { "epoch": 0.64, "learning_rate": 3.937206037433121e-05, "loss": 0.0083, "step": 32866 }, { "epoch": 0.64, "learning_rate": 3.937141341407398e-05, "loss": 0.0002, "step": 32868 }, { "epoch": 0.64, "learning_rate": 3.9370766453816746e-05, "loss": 0.0001, "step": 32870 }, { "epoch": 0.64, "learning_rate": 3.9370119493559515e-05, "loss": 0.0004, "step": 32872 }, { "epoch": 0.64, "learning_rate": 3.9369472533302284e-05, "loss": 0.0007, "step": 32874 }, { "epoch": 0.64, "learning_rate": 3.9368825573045046e-05, "loss": 0.0004, "step": 32876 }, { "epoch": 0.64, "learning_rate": 3.936817861278782e-05, "loss": 0.0, "step": 32878 }, { "epoch": 0.64, "learning_rate": 3.9367531652530584e-05, "loss": 0.0005, "step": 32880 }, { "epoch": 0.64, "learning_rate": 3.9366884692273354e-05, "loss": 0.001, "step": 32882 }, { "epoch": 0.64, "learning_rate": 3.936623773201612e-05, "loss": 0.0002, "step": 32884 }, { "epoch": 0.64, "learning_rate": 3.936559077175889e-05, "loss": 0.0091, "step": 32886 }, { "epoch": 0.64, "learning_rate": 3.936494381150166e-05, "loss": 0.0, "step": 32888 }, { "epoch": 0.64, "learning_rate": 3.936429685124443e-05, "loss": 0.0066, "step": 32890 }, { "epoch": 0.64, "learning_rate": 3.93636498909872e-05, "loss": 0.0, "step": 32892 }, { "epoch": 0.64, "learning_rate": 3.936300293072997e-05, "loss": 0.0034, "step": 32894 }, { "epoch": 0.64, "learning_rate": 3.936235597047274e-05, "loss": 0.0019, "step": 32896 }, { "epoch": 0.64, "learning_rate": 3.9361709010215506e-05, "loss": 0.0, "step": 32898 }, { "epoch": 0.64, "learning_rate": 3.9361062049958275e-05, "loss": 0.0007, "step": 32900 }, { "epoch": 0.64, "learning_rate": 3.9360415089701044e-05, "loss": 0.0, "step": 32902 }, { "epoch": 0.64, "learning_rate": 3.935976812944381e-05, "loss": 0.0006, "step": 32904 }, { "epoch": 0.64, "learning_rate": 3.935912116918658e-05, "loss": 0.0001, "step": 32906 }, { "epoch": 0.64, "learning_rate": 3.9358474208929345e-05, "loss": 0.0016, "step": 32908 }, { "epoch": 0.64, "learning_rate": 3.9357827248672114e-05, "loss": 0.0001, "step": 32910 }, { "epoch": 0.64, "learning_rate": 3.935718028841488e-05, "loss": 0.0009, "step": 32912 }, { "epoch": 0.64, "learning_rate": 3.935653332815765e-05, "loss": 0.0079, "step": 32914 }, { "epoch": 0.64, "learning_rate": 3.935588636790042e-05, "loss": 0.0003, "step": 32916 }, { "epoch": 0.64, "learning_rate": 3.935523940764319e-05, "loss": 0.0042, "step": 32918 }, { "epoch": 0.64, "learning_rate": 3.935459244738596e-05, "loss": 0.0, "step": 32920 }, { "epoch": 0.64, "learning_rate": 3.935394548712873e-05, "loss": 0.0, "step": 32922 }, { "epoch": 0.64, "learning_rate": 3.93532985268715e-05, "loss": 0.0063, "step": 32924 }, { "epoch": 0.64, "learning_rate": 3.935265156661426e-05, "loss": 0.0093, "step": 32926 }, { "epoch": 0.64, "learning_rate": 3.9352004606357036e-05, "loss": 0.0, "step": 32928 }, { "epoch": 0.64, "learning_rate": 3.93513576460998e-05, "loss": 0.0169, "step": 32930 }, { "epoch": 0.64, "learning_rate": 3.935071068584257e-05, "loss": 0.0, "step": 32932 }, { "epoch": 0.64, "learning_rate": 3.935006372558534e-05, "loss": 0.0001, "step": 32934 }, { "epoch": 0.64, "learning_rate": 3.9349416765328105e-05, "loss": 0.0133, "step": 32936 }, { "epoch": 0.64, "learning_rate": 3.934876980507088e-05, "loss": 0.0003, "step": 32938 }, { "epoch": 0.64, "learning_rate": 3.9348122844813643e-05, "loss": 0.0008, "step": 32940 }, { "epoch": 0.64, "learning_rate": 3.934747588455641e-05, "loss": 0.0, "step": 32942 }, { "epoch": 0.64, "learning_rate": 3.934682892429918e-05, "loss": 0.0055, "step": 32944 }, { "epoch": 0.64, "learning_rate": 3.934618196404195e-05, "loss": 0.0002, "step": 32946 }, { "epoch": 0.64, "learning_rate": 3.934553500378472e-05, "loss": 0.0001, "step": 32948 }, { "epoch": 0.64, "learning_rate": 3.934488804352749e-05, "loss": 0.0033, "step": 32950 }, { "epoch": 0.64, "learning_rate": 3.934424108327026e-05, "loss": 0.0, "step": 32952 }, { "epoch": 0.64, "learning_rate": 3.934359412301303e-05, "loss": 0.0094, "step": 32954 }, { "epoch": 0.64, "learning_rate": 3.9342947162755796e-05, "loss": 0.0002, "step": 32956 }, { "epoch": 0.64, "learning_rate": 3.934230020249856e-05, "loss": 0.0, "step": 32958 }, { "epoch": 0.64, "learning_rate": 3.9341653242241334e-05, "loss": 0.0081, "step": 32960 }, { "epoch": 0.64, "learning_rate": 3.9341006281984097e-05, "loss": 0.0, "step": 32962 }, { "epoch": 0.64, "learning_rate": 3.9340359321726866e-05, "loss": 0.0033, "step": 32964 }, { "epoch": 0.64, "learning_rate": 3.933971236146964e-05, "loss": 0.0003, "step": 32966 }, { "epoch": 0.64, "learning_rate": 3.9339065401212404e-05, "loss": 0.0001, "step": 32968 }, { "epoch": 0.64, "learning_rate": 3.933841844095517e-05, "loss": 0.0003, "step": 32970 }, { "epoch": 0.64, "learning_rate": 3.933777148069794e-05, "loss": 0.0012, "step": 32972 }, { "epoch": 0.64, "learning_rate": 3.933712452044071e-05, "loss": 0.0001, "step": 32974 }, { "epoch": 0.64, "learning_rate": 3.933647756018348e-05, "loss": 0.0, "step": 32976 }, { "epoch": 0.64, "learning_rate": 3.933583059992625e-05, "loss": 0.006, "step": 32978 }, { "epoch": 0.64, "learning_rate": 3.933518363966902e-05, "loss": 0.0, "step": 32980 }, { "epoch": 0.64, "learning_rate": 3.933453667941179e-05, "loss": 0.0001, "step": 32982 }, { "epoch": 0.64, "learning_rate": 3.9333889719154557e-05, "loss": 0.0006, "step": 32984 }, { "epoch": 0.64, "learning_rate": 3.933324275889732e-05, "loss": 0.0007, "step": 32986 }, { "epoch": 0.64, "learning_rate": 3.9332595798640095e-05, "loss": 0.0, "step": 32988 }, { "epoch": 0.64, "learning_rate": 3.933194883838286e-05, "loss": 0.0, "step": 32990 }, { "epoch": 0.64, "learning_rate": 3.9331301878125626e-05, "loss": 0.0023, "step": 32992 }, { "epoch": 0.64, "learning_rate": 3.9330654917868395e-05, "loss": 0.0, "step": 32994 }, { "epoch": 0.64, "learning_rate": 3.9330007957611164e-05, "loss": 0.0, "step": 32996 }, { "epoch": 0.64, "learning_rate": 3.932936099735394e-05, "loss": 0.0, "step": 32998 }, { "epoch": 0.64, "learning_rate": 3.93287140370967e-05, "loss": 0.0002, "step": 33000 }, { "epoch": 0.64, "learning_rate": 3.932806707683947e-05, "loss": 0.0003, "step": 33002 }, { "epoch": 0.64, "learning_rate": 3.932742011658224e-05, "loss": 0.0, "step": 33004 }, { "epoch": 0.64, "learning_rate": 3.932677315632501e-05, "loss": 0.0, "step": 33006 }, { "epoch": 0.64, "learning_rate": 3.932612619606777e-05, "loss": 0.0001, "step": 33008 }, { "epoch": 0.64, "learning_rate": 3.932547923581055e-05, "loss": 0.0, "step": 33010 }, { "epoch": 0.64, "learning_rate": 3.932483227555332e-05, "loss": 0.0023, "step": 33012 }, { "epoch": 0.64, "learning_rate": 3.9324185315296086e-05, "loss": 0.0, "step": 33014 }, { "epoch": 0.64, "learning_rate": 3.9323538355038855e-05, "loss": 0.0019, "step": 33016 }, { "epoch": 0.64, "learning_rate": 3.932289139478162e-05, "loss": 0.0003, "step": 33018 }, { "epoch": 0.64, "learning_rate": 3.932224443452439e-05, "loss": 0.0001, "step": 33020 }, { "epoch": 0.64, "learning_rate": 3.9321597474267156e-05, "loss": 0.0012, "step": 33022 }, { "epoch": 0.64, "learning_rate": 3.9320950514009925e-05, "loss": 0.0, "step": 33024 }, { "epoch": 0.64, "learning_rate": 3.9320303553752694e-05, "loss": 0.0, "step": 33026 }, { "epoch": 0.64, "learning_rate": 3.931965659349546e-05, "loss": 0.0004, "step": 33028 }, { "epoch": 0.64, "learning_rate": 3.931900963323823e-05, "loss": 0.0, "step": 33030 }, { "epoch": 0.64, "learning_rate": 3.9318362672981e-05, "loss": 0.0001, "step": 33032 }, { "epoch": 0.64, "learning_rate": 3.931771571272377e-05, "loss": 0.0, "step": 33034 }, { "epoch": 0.64, "learning_rate": 3.931706875246654e-05, "loss": 0.0053, "step": 33036 }, { "epoch": 0.64, "learning_rate": 3.931642179220931e-05, "loss": 0.0036, "step": 33038 }, { "epoch": 0.64, "learning_rate": 3.931577483195207e-05, "loss": 0.0, "step": 33040 }, { "epoch": 0.64, "learning_rate": 3.9315127871694846e-05, "loss": 0.0018, "step": 33042 }, { "epoch": 0.64, "learning_rate": 3.9314480911437615e-05, "loss": 0.0001, "step": 33044 }, { "epoch": 0.64, "learning_rate": 3.931383395118038e-05, "loss": 0.0, "step": 33046 }, { "epoch": 0.64, "learning_rate": 3.9313186990923154e-05, "loss": 0.0, "step": 33048 }, { "epoch": 0.64, "learning_rate": 3.9312540030665916e-05, "loss": 0.0001, "step": 33050 }, { "epoch": 0.64, "learning_rate": 3.9311893070408685e-05, "loss": 0.0, "step": 33052 }, { "epoch": 0.64, "learning_rate": 3.9311246110151454e-05, "loss": 0.0165, "step": 33054 }, { "epoch": 0.64, "learning_rate": 3.931059914989422e-05, "loss": 0.0, "step": 33056 }, { "epoch": 0.64, "learning_rate": 3.930995218963699e-05, "loss": 0.0001, "step": 33058 }, { "epoch": 0.64, "learning_rate": 3.930930522937976e-05, "loss": 0.0098, "step": 33060 }, { "epoch": 0.64, "learning_rate": 3.930865826912253e-05, "loss": 0.0003, "step": 33062 }, { "epoch": 0.64, "learning_rate": 3.93080113088653e-05, "loss": 0.0, "step": 33064 }, { "epoch": 0.64, "learning_rate": 3.930736434860807e-05, "loss": 0.0097, "step": 33066 }, { "epoch": 0.64, "learning_rate": 3.930671738835083e-05, "loss": 0.0002, "step": 33068 }, { "epoch": 0.64, "learning_rate": 3.930607042809361e-05, "loss": 0.0046, "step": 33070 }, { "epoch": 0.64, "learning_rate": 3.930542346783637e-05, "loss": 0.0008, "step": 33072 }, { "epoch": 0.64, "learning_rate": 3.930477650757914e-05, "loss": 0.0003, "step": 33074 }, { "epoch": 0.64, "learning_rate": 3.930412954732191e-05, "loss": 0.0062, "step": 33076 }, { "epoch": 0.64, "learning_rate": 3.9303482587064676e-05, "loss": 0.0015, "step": 33078 }, { "epoch": 0.64, "learning_rate": 3.930283562680745e-05, "loss": 0.0014, "step": 33080 }, { "epoch": 0.64, "learning_rate": 3.9302188666550215e-05, "loss": 0.0, "step": 33082 }, { "epoch": 0.64, "learning_rate": 3.9301541706292984e-05, "loss": 0.0887, "step": 33084 }, { "epoch": 0.64, "learning_rate": 3.930089474603575e-05, "loss": 0.0, "step": 33086 }, { "epoch": 0.64, "learning_rate": 3.930024778577852e-05, "loss": 0.0, "step": 33088 }, { "epoch": 0.64, "learning_rate": 3.929960082552129e-05, "loss": 0.0001, "step": 33090 }, { "epoch": 0.64, "learning_rate": 3.929895386526406e-05, "loss": 0.0, "step": 33092 }, { "epoch": 0.64, "learning_rate": 3.929830690500683e-05, "loss": 0.0003, "step": 33094 }, { "epoch": 0.64, "learning_rate": 3.92976599447496e-05, "loss": 0.0061, "step": 33096 }, { "epoch": 0.64, "learning_rate": 3.929701298449237e-05, "loss": 0.0001, "step": 33098 }, { "epoch": 0.64, "learning_rate": 3.929636602423513e-05, "loss": 0.0018, "step": 33100 }, { "epoch": 0.64, "learning_rate": 3.9295719063977905e-05, "loss": 0.0003, "step": 33102 }, { "epoch": 0.64, "learning_rate": 3.929507210372067e-05, "loss": 0.0001, "step": 33104 }, { "epoch": 0.64, "learning_rate": 3.929442514346344e-05, "loss": 0.0001, "step": 33106 }, { "epoch": 0.64, "learning_rate": 3.9293778183206206e-05, "loss": 0.0, "step": 33108 }, { "epoch": 0.64, "learning_rate": 3.9293131222948975e-05, "loss": 0.0, "step": 33110 }, { "epoch": 0.64, "learning_rate": 3.9292484262691744e-05, "loss": 0.0162, "step": 33112 }, { "epoch": 0.64, "learning_rate": 3.929183730243451e-05, "loss": 0.0, "step": 33114 }, { "epoch": 0.64, "learning_rate": 3.929119034217728e-05, "loss": 0.0078, "step": 33116 }, { "epoch": 0.64, "learning_rate": 3.929054338192005e-05, "loss": 0.0107, "step": 33118 }, { "epoch": 0.64, "learning_rate": 3.928989642166282e-05, "loss": 0.0001, "step": 33120 }, { "epoch": 0.64, "learning_rate": 3.928924946140559e-05, "loss": 0.0004, "step": 33122 }, { "epoch": 0.64, "learning_rate": 3.928860250114836e-05, "loss": 0.0031, "step": 33124 }, { "epoch": 0.64, "learning_rate": 3.928795554089113e-05, "loss": 0.0002, "step": 33126 }, { "epoch": 0.64, "learning_rate": 3.928730858063389e-05, "loss": 0.0001, "step": 33128 }, { "epoch": 0.64, "learning_rate": 3.9286661620376666e-05, "loss": 0.0012, "step": 33130 }, { "epoch": 0.64, "learning_rate": 3.928601466011943e-05, "loss": 0.0001, "step": 33132 }, { "epoch": 0.64, "learning_rate": 3.92853676998622e-05, "loss": 0.0, "step": 33134 }, { "epoch": 0.64, "learning_rate": 3.9284720739604966e-05, "loss": 0.0, "step": 33136 }, { "epoch": 0.64, "learning_rate": 3.9284073779347735e-05, "loss": 0.0001, "step": 33138 }, { "epoch": 0.64, "learning_rate": 3.9283426819090504e-05, "loss": 0.0012, "step": 33140 }, { "epoch": 0.64, "learning_rate": 3.9282779858833274e-05, "loss": 0.0, "step": 33142 }, { "epoch": 0.64, "learning_rate": 3.928213289857604e-05, "loss": 0.0001, "step": 33144 }, { "epoch": 0.64, "learning_rate": 3.928148593831881e-05, "loss": 0.0004, "step": 33146 }, { "epoch": 0.64, "learning_rate": 3.928083897806158e-05, "loss": 0.0, "step": 33148 }, { "epoch": 0.64, "learning_rate": 3.928019201780434e-05, "loss": 0.0001, "step": 33150 }, { "epoch": 0.64, "learning_rate": 3.927954505754712e-05, "loss": 0.0045, "step": 33152 }, { "epoch": 0.64, "learning_rate": 3.927889809728988e-05, "loss": 0.0, "step": 33154 }, { "epoch": 0.64, "learning_rate": 3.927825113703266e-05, "loss": 0.0007, "step": 33156 }, { "epoch": 0.64, "learning_rate": 3.9277604176775426e-05, "loss": 0.0001, "step": 33158 }, { "epoch": 0.64, "learning_rate": 3.927695721651819e-05, "loss": 0.0001, "step": 33160 }, { "epoch": 0.64, "learning_rate": 3.9276310256260964e-05, "loss": 0.0011, "step": 33162 }, { "epoch": 0.64, "learning_rate": 3.927566329600373e-05, "loss": 0.0165, "step": 33164 }, { "epoch": 0.64, "learning_rate": 3.9275016335746496e-05, "loss": 0.0021, "step": 33166 }, { "epoch": 0.64, "learning_rate": 3.9274369375489265e-05, "loss": 0.0, "step": 33168 }, { "epoch": 0.64, "learning_rate": 3.9273722415232034e-05, "loss": 0.0024, "step": 33170 }, { "epoch": 0.64, "learning_rate": 3.92730754549748e-05, "loss": 0.0001, "step": 33172 }, { "epoch": 0.64, "learning_rate": 3.927242849471757e-05, "loss": 0.0001, "step": 33174 }, { "epoch": 0.64, "learning_rate": 3.927178153446034e-05, "loss": 0.0004, "step": 33176 }, { "epoch": 0.64, "learning_rate": 3.927113457420311e-05, "loss": 0.013, "step": 33178 }, { "epoch": 0.64, "learning_rate": 3.927048761394588e-05, "loss": 0.0009, "step": 33180 }, { "epoch": 0.64, "learning_rate": 3.926984065368864e-05, "loss": 0.0002, "step": 33182 }, { "epoch": 0.64, "learning_rate": 3.926919369343142e-05, "loss": 0.0, "step": 33184 }, { "epoch": 0.64, "learning_rate": 3.926854673317418e-05, "loss": 0.0, "step": 33186 }, { "epoch": 0.64, "learning_rate": 3.926789977291695e-05, "loss": 0.0001, "step": 33188 }, { "epoch": 0.64, "learning_rate": 3.9267252812659725e-05, "loss": 0.0229, "step": 33190 }, { "epoch": 0.64, "learning_rate": 3.926660585240249e-05, "loss": 0.0007, "step": 33192 }, { "epoch": 0.64, "learning_rate": 3.9265958892145256e-05, "loss": 0.0339, "step": 33194 }, { "epoch": 0.64, "learning_rate": 3.9265311931888025e-05, "loss": 0.0, "step": 33196 }, { "epoch": 0.64, "learning_rate": 3.9264664971630794e-05, "loss": 0.0001, "step": 33198 }, { "epoch": 0.64, "learning_rate": 3.9264018011373563e-05, "loss": 0.0, "step": 33200 }, { "epoch": 0.64, "learning_rate": 3.926337105111633e-05, "loss": 0.0054, "step": 33202 }, { "epoch": 0.64, "learning_rate": 3.92627240908591e-05, "loss": 0.0001, "step": 33204 }, { "epoch": 0.64, "learning_rate": 3.926207713060187e-05, "loss": 0.0153, "step": 33206 }, { "epoch": 0.64, "learning_rate": 3.926143017034464e-05, "loss": 0.0, "step": 33208 }, { "epoch": 0.64, "learning_rate": 3.92607832100874e-05, "loss": 0.0451, "step": 33210 }, { "epoch": 0.64, "learning_rate": 3.926013624983018e-05, "loss": 0.0, "step": 33212 }, { "epoch": 0.64, "learning_rate": 3.925948928957294e-05, "loss": 0.0, "step": 33214 }, { "epoch": 0.64, "learning_rate": 3.925884232931571e-05, "loss": 0.0002, "step": 33216 }, { "epoch": 0.64, "learning_rate": 3.925819536905848e-05, "loss": 0.0004, "step": 33218 }, { "epoch": 0.64, "learning_rate": 3.925754840880125e-05, "loss": 0.0001, "step": 33220 }, { "epoch": 0.64, "learning_rate": 3.925690144854402e-05, "loss": 0.0251, "step": 33222 }, { "epoch": 0.64, "learning_rate": 3.9256254488286786e-05, "loss": 0.0, "step": 33224 }, { "epoch": 0.64, "learning_rate": 3.9255607528029555e-05, "loss": 0.0001, "step": 33226 }, { "epoch": 0.64, "learning_rate": 3.9254960567772324e-05, "loss": 0.0001, "step": 33228 }, { "epoch": 0.64, "learning_rate": 3.925431360751509e-05, "loss": 0.0001, "step": 33230 }, { "epoch": 0.64, "learning_rate": 3.9253666647257855e-05, "loss": 0.0, "step": 33232 }, { "epoch": 0.65, "learning_rate": 3.925301968700063e-05, "loss": 0.0, "step": 33234 }, { "epoch": 0.65, "learning_rate": 3.92523727267434e-05, "loss": 0.0, "step": 33236 }, { "epoch": 0.65, "learning_rate": 3.925172576648617e-05, "loss": 0.0056, "step": 33238 }, { "epoch": 0.65, "learning_rate": 3.925107880622894e-05, "loss": 0.0, "step": 33240 }, { "epoch": 0.65, "learning_rate": 3.92504318459717e-05, "loss": 0.0006, "step": 33242 }, { "epoch": 0.65, "learning_rate": 3.9249784885714476e-05, "loss": 0.0003, "step": 33244 }, { "epoch": 0.65, "learning_rate": 3.924913792545724e-05, "loss": 0.0007, "step": 33246 }, { "epoch": 0.65, "learning_rate": 3.924849096520001e-05, "loss": 0.0005, "step": 33248 }, { "epoch": 0.65, "learning_rate": 3.924784400494278e-05, "loss": 0.0119, "step": 33250 }, { "epoch": 0.65, "learning_rate": 3.9247197044685546e-05, "loss": 0.0002, "step": 33252 }, { "epoch": 0.65, "learning_rate": 3.9246550084428315e-05, "loss": 0.0001, "step": 33254 }, { "epoch": 0.65, "learning_rate": 3.9245903124171084e-05, "loss": 0.0002, "step": 33256 }, { "epoch": 0.65, "learning_rate": 3.924525616391385e-05, "loss": 0.0, "step": 33258 }, { "epoch": 0.65, "learning_rate": 3.924460920365662e-05, "loss": 0.0, "step": 33260 }, { "epoch": 0.65, "learning_rate": 3.924396224339939e-05, "loss": 0.0277, "step": 33262 }, { "epoch": 0.65, "learning_rate": 3.9243315283142154e-05, "loss": 0.0, "step": 33264 }, { "epoch": 0.65, "learning_rate": 3.924266832288493e-05, "loss": 0.0, "step": 33266 }, { "epoch": 0.65, "learning_rate": 3.92420213626277e-05, "loss": 0.0, "step": 33268 }, { "epoch": 0.65, "learning_rate": 3.924137440237046e-05, "loss": 0.0166, "step": 33270 }, { "epoch": 0.65, "learning_rate": 3.924072744211324e-05, "loss": 0.0005, "step": 33272 }, { "epoch": 0.65, "learning_rate": 3.9240080481856e-05, "loss": 0.0001, "step": 33274 }, { "epoch": 0.65, "learning_rate": 3.923943352159877e-05, "loss": 0.002, "step": 33276 }, { "epoch": 0.65, "learning_rate": 3.923878656134154e-05, "loss": 0.0081, "step": 33278 }, { "epoch": 0.65, "learning_rate": 3.9238139601084306e-05, "loss": 0.0012, "step": 33280 }, { "epoch": 0.65, "learning_rate": 3.9237492640827076e-05, "loss": 0.0001, "step": 33282 }, { "epoch": 0.65, "learning_rate": 3.9236845680569845e-05, "loss": 0.0001, "step": 33284 }, { "epoch": 0.65, "learning_rate": 3.9236198720312614e-05, "loss": 0.0, "step": 33286 }, { "epoch": 0.65, "learning_rate": 3.923555176005538e-05, "loss": 0.0004, "step": 33288 }, { "epoch": 0.65, "learning_rate": 3.923490479979815e-05, "loss": 0.0019, "step": 33290 }, { "epoch": 0.65, "learning_rate": 3.9234257839540914e-05, "loss": 0.0012, "step": 33292 }, { "epoch": 0.65, "learning_rate": 3.923361087928369e-05, "loss": 0.0, "step": 33294 }, { "epoch": 0.65, "learning_rate": 3.923296391902645e-05, "loss": 0.0009, "step": 33296 }, { "epoch": 0.65, "learning_rate": 3.923231695876923e-05, "loss": 0.0001, "step": 33298 }, { "epoch": 0.65, "learning_rate": 3.9231669998512e-05, "loss": 0.0, "step": 33300 }, { "epoch": 0.65, "learning_rate": 3.923102303825476e-05, "loss": 0.0001, "step": 33302 }, { "epoch": 0.65, "learning_rate": 3.9230376077997535e-05, "loss": 0.0001, "step": 33304 }, { "epoch": 0.65, "learning_rate": 3.92297291177403e-05, "loss": 0.0, "step": 33306 }, { "epoch": 0.65, "learning_rate": 3.922908215748307e-05, "loss": 0.0002, "step": 33308 }, { "epoch": 0.65, "learning_rate": 3.9228435197225836e-05, "loss": 0.0021, "step": 33310 }, { "epoch": 0.65, "learning_rate": 3.9227788236968605e-05, "loss": 0.0068, "step": 33312 }, { "epoch": 0.65, "learning_rate": 3.9227141276711374e-05, "loss": 0.0, "step": 33314 }, { "epoch": 0.65, "learning_rate": 3.922649431645414e-05, "loss": 0.0001, "step": 33316 }, { "epoch": 0.65, "learning_rate": 3.922584735619691e-05, "loss": 0.01, "step": 33318 }, { "epoch": 0.65, "learning_rate": 3.922520039593968e-05, "loss": 0.0001, "step": 33320 }, { "epoch": 0.65, "learning_rate": 3.922455343568245e-05, "loss": 0.0445, "step": 33322 }, { "epoch": 0.65, "learning_rate": 3.922390647542521e-05, "loss": 0.0001, "step": 33324 }, { "epoch": 0.65, "learning_rate": 3.922325951516799e-05, "loss": 0.0, "step": 33326 }, { "epoch": 0.65, "learning_rate": 3.922261255491075e-05, "loss": 0.0002, "step": 33328 }, { "epoch": 0.65, "learning_rate": 3.922196559465352e-05, "loss": 0.0003, "step": 33330 }, { "epoch": 0.65, "learning_rate": 3.922131863439629e-05, "loss": 0.0015, "step": 33332 }, { "epoch": 0.65, "learning_rate": 3.922067167413906e-05, "loss": 0.0049, "step": 33334 }, { "epoch": 0.65, "learning_rate": 3.922002471388183e-05, "loss": 0.0181, "step": 33336 }, { "epoch": 0.65, "learning_rate": 3.9219377753624596e-05, "loss": 0.0315, "step": 33338 }, { "epoch": 0.65, "learning_rate": 3.9218730793367365e-05, "loss": 0.0029, "step": 33340 }, { "epoch": 0.65, "learning_rate": 3.9218083833110135e-05, "loss": 0.0, "step": 33342 }, { "epoch": 0.65, "learning_rate": 3.9217436872852904e-05, "loss": 0.0, "step": 33344 }, { "epoch": 0.65, "learning_rate": 3.921678991259567e-05, "loss": 0.0, "step": 33346 }, { "epoch": 0.65, "learning_rate": 3.921614295233844e-05, "loss": 0.0002, "step": 33348 }, { "epoch": 0.65, "learning_rate": 3.921549599208121e-05, "loss": 0.0, "step": 33350 }, { "epoch": 0.65, "learning_rate": 3.921484903182397e-05, "loss": 0.0021, "step": 33352 }, { "epoch": 0.65, "learning_rate": 3.921420207156675e-05, "loss": 0.0001, "step": 33354 }, { "epoch": 0.65, "learning_rate": 3.921355511130951e-05, "loss": 0.0007, "step": 33356 }, { "epoch": 0.65, "learning_rate": 3.921290815105228e-05, "loss": 0.0, "step": 33358 }, { "epoch": 0.65, "learning_rate": 3.921226119079505e-05, "loss": 0.0002, "step": 33360 }, { "epoch": 0.65, "learning_rate": 3.921161423053782e-05, "loss": 0.0006, "step": 33362 }, { "epoch": 0.65, "learning_rate": 3.921096727028059e-05, "loss": 0.0127, "step": 33364 }, { "epoch": 0.65, "learning_rate": 3.921032031002336e-05, "loss": 0.0124, "step": 33366 }, { "epoch": 0.65, "learning_rate": 3.9209673349766126e-05, "loss": 0.0001, "step": 33368 }, { "epoch": 0.65, "learning_rate": 3.9209026389508895e-05, "loss": 0.0001, "step": 33370 }, { "epoch": 0.65, "learning_rate": 3.9208379429251664e-05, "loss": 0.0002, "step": 33372 }, { "epoch": 0.65, "learning_rate": 3.9207732468994426e-05, "loss": 0.0152, "step": 33374 }, { "epoch": 0.65, "learning_rate": 3.92070855087372e-05, "loss": 0.0, "step": 33376 }, { "epoch": 0.65, "learning_rate": 3.920643854847997e-05, "loss": 0.0, "step": 33378 }, { "epoch": 0.65, "learning_rate": 3.920579158822274e-05, "loss": 0.0, "step": 33380 }, { "epoch": 0.65, "learning_rate": 3.920514462796551e-05, "loss": 0.0123, "step": 33382 }, { "epoch": 0.65, "learning_rate": 3.920449766770827e-05, "loss": 0.0001, "step": 33384 }, { "epoch": 0.65, "learning_rate": 3.920385070745105e-05, "loss": 0.0001, "step": 33386 }, { "epoch": 0.65, "learning_rate": 3.920320374719381e-05, "loss": 0.0, "step": 33388 }, { "epoch": 0.65, "learning_rate": 3.920255678693658e-05, "loss": 0.0, "step": 33390 }, { "epoch": 0.65, "learning_rate": 3.920190982667935e-05, "loss": 0.0002, "step": 33392 }, { "epoch": 0.65, "learning_rate": 3.920126286642212e-05, "loss": 0.0003, "step": 33394 }, { "epoch": 0.65, "learning_rate": 3.9200615906164886e-05, "loss": 0.0001, "step": 33396 }, { "epoch": 0.65, "learning_rate": 3.9199968945907655e-05, "loss": 0.0, "step": 33398 }, { "epoch": 0.65, "learning_rate": 3.9199321985650424e-05, "loss": 0.0002, "step": 33400 }, { "epoch": 0.65, "learning_rate": 3.9198675025393194e-05, "loss": 0.0051, "step": 33402 }, { "epoch": 0.65, "learning_rate": 3.919802806513596e-05, "loss": 0.0105, "step": 33404 }, { "epoch": 0.65, "learning_rate": 3.9197381104878725e-05, "loss": 0.0001, "step": 33406 }, { "epoch": 0.65, "learning_rate": 3.91967341446215e-05, "loss": 0.0, "step": 33408 }, { "epoch": 0.65, "learning_rate": 3.919608718436426e-05, "loss": 0.001, "step": 33410 }, { "epoch": 0.65, "learning_rate": 3.919544022410703e-05, "loss": 0.0003, "step": 33412 }, { "epoch": 0.65, "learning_rate": 3.919479326384981e-05, "loss": 0.0, "step": 33414 }, { "epoch": 0.65, "learning_rate": 3.919414630359257e-05, "loss": 0.0001, "step": 33416 }, { "epoch": 0.65, "learning_rate": 3.919349934333534e-05, "loss": 0.0, "step": 33418 }, { "epoch": 0.65, "learning_rate": 3.919285238307811e-05, "loss": 0.0029, "step": 33420 }, { "epoch": 0.65, "learning_rate": 3.919220542282088e-05, "loss": 0.0, "step": 33422 }, { "epoch": 0.65, "learning_rate": 3.919155846256365e-05, "loss": 0.0017, "step": 33424 }, { "epoch": 0.65, "learning_rate": 3.9190911502306416e-05, "loss": 0.0172, "step": 33426 }, { "epoch": 0.65, "learning_rate": 3.9190264542049185e-05, "loss": 0.0004, "step": 33428 }, { "epoch": 0.65, "learning_rate": 3.9189617581791954e-05, "loss": 0.0001, "step": 33430 }, { "epoch": 0.65, "learning_rate": 3.918897062153472e-05, "loss": 0.0008, "step": 33432 }, { "epoch": 0.65, "learning_rate": 3.9188323661277485e-05, "loss": 0.0, "step": 33434 }, { "epoch": 0.65, "learning_rate": 3.918767670102026e-05, "loss": 0.0009, "step": 33436 }, { "epoch": 0.65, "learning_rate": 3.9187029740763023e-05, "loss": 0.0054, "step": 33438 }, { "epoch": 0.65, "learning_rate": 3.91863827805058e-05, "loss": 0.008, "step": 33440 }, { "epoch": 0.65, "learning_rate": 3.918573582024856e-05, "loss": 0.0061, "step": 33442 }, { "epoch": 0.65, "learning_rate": 3.918508885999133e-05, "loss": 0.0, "step": 33444 }, { "epoch": 0.65, "learning_rate": 3.9184441899734107e-05, "loss": 0.0001, "step": 33446 }, { "epoch": 0.65, "learning_rate": 3.918379493947687e-05, "loss": 0.0, "step": 33448 }, { "epoch": 0.65, "learning_rate": 3.918314797921964e-05, "loss": 0.0001, "step": 33450 }, { "epoch": 0.65, "learning_rate": 3.918250101896241e-05, "loss": 0.0, "step": 33452 }, { "epoch": 0.65, "learning_rate": 3.9181854058705176e-05, "loss": 0.0083, "step": 33454 }, { "epoch": 0.65, "learning_rate": 3.918120709844794e-05, "loss": 0.0001, "step": 33456 }, { "epoch": 0.65, "learning_rate": 3.9180560138190714e-05, "loss": 0.0, "step": 33458 }, { "epoch": 0.65, "learning_rate": 3.9179913177933483e-05, "loss": 0.0006, "step": 33460 }, { "epoch": 0.65, "learning_rate": 3.917926621767625e-05, "loss": 0.0, "step": 33462 }, { "epoch": 0.65, "learning_rate": 3.917861925741902e-05, "loss": 0.0039, "step": 33464 }, { "epoch": 0.65, "learning_rate": 3.9177972297161784e-05, "loss": 0.0, "step": 33466 }, { "epoch": 0.65, "learning_rate": 3.917732533690456e-05, "loss": 0.0005, "step": 33468 }, { "epoch": 0.65, "learning_rate": 3.917667837664732e-05, "loss": 0.0023, "step": 33470 }, { "epoch": 0.65, "learning_rate": 3.917603141639009e-05, "loss": 0.0, "step": 33472 }, { "epoch": 0.65, "learning_rate": 3.917538445613286e-05, "loss": 0.0, "step": 33474 }, { "epoch": 0.65, "learning_rate": 3.917473749587563e-05, "loss": 0.0174, "step": 33476 }, { "epoch": 0.65, "learning_rate": 3.91740905356184e-05, "loss": 0.0321, "step": 33478 }, { "epoch": 0.65, "learning_rate": 3.917344357536117e-05, "loss": 0.0, "step": 33480 }, { "epoch": 0.65, "learning_rate": 3.9172796615103937e-05, "loss": 0.0, "step": 33482 }, { "epoch": 0.65, "learning_rate": 3.9172149654846706e-05, "loss": 0.0001, "step": 33484 }, { "epoch": 0.65, "learning_rate": 3.9171502694589475e-05, "loss": 0.0001, "step": 33486 }, { "epoch": 0.65, "learning_rate": 3.917085573433224e-05, "loss": 0.0002, "step": 33488 }, { "epoch": 0.65, "learning_rate": 3.917020877407501e-05, "loss": 0.0001, "step": 33490 }, { "epoch": 0.65, "learning_rate": 3.916956181381778e-05, "loss": 0.0, "step": 33492 }, { "epoch": 0.65, "learning_rate": 3.9168914853560544e-05, "loss": 0.0, "step": 33494 }, { "epoch": 0.65, "learning_rate": 3.916826789330332e-05, "loss": 0.0, "step": 33496 }, { "epoch": 0.65, "learning_rate": 3.916762093304608e-05, "loss": 0.0003, "step": 33498 }, { "epoch": 0.65, "learning_rate": 3.916697397278885e-05, "loss": 0.0001, "step": 33500 }, { "epoch": 0.65, "learning_rate": 3.916632701253162e-05, "loss": 0.0001, "step": 33502 }, { "epoch": 0.65, "learning_rate": 3.916568005227439e-05, "loss": 0.0001, "step": 33504 }, { "epoch": 0.65, "learning_rate": 3.916503309201716e-05, "loss": 0.0005, "step": 33506 }, { "epoch": 0.65, "learning_rate": 3.916438613175993e-05, "loss": 0.0067, "step": 33508 }, { "epoch": 0.65, "learning_rate": 3.91637391715027e-05, "loss": 0.0002, "step": 33510 }, { "epoch": 0.65, "learning_rate": 3.9163092211245466e-05, "loss": 0.0, "step": 33512 }, { "epoch": 0.65, "learning_rate": 3.9162445250988235e-05, "loss": 0.0, "step": 33514 }, { "epoch": 0.65, "learning_rate": 3.9161798290731e-05, "loss": 0.0, "step": 33516 }, { "epoch": 0.65, "learning_rate": 3.916115133047377e-05, "loss": 0.0004, "step": 33518 }, { "epoch": 0.65, "learning_rate": 3.9160504370216536e-05, "loss": 0.0001, "step": 33520 }, { "epoch": 0.65, "learning_rate": 3.915985740995931e-05, "loss": 0.0003, "step": 33522 }, { "epoch": 0.65, "learning_rate": 3.915921044970208e-05, "loss": 0.0, "step": 33524 }, { "epoch": 0.65, "learning_rate": 3.915856348944484e-05, "loss": 0.0026, "step": 33526 }, { "epoch": 0.65, "learning_rate": 3.915791652918762e-05, "loss": 0.0022, "step": 33528 }, { "epoch": 0.65, "learning_rate": 3.915726956893038e-05, "loss": 0.0, "step": 33530 }, { "epoch": 0.65, "learning_rate": 3.915662260867315e-05, "loss": 0.0133, "step": 33532 }, { "epoch": 0.65, "learning_rate": 3.915597564841592e-05, "loss": 0.0, "step": 33534 }, { "epoch": 0.65, "learning_rate": 3.915532868815869e-05, "loss": 0.0008, "step": 33536 }, { "epoch": 0.65, "learning_rate": 3.915468172790146e-05, "loss": 0.01, "step": 33538 }, { "epoch": 0.65, "learning_rate": 3.9154034767644226e-05, "loss": 0.0003, "step": 33540 }, { "epoch": 0.65, "learning_rate": 3.9153387807386996e-05, "loss": 0.0065, "step": 33542 }, { "epoch": 0.65, "learning_rate": 3.9152740847129765e-05, "loss": 0.0001, "step": 33544 }, { "epoch": 0.65, "learning_rate": 3.9152093886872534e-05, "loss": 0.0, "step": 33546 }, { "epoch": 0.65, "learning_rate": 3.9151446926615296e-05, "loss": 0.0051, "step": 33548 }, { "epoch": 0.65, "learning_rate": 3.915079996635807e-05, "loss": 0.0001, "step": 33550 }, { "epoch": 0.65, "learning_rate": 3.9150153006100834e-05, "loss": 0.0, "step": 33552 }, { "epoch": 0.65, "learning_rate": 3.91495060458436e-05, "loss": 0.01, "step": 33554 }, { "epoch": 0.65, "learning_rate": 3.914885908558637e-05, "loss": 0.0002, "step": 33556 }, { "epoch": 0.65, "learning_rate": 3.914821212532914e-05, "loss": 0.0001, "step": 33558 }, { "epoch": 0.65, "learning_rate": 3.914756516507191e-05, "loss": 0.0, "step": 33560 }, { "epoch": 0.65, "learning_rate": 3.914691820481468e-05, "loss": 0.0001, "step": 33562 }, { "epoch": 0.65, "learning_rate": 3.914627124455745e-05, "loss": 0.0002, "step": 33564 }, { "epoch": 0.65, "learning_rate": 3.914562428430022e-05, "loss": 0.0003, "step": 33566 }, { "epoch": 0.65, "learning_rate": 3.914497732404299e-05, "loss": 0.0002, "step": 33568 }, { "epoch": 0.65, "learning_rate": 3.9144330363785756e-05, "loss": 0.0008, "step": 33570 }, { "epoch": 0.65, "learning_rate": 3.9143683403528525e-05, "loss": 0.0004, "step": 33572 }, { "epoch": 0.65, "learning_rate": 3.9143036443271294e-05, "loss": 0.0136, "step": 33574 }, { "epoch": 0.65, "learning_rate": 3.9142389483014056e-05, "loss": 0.0, "step": 33576 }, { "epoch": 0.65, "learning_rate": 3.914174252275683e-05, "loss": 0.0, "step": 33578 }, { "epoch": 0.65, "learning_rate": 3.9141095562499595e-05, "loss": 0.0, "step": 33580 }, { "epoch": 0.65, "learning_rate": 3.914044860224237e-05, "loss": 0.0002, "step": 33582 }, { "epoch": 0.65, "learning_rate": 3.913980164198513e-05, "loss": 0.0, "step": 33584 }, { "epoch": 0.65, "learning_rate": 3.91391546817279e-05, "loss": 0.0, "step": 33586 }, { "epoch": 0.65, "learning_rate": 3.913850772147067e-05, "loss": 0.0, "step": 33588 }, { "epoch": 0.65, "learning_rate": 3.913786076121344e-05, "loss": 0.0001, "step": 33590 }, { "epoch": 0.65, "learning_rate": 3.913721380095621e-05, "loss": 0.0072, "step": 33592 }, { "epoch": 0.65, "learning_rate": 3.913656684069898e-05, "loss": 0.0008, "step": 33594 }, { "epoch": 0.65, "learning_rate": 3.913591988044175e-05, "loss": 0.0, "step": 33596 }, { "epoch": 0.65, "learning_rate": 3.913527292018451e-05, "loss": 0.001, "step": 33598 }, { "epoch": 0.65, "learning_rate": 3.9134625959927285e-05, "loss": 0.0034, "step": 33600 }, { "epoch": 0.65, "learning_rate": 3.9133978999670055e-05, "loss": 0.0005, "step": 33602 }, { "epoch": 0.65, "learning_rate": 3.9133332039412824e-05, "loss": 0.0, "step": 33604 }, { "epoch": 0.65, "learning_rate": 3.913268507915559e-05, "loss": 0.0136, "step": 33606 }, { "epoch": 0.65, "learning_rate": 3.9132038118898355e-05, "loss": 0.0, "step": 33608 }, { "epoch": 0.65, "learning_rate": 3.913139115864113e-05, "loss": 0.0137, "step": 33610 }, { "epoch": 0.65, "learning_rate": 3.913074419838389e-05, "loss": 0.0, "step": 33612 }, { "epoch": 0.65, "learning_rate": 3.913009723812666e-05, "loss": 0.0055, "step": 33614 }, { "epoch": 0.65, "learning_rate": 3.912945027786943e-05, "loss": 0.0002, "step": 33616 }, { "epoch": 0.65, "learning_rate": 3.91288033176122e-05, "loss": 0.0035, "step": 33618 }, { "epoch": 0.65, "learning_rate": 3.912815635735497e-05, "loss": 0.0001, "step": 33620 }, { "epoch": 0.65, "learning_rate": 3.912750939709774e-05, "loss": 0.0001, "step": 33622 }, { "epoch": 0.65, "learning_rate": 3.912686243684051e-05, "loss": 0.0, "step": 33624 }, { "epoch": 0.65, "learning_rate": 3.912621547658328e-05, "loss": 0.0001, "step": 33626 }, { "epoch": 0.65, "learning_rate": 3.9125568516326046e-05, "loss": 0.003, "step": 33628 }, { "epoch": 0.65, "learning_rate": 3.912492155606881e-05, "loss": 0.0029, "step": 33630 }, { "epoch": 0.65, "learning_rate": 3.9124274595811584e-05, "loss": 0.0181, "step": 33632 }, { "epoch": 0.65, "learning_rate": 3.9123627635554346e-05, "loss": 0.0, "step": 33634 }, { "epoch": 0.65, "learning_rate": 3.9122980675297115e-05, "loss": 0.0015, "step": 33636 }, { "epoch": 0.65, "learning_rate": 3.912233371503989e-05, "loss": 0.0009, "step": 33638 }, { "epoch": 0.65, "learning_rate": 3.9121686754782654e-05, "loss": 0.0003, "step": 33640 }, { "epoch": 0.65, "learning_rate": 3.912103979452542e-05, "loss": 0.0001, "step": 33642 }, { "epoch": 0.65, "learning_rate": 3.912039283426819e-05, "loss": 0.0246, "step": 33644 }, { "epoch": 0.65, "learning_rate": 3.911974587401096e-05, "loss": 0.0023, "step": 33646 }, { "epoch": 0.65, "learning_rate": 3.911909891375373e-05, "loss": 0.0062, "step": 33648 }, { "epoch": 0.65, "learning_rate": 3.91184519534965e-05, "loss": 0.0, "step": 33650 }, { "epoch": 0.65, "learning_rate": 3.911780499323927e-05, "loss": 0.0005, "step": 33652 }, { "epoch": 0.65, "learning_rate": 3.911715803298204e-05, "loss": 0.0, "step": 33654 }, { "epoch": 0.65, "learning_rate": 3.9116511072724806e-05, "loss": 0.0055, "step": 33656 }, { "epoch": 0.65, "learning_rate": 3.911586411246757e-05, "loss": 0.0014, "step": 33658 }, { "epoch": 0.65, "learning_rate": 3.9115217152210344e-05, "loss": 0.0001, "step": 33660 }, { "epoch": 0.65, "learning_rate": 3.911457019195311e-05, "loss": 0.0094, "step": 33662 }, { "epoch": 0.65, "learning_rate": 3.911392323169588e-05, "loss": 0.0002, "step": 33664 }, { "epoch": 0.65, "learning_rate": 3.9113276271438645e-05, "loss": 0.0016, "step": 33666 }, { "epoch": 0.65, "learning_rate": 3.9112629311181414e-05, "loss": 0.0003, "step": 33668 }, { "epoch": 0.65, "learning_rate": 3.911198235092419e-05, "loss": 0.0006, "step": 33670 }, { "epoch": 0.65, "learning_rate": 3.911133539066695e-05, "loss": 0.001, "step": 33672 }, { "epoch": 0.65, "learning_rate": 3.911068843040972e-05, "loss": 0.0, "step": 33674 }, { "epoch": 0.65, "learning_rate": 3.911004147015249e-05, "loss": 0.0, "step": 33676 }, { "epoch": 0.65, "learning_rate": 3.910939450989526e-05, "loss": 0.0, "step": 33678 }, { "epoch": 0.65, "learning_rate": 3.910874754963803e-05, "loss": 0.0002, "step": 33680 }, { "epoch": 0.65, "learning_rate": 3.91081005893808e-05, "loss": 0.0003, "step": 33682 }, { "epoch": 0.65, "learning_rate": 3.910745362912357e-05, "loss": 0.0, "step": 33684 }, { "epoch": 0.65, "learning_rate": 3.9106806668866336e-05, "loss": 0.001, "step": 33686 }, { "epoch": 0.65, "learning_rate": 3.9106159708609105e-05, "loss": 0.0026, "step": 33688 }, { "epoch": 0.65, "learning_rate": 3.910551274835187e-05, "loss": 0.0, "step": 33690 }, { "epoch": 0.65, "learning_rate": 3.910486578809464e-05, "loss": 0.0, "step": 33692 }, { "epoch": 0.65, "learning_rate": 3.9104218827837405e-05, "loss": 0.0005, "step": 33694 }, { "epoch": 0.65, "learning_rate": 3.9103571867580174e-05, "loss": 0.0003, "step": 33696 }, { "epoch": 0.65, "learning_rate": 3.9102924907322943e-05, "loss": 0.0001, "step": 33698 }, { "epoch": 0.65, "learning_rate": 3.910227794706571e-05, "loss": 0.0008, "step": 33700 }, { "epoch": 0.65, "learning_rate": 3.910163098680848e-05, "loss": 0.0, "step": 33702 }, { "epoch": 0.65, "learning_rate": 3.910098402655125e-05, "loss": 0.0, "step": 33704 }, { "epoch": 0.65, "learning_rate": 3.910033706629402e-05, "loss": 0.0008, "step": 33706 }, { "epoch": 0.65, "learning_rate": 3.909969010603679e-05, "loss": 0.0005, "step": 33708 }, { "epoch": 0.65, "learning_rate": 3.909904314577956e-05, "loss": 0.0001, "step": 33710 }, { "epoch": 0.65, "learning_rate": 3.909839618552232e-05, "loss": 0.0, "step": 33712 }, { "epoch": 0.65, "learning_rate": 3.9097749225265096e-05, "loss": 0.0004, "step": 33714 }, { "epoch": 0.65, "learning_rate": 3.9097102265007865e-05, "loss": 0.0, "step": 33716 }, { "epoch": 0.65, "learning_rate": 3.909645530475063e-05, "loss": 0.0, "step": 33718 }, { "epoch": 0.65, "learning_rate": 3.90958083444934e-05, "loss": 0.0, "step": 33720 }, { "epoch": 0.65, "learning_rate": 3.9095161384236166e-05, "loss": 0.0, "step": 33722 }, { "epoch": 0.65, "learning_rate": 3.909451442397894e-05, "loss": 0.0009, "step": 33724 }, { "epoch": 0.65, "learning_rate": 3.9093867463721704e-05, "loss": 0.0118, "step": 33726 }, { "epoch": 0.65, "learning_rate": 3.909322050346447e-05, "loss": 0.0043, "step": 33728 }, { "epoch": 0.65, "learning_rate": 3.909257354320724e-05, "loss": 0.0002, "step": 33730 }, { "epoch": 0.65, "learning_rate": 3.909192658295001e-05, "loss": 0.0001, "step": 33732 }, { "epoch": 0.65, "learning_rate": 3.909127962269278e-05, "loss": 0.0, "step": 33734 }, { "epoch": 0.65, "learning_rate": 3.909063266243555e-05, "loss": 0.014, "step": 33736 }, { "epoch": 0.65, "learning_rate": 3.908998570217832e-05, "loss": 0.0009, "step": 33738 }, { "epoch": 0.65, "learning_rate": 3.908933874192108e-05, "loss": 0.1176, "step": 33740 }, { "epoch": 0.65, "learning_rate": 3.9088691781663857e-05, "loss": 0.0001, "step": 33742 }, { "epoch": 0.65, "learning_rate": 3.908804482140662e-05, "loss": 0.0148, "step": 33744 }, { "epoch": 0.65, "learning_rate": 3.9087397861149395e-05, "loss": 0.01, "step": 33746 }, { "epoch": 0.66, "learning_rate": 3.9086750900892164e-05, "loss": 0.0015, "step": 33748 }, { "epoch": 0.66, "learning_rate": 3.9086103940634926e-05, "loss": 0.0, "step": 33750 }, { "epoch": 0.66, "learning_rate": 3.90854569803777e-05, "loss": 0.0001, "step": 33752 }, { "epoch": 0.66, "learning_rate": 3.9084810020120464e-05, "loss": 0.0008, "step": 33754 }, { "epoch": 0.66, "learning_rate": 3.908416305986323e-05, "loss": 0.0026, "step": 33756 }, { "epoch": 0.66, "learning_rate": 3.9083516099606e-05, "loss": 0.0078, "step": 33758 }, { "epoch": 0.66, "learning_rate": 3.908286913934877e-05, "loss": 0.0, "step": 33760 }, { "epoch": 0.66, "learning_rate": 3.908222217909154e-05, "loss": 0.0001, "step": 33762 }, { "epoch": 0.66, "learning_rate": 3.908157521883431e-05, "loss": 0.0013, "step": 33764 }, { "epoch": 0.66, "learning_rate": 3.908092825857708e-05, "loss": 0.0002, "step": 33766 }, { "epoch": 0.66, "learning_rate": 3.908028129831985e-05, "loss": 0.0004, "step": 33768 }, { "epoch": 0.66, "learning_rate": 3.907963433806262e-05, "loss": 0.0, "step": 33770 }, { "epoch": 0.66, "learning_rate": 3.907898737780538e-05, "loss": 0.0, "step": 33772 }, { "epoch": 0.66, "learning_rate": 3.9078340417548155e-05, "loss": 0.0, "step": 33774 }, { "epoch": 0.66, "learning_rate": 3.907769345729092e-05, "loss": 0.0048, "step": 33776 }, { "epoch": 0.66, "learning_rate": 3.9077046497033686e-05, "loss": 0.0001, "step": 33778 }, { "epoch": 0.66, "learning_rate": 3.907639953677646e-05, "loss": 0.0, "step": 33780 }, { "epoch": 0.66, "learning_rate": 3.9075752576519225e-05, "loss": 0.0004, "step": 33782 }, { "epoch": 0.66, "learning_rate": 3.9075105616261994e-05, "loss": 0.0008, "step": 33784 }, { "epoch": 0.66, "learning_rate": 3.907445865600476e-05, "loss": 0.0, "step": 33786 }, { "epoch": 0.66, "learning_rate": 3.907381169574753e-05, "loss": 0.0011, "step": 33788 }, { "epoch": 0.66, "learning_rate": 3.90731647354903e-05, "loss": 0.0017, "step": 33790 }, { "epoch": 0.66, "learning_rate": 3.907251777523307e-05, "loss": 0.0287, "step": 33792 }, { "epoch": 0.66, "learning_rate": 3.907187081497584e-05, "loss": 0.0, "step": 33794 }, { "epoch": 0.66, "learning_rate": 3.907122385471861e-05, "loss": 0.0099, "step": 33796 }, { "epoch": 0.66, "learning_rate": 3.907057689446138e-05, "loss": 0.0001, "step": 33798 }, { "epoch": 0.66, "learning_rate": 3.906992993420414e-05, "loss": 0.0003, "step": 33800 }, { "epoch": 0.66, "learning_rate": 3.9069282973946916e-05, "loss": 0.0001, "step": 33802 }, { "epoch": 0.66, "learning_rate": 3.906863601368968e-05, "loss": 0.0, "step": 33804 }, { "epoch": 0.66, "learning_rate": 3.9067989053432454e-05, "loss": 0.0, "step": 33806 }, { "epoch": 0.66, "learning_rate": 3.9067342093175216e-05, "loss": 0.0, "step": 33808 }, { "epoch": 0.66, "learning_rate": 3.9066695132917985e-05, "loss": 0.0, "step": 33810 }, { "epoch": 0.66, "learning_rate": 3.9066048172660754e-05, "loss": 0.0001, "step": 33812 }, { "epoch": 0.66, "learning_rate": 3.906540121240352e-05, "loss": 0.0056, "step": 33814 }, { "epoch": 0.66, "learning_rate": 3.906475425214629e-05, "loss": 0.0005, "step": 33816 }, { "epoch": 0.66, "learning_rate": 3.906410729188906e-05, "loss": 0.0104, "step": 33818 }, { "epoch": 0.66, "learning_rate": 3.906346033163183e-05, "loss": 0.0049, "step": 33820 }, { "epoch": 0.66, "learning_rate": 3.906281337137459e-05, "loss": 0.0001, "step": 33822 }, { "epoch": 0.66, "learning_rate": 3.906216641111737e-05, "loss": 0.0001, "step": 33824 }, { "epoch": 0.66, "learning_rate": 3.906151945086014e-05, "loss": 0.0029, "step": 33826 }, { "epoch": 0.66, "learning_rate": 3.906087249060291e-05, "loss": 0.0, "step": 33828 }, { "epoch": 0.66, "learning_rate": 3.9060225530345676e-05, "loss": 0.0, "step": 33830 }, { "epoch": 0.66, "learning_rate": 3.905957857008844e-05, "loss": 0.0003, "step": 33832 }, { "epoch": 0.66, "learning_rate": 3.9058931609831214e-05, "loss": 0.0006, "step": 33834 }, { "epoch": 0.66, "learning_rate": 3.9058284649573976e-05, "loss": 0.0119, "step": 33836 }, { "epoch": 0.66, "learning_rate": 3.9057637689316745e-05, "loss": 0.0, "step": 33838 }, { "epoch": 0.66, "learning_rate": 3.9056990729059515e-05, "loss": 0.0004, "step": 33840 }, { "epoch": 0.66, "learning_rate": 3.9056343768802284e-05, "loss": 0.0006, "step": 33842 }, { "epoch": 0.66, "learning_rate": 3.905569680854505e-05, "loss": 0.0, "step": 33844 }, { "epoch": 0.66, "learning_rate": 3.905504984828782e-05, "loss": 0.0, "step": 33846 }, { "epoch": 0.66, "learning_rate": 3.905440288803059e-05, "loss": 0.0001, "step": 33848 }, { "epoch": 0.66, "learning_rate": 3.905375592777336e-05, "loss": 0.0048, "step": 33850 }, { "epoch": 0.66, "learning_rate": 3.905310896751613e-05, "loss": 0.0003, "step": 33852 }, { "epoch": 0.66, "learning_rate": 3.905246200725889e-05, "loss": 0.005, "step": 33854 }, { "epoch": 0.66, "learning_rate": 3.905181504700167e-05, "loss": 0.0, "step": 33856 }, { "epoch": 0.66, "learning_rate": 3.9051168086744436e-05, "loss": 0.0121, "step": 33858 }, { "epoch": 0.66, "learning_rate": 3.90505211264872e-05, "loss": 0.0002, "step": 33860 }, { "epoch": 0.66, "learning_rate": 3.9049874166229974e-05, "loss": 0.0017, "step": 33862 }, { "epoch": 0.66, "learning_rate": 3.904922720597274e-05, "loss": 0.0, "step": 33864 }, { "epoch": 0.66, "learning_rate": 3.904858024571551e-05, "loss": 0.0, "step": 33866 }, { "epoch": 0.66, "learning_rate": 3.9047933285458275e-05, "loss": 0.0035, "step": 33868 }, { "epoch": 0.66, "learning_rate": 3.9047286325201044e-05, "loss": 0.0003, "step": 33870 }, { "epoch": 0.66, "learning_rate": 3.904663936494381e-05, "loss": 0.0001, "step": 33872 }, { "epoch": 0.66, "learning_rate": 3.904599240468658e-05, "loss": 0.0, "step": 33874 }, { "epoch": 0.66, "learning_rate": 3.904534544442935e-05, "loss": 0.0001, "step": 33876 }, { "epoch": 0.66, "learning_rate": 3.904469848417212e-05, "loss": 0.0002, "step": 33878 }, { "epoch": 0.66, "learning_rate": 3.904405152391489e-05, "loss": 0.0024, "step": 33880 }, { "epoch": 0.66, "learning_rate": 3.904340456365765e-05, "loss": 0.0037, "step": 33882 }, { "epoch": 0.66, "learning_rate": 3.904275760340043e-05, "loss": 0.0015, "step": 33884 }, { "epoch": 0.66, "learning_rate": 3.904211064314319e-05, "loss": 0.0002, "step": 33886 }, { "epoch": 0.66, "learning_rate": 3.9041463682885966e-05, "loss": 0.0004, "step": 33888 }, { "epoch": 0.66, "learning_rate": 3.904081672262873e-05, "loss": 0.0, "step": 33890 }, { "epoch": 0.66, "learning_rate": 3.90401697623715e-05, "loss": 0.0001, "step": 33892 }, { "epoch": 0.66, "learning_rate": 3.903952280211427e-05, "loss": 0.0004, "step": 33894 }, { "epoch": 0.66, "learning_rate": 3.9038875841857035e-05, "loss": 0.0002, "step": 33896 }, { "epoch": 0.66, "learning_rate": 3.9038228881599804e-05, "loss": 0.0044, "step": 33898 }, { "epoch": 0.66, "learning_rate": 3.9037581921342574e-05, "loss": 0.0007, "step": 33900 }, { "epoch": 0.66, "learning_rate": 3.903693496108534e-05, "loss": 0.0004, "step": 33902 }, { "epoch": 0.66, "learning_rate": 3.903628800082811e-05, "loss": 0.0, "step": 33904 }, { "epoch": 0.66, "learning_rate": 3.903564104057088e-05, "loss": 0.0003, "step": 33906 }, { "epoch": 0.66, "learning_rate": 3.903499408031365e-05, "loss": 0.0, "step": 33908 }, { "epoch": 0.66, "learning_rate": 3.903434712005642e-05, "loss": 0.0001, "step": 33910 }, { "epoch": 0.66, "learning_rate": 3.903370015979919e-05, "loss": 0.0, "step": 33912 }, { "epoch": 0.66, "learning_rate": 3.903305319954195e-05, "loss": 0.0, "step": 33914 }, { "epoch": 0.66, "learning_rate": 3.9032406239284726e-05, "loss": 0.0001, "step": 33916 }, { "epoch": 0.66, "learning_rate": 3.903175927902749e-05, "loss": 0.0001, "step": 33918 }, { "epoch": 0.66, "learning_rate": 3.903111231877026e-05, "loss": 0.0, "step": 33920 }, { "epoch": 0.66, "learning_rate": 3.903046535851303e-05, "loss": 0.0046, "step": 33922 }, { "epoch": 0.66, "learning_rate": 3.9029818398255796e-05, "loss": 0.0072, "step": 33924 }, { "epoch": 0.66, "learning_rate": 3.9029171437998565e-05, "loss": 0.0002, "step": 33926 }, { "epoch": 0.66, "learning_rate": 3.9028524477741334e-05, "loss": 0.0002, "step": 33928 }, { "epoch": 0.66, "learning_rate": 3.90278775174841e-05, "loss": 0.0006, "step": 33930 }, { "epoch": 0.66, "learning_rate": 3.902723055722687e-05, "loss": 0.0001, "step": 33932 }, { "epoch": 0.66, "learning_rate": 3.902658359696964e-05, "loss": 0.0047, "step": 33934 }, { "epoch": 0.66, "learning_rate": 3.9025936636712404e-05, "loss": 0.0025, "step": 33936 }, { "epoch": 0.66, "learning_rate": 3.902528967645518e-05, "loss": 0.0057, "step": 33938 }, { "epoch": 0.66, "learning_rate": 3.902464271619795e-05, "loss": 0.0001, "step": 33940 }, { "epoch": 0.66, "learning_rate": 3.902399575594071e-05, "loss": 0.0001, "step": 33942 }, { "epoch": 0.66, "learning_rate": 3.902334879568349e-05, "loss": 0.0013, "step": 33944 }, { "epoch": 0.66, "learning_rate": 3.902270183542625e-05, "loss": 0.0037, "step": 33946 }, { "epoch": 0.66, "learning_rate": 3.9022054875169025e-05, "loss": 0.0002, "step": 33948 }, { "epoch": 0.66, "learning_rate": 3.902140791491179e-05, "loss": 0.022, "step": 33950 }, { "epoch": 0.66, "learning_rate": 3.9020760954654556e-05, "loss": 0.0003, "step": 33952 }, { "epoch": 0.66, "learning_rate": 3.9020113994397325e-05, "loss": 0.0025, "step": 33954 }, { "epoch": 0.66, "learning_rate": 3.9019467034140094e-05, "loss": 0.0, "step": 33956 }, { "epoch": 0.66, "learning_rate": 3.9018820073882863e-05, "loss": 0.0, "step": 33958 }, { "epoch": 0.66, "learning_rate": 3.901817311362563e-05, "loss": 0.0039, "step": 33960 }, { "epoch": 0.66, "learning_rate": 3.90175261533684e-05, "loss": 0.0, "step": 33962 }, { "epoch": 0.66, "learning_rate": 3.9016879193111164e-05, "loss": 0.0, "step": 33964 }, { "epoch": 0.66, "learning_rate": 3.901623223285394e-05, "loss": 0.0001, "step": 33966 }, { "epoch": 0.66, "learning_rate": 3.90155852725967e-05, "loss": 0.0, "step": 33968 }, { "epoch": 0.66, "learning_rate": 3.901493831233948e-05, "loss": 0.0, "step": 33970 }, { "epoch": 0.66, "learning_rate": 3.901429135208225e-05, "loss": 0.0, "step": 33972 }, { "epoch": 0.66, "learning_rate": 3.901364439182501e-05, "loss": 0.0047, "step": 33974 }, { "epoch": 0.66, "learning_rate": 3.9012997431567785e-05, "loss": 0.0, "step": 33976 }, { "epoch": 0.66, "learning_rate": 3.901235047131055e-05, "loss": 0.003, "step": 33978 }, { "epoch": 0.66, "learning_rate": 3.9011703511053317e-05, "loss": 0.0001, "step": 33980 }, { "epoch": 0.66, "learning_rate": 3.9011056550796086e-05, "loss": 0.0, "step": 33982 }, { "epoch": 0.66, "learning_rate": 3.9010409590538855e-05, "loss": 0.0, "step": 33984 }, { "epoch": 0.66, "learning_rate": 3.9009762630281624e-05, "loss": 0.0014, "step": 33986 }, { "epoch": 0.66, "learning_rate": 3.900911567002439e-05, "loss": 0.002, "step": 33988 }, { "epoch": 0.66, "learning_rate": 3.900846870976716e-05, "loss": 0.0001, "step": 33990 }, { "epoch": 0.66, "learning_rate": 3.900782174950993e-05, "loss": 0.0, "step": 33992 }, { "epoch": 0.66, "learning_rate": 3.90071747892527e-05, "loss": 0.0001, "step": 33994 }, { "epoch": 0.66, "learning_rate": 3.900652782899546e-05, "loss": 0.0001, "step": 33996 }, { "epoch": 0.66, "learning_rate": 3.900588086873824e-05, "loss": 0.0001, "step": 33998 }, { "epoch": 0.66, "learning_rate": 3.9005233908481e-05, "loss": 0.0001, "step": 34000 }, { "epoch": 0.66, "learning_rate": 3.900458694822377e-05, "loss": 0.0, "step": 34002 }, { "epoch": 0.66, "learning_rate": 3.9003939987966546e-05, "loss": 0.0076, "step": 34004 }, { "epoch": 0.66, "learning_rate": 3.900329302770931e-05, "loss": 0.0001, "step": 34006 }, { "epoch": 0.66, "learning_rate": 3.9002646067452084e-05, "loss": 0.0, "step": 34008 }, { "epoch": 0.66, "learning_rate": 3.9001999107194846e-05, "loss": 0.0137, "step": 34010 }, { "epoch": 0.66, "learning_rate": 3.9001352146937615e-05, "loss": 0.0002, "step": 34012 }, { "epoch": 0.66, "learning_rate": 3.9000705186680384e-05, "loss": 0.0063, "step": 34014 }, { "epoch": 0.66, "learning_rate": 3.900005822642315e-05, "loss": 0.0, "step": 34016 }, { "epoch": 0.66, "learning_rate": 3.899941126616592e-05, "loss": 0.0, "step": 34018 }, { "epoch": 0.66, "learning_rate": 3.899876430590869e-05, "loss": 0.0, "step": 34020 }, { "epoch": 0.66, "learning_rate": 3.899811734565146e-05, "loss": 0.0, "step": 34022 }, { "epoch": 0.66, "learning_rate": 3.899747038539422e-05, "loss": 0.0036, "step": 34024 }, { "epoch": 0.66, "learning_rate": 3.8996823425137e-05, "loss": 0.0014, "step": 34026 }, { "epoch": 0.66, "learning_rate": 3.899617646487976e-05, "loss": 0.0001, "step": 34028 }, { "epoch": 0.66, "learning_rate": 3.899552950462254e-05, "loss": 0.0024, "step": 34030 }, { "epoch": 0.66, "learning_rate": 3.89948825443653e-05, "loss": 0.0009, "step": 34032 }, { "epoch": 0.66, "learning_rate": 3.899423558410807e-05, "loss": 0.0001, "step": 34034 }, { "epoch": 0.66, "learning_rate": 3.899358862385084e-05, "loss": 0.0107, "step": 34036 }, { "epoch": 0.66, "learning_rate": 3.8992941663593606e-05, "loss": 0.0002, "step": 34038 }, { "epoch": 0.66, "learning_rate": 3.8992294703336376e-05, "loss": 0.0071, "step": 34040 }, { "epoch": 0.66, "learning_rate": 3.8991647743079145e-05, "loss": 0.0698, "step": 34042 }, { "epoch": 0.66, "learning_rate": 3.8991000782821914e-05, "loss": 0.0012, "step": 34044 }, { "epoch": 0.66, "learning_rate": 3.8990353822564676e-05, "loss": 0.0, "step": 34046 }, { "epoch": 0.66, "learning_rate": 3.898970686230745e-05, "loss": 0.0002, "step": 34048 }, { "epoch": 0.66, "learning_rate": 3.898905990205022e-05, "loss": 0.0016, "step": 34050 }, { "epoch": 0.66, "learning_rate": 3.898841294179299e-05, "loss": 0.0001, "step": 34052 }, { "epoch": 0.66, "learning_rate": 3.898776598153576e-05, "loss": 0.021, "step": 34054 }, { "epoch": 0.66, "learning_rate": 3.898711902127852e-05, "loss": 0.0006, "step": 34056 }, { "epoch": 0.66, "learning_rate": 3.89864720610213e-05, "loss": 0.0, "step": 34058 }, { "epoch": 0.66, "learning_rate": 3.898582510076406e-05, "loss": 0.0089, "step": 34060 }, { "epoch": 0.66, "learning_rate": 3.898517814050683e-05, "loss": 0.0, "step": 34062 }, { "epoch": 0.66, "learning_rate": 3.89845311802496e-05, "loss": 0.0, "step": 34064 }, { "epoch": 0.66, "learning_rate": 3.898388421999237e-05, "loss": 0.0, "step": 34066 }, { "epoch": 0.66, "learning_rate": 3.8983237259735136e-05, "loss": 0.0, "step": 34068 }, { "epoch": 0.66, "learning_rate": 3.8982590299477905e-05, "loss": 0.0, "step": 34070 }, { "epoch": 0.66, "learning_rate": 3.8981943339220674e-05, "loss": 0.0001, "step": 34072 }, { "epoch": 0.66, "learning_rate": 3.898129637896344e-05, "loss": 0.0, "step": 34074 }, { "epoch": 0.66, "learning_rate": 3.898064941870621e-05, "loss": 0.0004, "step": 34076 }, { "epoch": 0.66, "learning_rate": 3.8980002458448975e-05, "loss": 0.0001, "step": 34078 }, { "epoch": 0.66, "learning_rate": 3.897935549819175e-05, "loss": 0.008, "step": 34080 }, { "epoch": 0.66, "learning_rate": 3.897870853793452e-05, "loss": 0.0, "step": 34082 }, { "epoch": 0.66, "learning_rate": 3.897806157767728e-05, "loss": 0.0042, "step": 34084 }, { "epoch": 0.66, "learning_rate": 3.897741461742006e-05, "loss": 0.0003, "step": 34086 }, { "epoch": 0.66, "learning_rate": 3.897676765716282e-05, "loss": 0.0, "step": 34088 }, { "epoch": 0.66, "learning_rate": 3.8976120696905596e-05, "loss": 0.0, "step": 34090 }, { "epoch": 0.66, "learning_rate": 3.897547373664836e-05, "loss": 0.0155, "step": 34092 }, { "epoch": 0.66, "learning_rate": 3.897482677639113e-05, "loss": 0.0044, "step": 34094 }, { "epoch": 0.66, "learning_rate": 3.8974179816133896e-05, "loss": 0.0001, "step": 34096 }, { "epoch": 0.66, "learning_rate": 3.8973532855876665e-05, "loss": 0.0001, "step": 34098 }, { "epoch": 0.66, "learning_rate": 3.8972885895619435e-05, "loss": 0.0, "step": 34100 }, { "epoch": 0.66, "learning_rate": 3.8972238935362204e-05, "loss": 0.0001, "step": 34102 }, { "epoch": 0.66, "learning_rate": 3.897159197510497e-05, "loss": 0.0, "step": 34104 }, { "epoch": 0.66, "learning_rate": 3.8970945014847735e-05, "loss": 0.0, "step": 34106 }, { "epoch": 0.66, "learning_rate": 3.897029805459051e-05, "loss": 0.0002, "step": 34108 }, { "epoch": 0.66, "learning_rate": 3.896965109433327e-05, "loss": 0.0, "step": 34110 }, { "epoch": 0.66, "learning_rate": 3.896900413407605e-05, "loss": 0.0014, "step": 34112 }, { "epoch": 0.66, "learning_rate": 3.896835717381881e-05, "loss": 0.0004, "step": 34114 }, { "epoch": 0.66, "learning_rate": 3.896771021356158e-05, "loss": 0.0162, "step": 34116 }, { "epoch": 0.66, "learning_rate": 3.8967063253304356e-05, "loss": 0.0002, "step": 34118 }, { "epoch": 0.66, "learning_rate": 3.896641629304712e-05, "loss": 0.0007, "step": 34120 }, { "epoch": 0.66, "learning_rate": 3.896576933278989e-05, "loss": 0.0018, "step": 34122 }, { "epoch": 0.66, "learning_rate": 3.896512237253266e-05, "loss": 0.0001, "step": 34124 }, { "epoch": 0.66, "learning_rate": 3.8964475412275426e-05, "loss": 0.0001, "step": 34126 }, { "epoch": 0.66, "learning_rate": 3.8963828452018195e-05, "loss": 0.0002, "step": 34128 }, { "epoch": 0.66, "learning_rate": 3.8963181491760964e-05, "loss": 0.006, "step": 34130 }, { "epoch": 0.66, "learning_rate": 3.896253453150373e-05, "loss": 0.0001, "step": 34132 }, { "epoch": 0.66, "learning_rate": 3.89618875712465e-05, "loss": 0.0368, "step": 34134 }, { "epoch": 0.66, "learning_rate": 3.896124061098927e-05, "loss": 0.0001, "step": 34136 }, { "epoch": 0.66, "learning_rate": 3.8960593650732034e-05, "loss": 0.0008, "step": 34138 }, { "epoch": 0.66, "learning_rate": 3.895994669047481e-05, "loss": 0.0, "step": 34140 }, { "epoch": 0.66, "learning_rate": 3.895929973021757e-05, "loss": 0.0044, "step": 34142 }, { "epoch": 0.66, "learning_rate": 3.895865276996034e-05, "loss": 0.0079, "step": 34144 }, { "epoch": 0.66, "learning_rate": 3.895800580970311e-05, "loss": 0.0001, "step": 34146 }, { "epoch": 0.66, "learning_rate": 3.895735884944588e-05, "loss": 0.0002, "step": 34148 }, { "epoch": 0.66, "learning_rate": 3.8956711889188655e-05, "loss": 0.0, "step": 34150 }, { "epoch": 0.66, "learning_rate": 3.895606492893142e-05, "loss": 0.0001, "step": 34152 }, { "epoch": 0.66, "learning_rate": 3.8955417968674186e-05, "loss": 0.003, "step": 34154 }, { "epoch": 0.66, "learning_rate": 3.8954771008416955e-05, "loss": 0.0143, "step": 34156 }, { "epoch": 0.66, "learning_rate": 3.8954124048159724e-05, "loss": 0.0, "step": 34158 }, { "epoch": 0.66, "learning_rate": 3.8953477087902494e-05, "loss": 0.0, "step": 34160 }, { "epoch": 0.66, "learning_rate": 3.895283012764526e-05, "loss": 0.0, "step": 34162 }, { "epoch": 0.66, "learning_rate": 3.895218316738803e-05, "loss": 0.0134, "step": 34164 }, { "epoch": 0.66, "learning_rate": 3.8951536207130794e-05, "loss": 0.0, "step": 34166 }, { "epoch": 0.66, "learning_rate": 3.895088924687357e-05, "loss": 0.0, "step": 34168 }, { "epoch": 0.66, "learning_rate": 3.895024228661633e-05, "loss": 0.0, "step": 34170 }, { "epoch": 0.66, "learning_rate": 3.894959532635911e-05, "loss": 0.002, "step": 34172 }, { "epoch": 0.66, "learning_rate": 3.894894836610187e-05, "loss": 0.0, "step": 34174 }, { "epoch": 0.66, "learning_rate": 3.894830140584464e-05, "loss": 0.0, "step": 34176 }, { "epoch": 0.66, "learning_rate": 3.894765444558741e-05, "loss": 0.0, "step": 34178 }, { "epoch": 0.66, "learning_rate": 3.894700748533018e-05, "loss": 0.0366, "step": 34180 }, { "epoch": 0.66, "learning_rate": 3.894636052507295e-05, "loss": 0.0012, "step": 34182 }, { "epoch": 0.66, "learning_rate": 3.8945713564815716e-05, "loss": 0.0004, "step": 34184 }, { "epoch": 0.66, "learning_rate": 3.8945066604558485e-05, "loss": 0.0001, "step": 34186 }, { "epoch": 0.66, "learning_rate": 3.894441964430125e-05, "loss": 0.0, "step": 34188 }, { "epoch": 0.66, "learning_rate": 3.894377268404402e-05, "loss": 0.0, "step": 34190 }, { "epoch": 0.66, "learning_rate": 3.8943125723786785e-05, "loss": 0.0001, "step": 34192 }, { "epoch": 0.66, "learning_rate": 3.894247876352956e-05, "loss": 0.0451, "step": 34194 }, { "epoch": 0.66, "learning_rate": 3.894183180327233e-05, "loss": 0.0, "step": 34196 }, { "epoch": 0.66, "learning_rate": 3.894118484301509e-05, "loss": 0.0001, "step": 34198 }, { "epoch": 0.66, "learning_rate": 3.894053788275787e-05, "loss": 0.0, "step": 34200 }, { "epoch": 0.66, "learning_rate": 3.893989092250063e-05, "loss": 0.0, "step": 34202 }, { "epoch": 0.66, "learning_rate": 3.89392439622434e-05, "loss": 0.0, "step": 34204 }, { "epoch": 0.66, "learning_rate": 3.893859700198617e-05, "loss": 0.0, "step": 34206 }, { "epoch": 0.66, "learning_rate": 3.893795004172894e-05, "loss": 0.0, "step": 34208 }, { "epoch": 0.66, "learning_rate": 3.893730308147171e-05, "loss": 0.0006, "step": 34210 }, { "epoch": 0.66, "learning_rate": 3.8936656121214476e-05, "loss": 0.0003, "step": 34212 }, { "epoch": 0.66, "learning_rate": 3.8936009160957245e-05, "loss": 0.0002, "step": 34214 }, { "epoch": 0.66, "learning_rate": 3.8935362200700014e-05, "loss": 0.0003, "step": 34216 }, { "epoch": 0.66, "learning_rate": 3.8934715240442783e-05, "loss": 0.002, "step": 34218 }, { "epoch": 0.66, "learning_rate": 3.8934068280185546e-05, "loss": 0.0012, "step": 34220 }, { "epoch": 0.66, "learning_rate": 3.893342131992832e-05, "loss": 0.0, "step": 34222 }, { "epoch": 0.66, "learning_rate": 3.8932774359671084e-05, "loss": 0.0001, "step": 34224 }, { "epoch": 0.66, "learning_rate": 3.893212739941385e-05, "loss": 0.0, "step": 34226 }, { "epoch": 0.66, "learning_rate": 3.893148043915663e-05, "loss": 0.0005, "step": 34228 }, { "epoch": 0.66, "learning_rate": 3.893083347889939e-05, "loss": 0.0002, "step": 34230 }, { "epoch": 0.66, "learning_rate": 3.893018651864217e-05, "loss": 0.0004, "step": 34232 }, { "epoch": 0.66, "learning_rate": 3.892953955838493e-05, "loss": 0.0006, "step": 34234 }, { "epoch": 0.66, "learning_rate": 3.89288925981277e-05, "loss": 0.0183, "step": 34236 }, { "epoch": 0.66, "learning_rate": 3.892824563787047e-05, "loss": 0.0002, "step": 34238 }, { "epoch": 0.66, "learning_rate": 3.8927598677613237e-05, "loss": 0.0, "step": 34240 }, { "epoch": 0.66, "learning_rate": 3.8926951717356006e-05, "loss": 0.0, "step": 34242 }, { "epoch": 0.66, "learning_rate": 3.8926304757098775e-05, "loss": 0.0001, "step": 34244 }, { "epoch": 0.66, "learning_rate": 3.8925657796841544e-05, "loss": 0.0, "step": 34246 }, { "epoch": 0.66, "learning_rate": 3.8925010836584306e-05, "loss": 0.0157, "step": 34248 }, { "epoch": 0.66, "learning_rate": 3.892436387632708e-05, "loss": 0.0006, "step": 34250 }, { "epoch": 0.66, "learning_rate": 3.8923716916069844e-05, "loss": 0.0002, "step": 34252 }, { "epoch": 0.66, "learning_rate": 3.892306995581262e-05, "loss": 0.0, "step": 34254 }, { "epoch": 0.66, "learning_rate": 3.892242299555538e-05, "loss": 0.0005, "step": 34256 }, { "epoch": 0.66, "learning_rate": 3.892177603529815e-05, "loss": 0.0001, "step": 34258 }, { "epoch": 0.66, "learning_rate": 3.892112907504093e-05, "loss": 0.0, "step": 34260 }, { "epoch": 0.66, "learning_rate": 3.892048211478369e-05, "loss": 0.0207, "step": 34262 }, { "epoch": 0.67, "learning_rate": 3.891983515452646e-05, "loss": 0.0, "step": 34264 }, { "epoch": 0.67, "learning_rate": 3.891918819426923e-05, "loss": 0.0021, "step": 34266 }, { "epoch": 0.67, "learning_rate": 3.8918541234012e-05, "loss": 0.0206, "step": 34268 }, { "epoch": 0.67, "learning_rate": 3.891789427375476e-05, "loss": 0.0072, "step": 34270 }, { "epoch": 0.67, "learning_rate": 3.8917247313497535e-05, "loss": 0.0017, "step": 34272 }, { "epoch": 0.67, "learning_rate": 3.8916600353240304e-05, "loss": 0.0, "step": 34274 }, { "epoch": 0.67, "learning_rate": 3.891595339298307e-05, "loss": 0.0062, "step": 34276 }, { "epoch": 0.67, "learning_rate": 3.891530643272584e-05, "loss": 0.009, "step": 34278 }, { "epoch": 0.67, "learning_rate": 3.8914659472468605e-05, "loss": 0.0, "step": 34280 }, { "epoch": 0.67, "learning_rate": 3.891401251221138e-05, "loss": 0.0, "step": 34282 }, { "epoch": 0.67, "learning_rate": 3.891336555195414e-05, "loss": 0.0, "step": 34284 }, { "epoch": 0.67, "learning_rate": 3.891271859169691e-05, "loss": 0.0179, "step": 34286 }, { "epoch": 0.67, "learning_rate": 3.891207163143968e-05, "loss": 0.0, "step": 34288 }, { "epoch": 0.67, "learning_rate": 3.891142467118245e-05, "loss": 0.0005, "step": 34290 }, { "epoch": 0.67, "learning_rate": 3.891077771092522e-05, "loss": 0.0048, "step": 34292 }, { "epoch": 0.67, "learning_rate": 3.891013075066799e-05, "loss": 0.0022, "step": 34294 }, { "epoch": 0.67, "learning_rate": 3.890948379041076e-05, "loss": 0.0, "step": 34296 }, { "epoch": 0.67, "learning_rate": 3.8908836830153526e-05, "loss": 0.0, "step": 34298 }, { "epoch": 0.67, "learning_rate": 3.8908189869896296e-05, "loss": 0.0, "step": 34300 }, { "epoch": 0.67, "learning_rate": 3.890754290963906e-05, "loss": 0.0003, "step": 34302 }, { "epoch": 0.67, "learning_rate": 3.8906895949381834e-05, "loss": 0.0004, "step": 34304 }, { "epoch": 0.67, "learning_rate": 3.89062489891246e-05, "loss": 0.0, "step": 34306 }, { "epoch": 0.67, "learning_rate": 3.8905602028867365e-05, "loss": 0.0, "step": 34308 }, { "epoch": 0.67, "learning_rate": 3.890495506861014e-05, "loss": 0.0084, "step": 34310 }, { "epoch": 0.67, "learning_rate": 3.89043081083529e-05, "loss": 0.0, "step": 34312 }, { "epoch": 0.67, "learning_rate": 3.890366114809568e-05, "loss": 0.0001, "step": 34314 }, { "epoch": 0.67, "learning_rate": 3.890301418783844e-05, "loss": 0.0024, "step": 34316 }, { "epoch": 0.67, "learning_rate": 3.890236722758121e-05, "loss": 0.0, "step": 34318 }, { "epoch": 0.67, "learning_rate": 3.890172026732398e-05, "loss": 0.0, "step": 34320 }, { "epoch": 0.67, "learning_rate": 3.890107330706675e-05, "loss": 0.0, "step": 34322 }, { "epoch": 0.67, "learning_rate": 3.890042634680952e-05, "loss": 0.0, "step": 34324 }, { "epoch": 0.67, "learning_rate": 3.889977938655229e-05, "loss": 0.0001, "step": 34326 }, { "epoch": 0.67, "learning_rate": 3.8899132426295056e-05, "loss": 0.0006, "step": 34328 }, { "epoch": 0.67, "learning_rate": 3.889848546603782e-05, "loss": 0.0, "step": 34330 }, { "epoch": 0.67, "learning_rate": 3.8897838505780594e-05, "loss": 0.0014, "step": 34332 }, { "epoch": 0.67, "learning_rate": 3.8897191545523356e-05, "loss": 0.0256, "step": 34334 }, { "epoch": 0.67, "learning_rate": 3.889654458526613e-05, "loss": 0.0007, "step": 34336 }, { "epoch": 0.67, "learning_rate": 3.88958976250089e-05, "loss": 0.0, "step": 34338 }, { "epoch": 0.67, "learning_rate": 3.8895250664751664e-05, "loss": 0.0, "step": 34340 }, { "epoch": 0.67, "learning_rate": 3.889460370449444e-05, "loss": 0.0007, "step": 34342 }, { "epoch": 0.67, "learning_rate": 3.88939567442372e-05, "loss": 0.0, "step": 34344 }, { "epoch": 0.67, "learning_rate": 3.889330978397997e-05, "loss": 0.0001, "step": 34346 }, { "epoch": 0.67, "learning_rate": 3.889266282372274e-05, "loss": 0.0063, "step": 34348 }, { "epoch": 0.67, "learning_rate": 3.889201586346551e-05, "loss": 0.0284, "step": 34350 }, { "epoch": 0.67, "learning_rate": 3.889136890320828e-05, "loss": 0.0015, "step": 34352 }, { "epoch": 0.67, "learning_rate": 3.889072194295105e-05, "loss": 0.0005, "step": 34354 }, { "epoch": 0.67, "learning_rate": 3.8890074982693816e-05, "loss": 0.0088, "step": 34356 }, { "epoch": 0.67, "learning_rate": 3.8889428022436585e-05, "loss": 0.0108, "step": 34358 }, { "epoch": 0.67, "learning_rate": 3.8888781062179355e-05, "loss": 0.0, "step": 34360 }, { "epoch": 0.67, "learning_rate": 3.888813410192212e-05, "loss": 0.0, "step": 34362 }, { "epoch": 0.67, "learning_rate": 3.888748714166489e-05, "loss": 0.0, "step": 34364 }, { "epoch": 0.67, "learning_rate": 3.8886840181407655e-05, "loss": 0.0024, "step": 34366 }, { "epoch": 0.67, "learning_rate": 3.8886193221150424e-05, "loss": 0.0222, "step": 34368 }, { "epoch": 0.67, "learning_rate": 3.888554626089319e-05, "loss": 0.0, "step": 34370 }, { "epoch": 0.67, "learning_rate": 3.888489930063596e-05, "loss": 0.0283, "step": 34372 }, { "epoch": 0.67, "learning_rate": 3.888425234037874e-05, "loss": 0.0006, "step": 34374 }, { "epoch": 0.67, "learning_rate": 3.88836053801215e-05, "loss": 0.0, "step": 34376 }, { "epoch": 0.67, "learning_rate": 3.888295841986427e-05, "loss": 0.0007, "step": 34378 }, { "epoch": 0.67, "learning_rate": 3.888231145960704e-05, "loss": 0.0, "step": 34380 }, { "epoch": 0.67, "learning_rate": 3.888166449934981e-05, "loss": 0.0113, "step": 34382 }, { "epoch": 0.67, "learning_rate": 3.888101753909258e-05, "loss": 0.0, "step": 34384 }, { "epoch": 0.67, "learning_rate": 3.8880370578835346e-05, "loss": 0.0004, "step": 34386 }, { "epoch": 0.67, "learning_rate": 3.8879723618578115e-05, "loss": 0.0, "step": 34388 }, { "epoch": 0.67, "learning_rate": 3.887907665832088e-05, "loss": 0.0, "step": 34390 }, { "epoch": 0.67, "learning_rate": 3.887842969806365e-05, "loss": 0.0001, "step": 34392 }, { "epoch": 0.67, "learning_rate": 3.8877782737806415e-05, "loss": 0.0004, "step": 34394 }, { "epoch": 0.67, "learning_rate": 3.887713577754919e-05, "loss": 0.0068, "step": 34396 }, { "epoch": 0.67, "learning_rate": 3.8876488817291954e-05, "loss": 0.0065, "step": 34398 }, { "epoch": 0.67, "learning_rate": 3.887584185703472e-05, "loss": 0.0, "step": 34400 }, { "epoch": 0.67, "learning_rate": 3.887519489677749e-05, "loss": 0.0, "step": 34402 }, { "epoch": 0.67, "learning_rate": 3.887454793652026e-05, "loss": 0.0163, "step": 34404 }, { "epoch": 0.67, "learning_rate": 3.887390097626303e-05, "loss": 0.0038, "step": 34406 }, { "epoch": 0.67, "learning_rate": 3.88732540160058e-05, "loss": 0.0, "step": 34408 }, { "epoch": 0.67, "learning_rate": 3.887260705574857e-05, "loss": 0.0051, "step": 34410 }, { "epoch": 0.67, "learning_rate": 3.887196009549133e-05, "loss": 0.0001, "step": 34412 }, { "epoch": 0.67, "learning_rate": 3.8871313135234106e-05, "loss": 0.0001, "step": 34414 }, { "epoch": 0.67, "learning_rate": 3.887066617497687e-05, "loss": 0.0087, "step": 34416 }, { "epoch": 0.67, "learning_rate": 3.8870019214719644e-05, "loss": 0.001, "step": 34418 }, { "epoch": 0.67, "learning_rate": 3.8869372254462414e-05, "loss": 0.0, "step": 34420 }, { "epoch": 0.67, "learning_rate": 3.8868725294205176e-05, "loss": 0.005, "step": 34422 }, { "epoch": 0.67, "learning_rate": 3.886807833394795e-05, "loss": 0.0, "step": 34424 }, { "epoch": 0.67, "learning_rate": 3.8867431373690714e-05, "loss": 0.0002, "step": 34426 }, { "epoch": 0.67, "learning_rate": 3.886678441343348e-05, "loss": 0.0, "step": 34428 }, { "epoch": 0.67, "learning_rate": 3.886613745317625e-05, "loss": 0.0004, "step": 34430 }, { "epoch": 0.67, "learning_rate": 3.886549049291902e-05, "loss": 0.0212, "step": 34432 }, { "epoch": 0.67, "learning_rate": 3.886484353266179e-05, "loss": 0.0002, "step": 34434 }, { "epoch": 0.67, "learning_rate": 3.886419657240456e-05, "loss": 0.0001, "step": 34436 }, { "epoch": 0.67, "learning_rate": 3.886354961214733e-05, "loss": 0.0001, "step": 34438 }, { "epoch": 0.67, "learning_rate": 3.88629026518901e-05, "loss": 0.0011, "step": 34440 }, { "epoch": 0.67, "learning_rate": 3.886225569163287e-05, "loss": 0.0, "step": 34442 }, { "epoch": 0.67, "learning_rate": 3.886160873137563e-05, "loss": 0.0, "step": 34444 }, { "epoch": 0.67, "learning_rate": 3.8860961771118405e-05, "loss": 0.0005, "step": 34446 }, { "epoch": 0.67, "learning_rate": 3.886031481086117e-05, "loss": 0.0001, "step": 34448 }, { "epoch": 0.67, "learning_rate": 3.8859667850603936e-05, "loss": 0.0001, "step": 34450 }, { "epoch": 0.67, "learning_rate": 3.885902089034671e-05, "loss": 0.0003, "step": 34452 }, { "epoch": 0.67, "learning_rate": 3.8858373930089474e-05, "loss": 0.0, "step": 34454 }, { "epoch": 0.67, "learning_rate": 3.885772696983225e-05, "loss": 0.0008, "step": 34456 }, { "epoch": 0.67, "learning_rate": 3.885708000957501e-05, "loss": 0.0, "step": 34458 }, { "epoch": 0.67, "learning_rate": 3.885643304931778e-05, "loss": 0.0002, "step": 34460 }, { "epoch": 0.67, "learning_rate": 3.885578608906055e-05, "loss": 0.0, "step": 34462 }, { "epoch": 0.67, "learning_rate": 3.885513912880332e-05, "loss": 0.0, "step": 34464 }, { "epoch": 0.67, "learning_rate": 3.885449216854609e-05, "loss": 0.0, "step": 34466 }, { "epoch": 0.67, "learning_rate": 3.885384520828886e-05, "loss": 0.0, "step": 34468 }, { "epoch": 0.67, "learning_rate": 3.885319824803163e-05, "loss": 0.0, "step": 34470 }, { "epoch": 0.67, "learning_rate": 3.885255128777439e-05, "loss": 0.0001, "step": 34472 }, { "epoch": 0.67, "learning_rate": 3.8851904327517165e-05, "loss": 0.0, "step": 34474 }, { "epoch": 0.67, "learning_rate": 3.885125736725993e-05, "loss": 0.0023, "step": 34476 }, { "epoch": 0.67, "learning_rate": 3.8850610407002703e-05, "loss": 0.0001, "step": 34478 }, { "epoch": 0.67, "learning_rate": 3.8849963446745466e-05, "loss": 0.0002, "step": 34480 }, { "epoch": 0.67, "learning_rate": 3.8849316486488235e-05, "loss": 0.0007, "step": 34482 }, { "epoch": 0.67, "learning_rate": 3.884866952623101e-05, "loss": 0.0018, "step": 34484 }, { "epoch": 0.67, "learning_rate": 3.884802256597377e-05, "loss": 0.0007, "step": 34486 }, { "epoch": 0.67, "learning_rate": 3.884737560571654e-05, "loss": 0.0, "step": 34488 }, { "epoch": 0.67, "learning_rate": 3.884672864545931e-05, "loss": 0.0002, "step": 34490 }, { "epoch": 0.67, "learning_rate": 3.884608168520208e-05, "loss": 0.0001, "step": 34492 }, { "epoch": 0.67, "learning_rate": 3.884543472494484e-05, "loss": 0.0, "step": 34494 }, { "epoch": 0.67, "learning_rate": 3.884478776468762e-05, "loss": 0.0, "step": 34496 }, { "epoch": 0.67, "learning_rate": 3.884414080443039e-05, "loss": 0.0001, "step": 34498 }, { "epoch": 0.67, "learning_rate": 3.8843493844173157e-05, "loss": 0.0, "step": 34500 }, { "epoch": 0.67, "learning_rate": 3.8842846883915926e-05, "loss": 0.0, "step": 34502 }, { "epoch": 0.67, "learning_rate": 3.884219992365869e-05, "loss": 0.0, "step": 34504 }, { "epoch": 0.67, "learning_rate": 3.8841552963401464e-05, "loss": 0.0002, "step": 34506 }, { "epoch": 0.67, "learning_rate": 3.8840906003144226e-05, "loss": 0.0001, "step": 34508 }, { "epoch": 0.67, "learning_rate": 3.8840259042886995e-05, "loss": 0.0, "step": 34510 }, { "epoch": 0.67, "learning_rate": 3.8839612082629764e-05, "loss": 0.0, "step": 34512 }, { "epoch": 0.67, "learning_rate": 3.883896512237253e-05, "loss": 0.0001, "step": 34514 }, { "epoch": 0.67, "learning_rate": 3.88383181621153e-05, "loss": 0.0, "step": 34516 }, { "epoch": 0.67, "learning_rate": 3.883767120185807e-05, "loss": 0.0001, "step": 34518 }, { "epoch": 0.67, "learning_rate": 3.883702424160084e-05, "loss": 0.0, "step": 34520 }, { "epoch": 0.67, "learning_rate": 3.883637728134361e-05, "loss": 0.0001, "step": 34522 }, { "epoch": 0.67, "learning_rate": 3.883573032108638e-05, "loss": 0.0003, "step": 34524 }, { "epoch": 0.67, "learning_rate": 3.883508336082914e-05, "loss": 0.0, "step": 34526 }, { "epoch": 0.67, "learning_rate": 3.883443640057192e-05, "loss": 0.0, "step": 34528 }, { "epoch": 0.67, "learning_rate": 3.8833789440314686e-05, "loss": 0.0, "step": 34530 }, { "epoch": 0.67, "learning_rate": 3.883314248005745e-05, "loss": 0.0001, "step": 34532 }, { "epoch": 0.67, "learning_rate": 3.8832495519800224e-05, "loss": 0.0, "step": 34534 }, { "epoch": 0.67, "learning_rate": 3.8831848559542987e-05, "loss": 0.0002, "step": 34536 }, { "epoch": 0.67, "learning_rate": 3.883120159928576e-05, "loss": 0.0064, "step": 34538 }, { "epoch": 0.67, "learning_rate": 3.8830554639028525e-05, "loss": 0.0, "step": 34540 }, { "epoch": 0.67, "learning_rate": 3.8829907678771294e-05, "loss": 0.0, "step": 34542 }, { "epoch": 0.67, "learning_rate": 3.882926071851406e-05, "loss": 0.0059, "step": 34544 }, { "epoch": 0.67, "learning_rate": 3.882861375825683e-05, "loss": 0.0001, "step": 34546 }, { "epoch": 0.67, "learning_rate": 3.88279667979996e-05, "loss": 0.0019, "step": 34548 }, { "epoch": 0.67, "learning_rate": 3.882731983774237e-05, "loss": 0.0022, "step": 34550 }, { "epoch": 0.67, "learning_rate": 3.882667287748514e-05, "loss": 0.0001, "step": 34552 }, { "epoch": 0.67, "learning_rate": 3.88260259172279e-05, "loss": 0.0048, "step": 34554 }, { "epoch": 0.67, "learning_rate": 3.882537895697068e-05, "loss": 0.0052, "step": 34556 }, { "epoch": 0.67, "learning_rate": 3.882473199671344e-05, "loss": 0.0001, "step": 34558 }, { "epoch": 0.67, "learning_rate": 3.8824085036456216e-05, "loss": 0.0001, "step": 34560 }, { "epoch": 0.67, "learning_rate": 3.8823438076198985e-05, "loss": 0.0004, "step": 34562 }, { "epoch": 0.67, "learning_rate": 3.882279111594175e-05, "loss": 0.0001, "step": 34564 }, { "epoch": 0.67, "learning_rate": 3.882214415568452e-05, "loss": 0.0052, "step": 34566 }, { "epoch": 0.67, "learning_rate": 3.8821497195427285e-05, "loss": 0.0001, "step": 34568 }, { "epoch": 0.67, "learning_rate": 3.8820850235170054e-05, "loss": 0.0001, "step": 34570 }, { "epoch": 0.67, "learning_rate": 3.882020327491282e-05, "loss": 0.0002, "step": 34572 }, { "epoch": 0.67, "learning_rate": 3.881955631465559e-05, "loss": 0.0, "step": 34574 }, { "epoch": 0.67, "learning_rate": 3.881890935439836e-05, "loss": 0.0019, "step": 34576 }, { "epoch": 0.67, "learning_rate": 3.881826239414113e-05, "loss": 0.0149, "step": 34578 }, { "epoch": 0.67, "learning_rate": 3.88176154338839e-05, "loss": 0.0046, "step": 34580 }, { "epoch": 0.67, "learning_rate": 3.881696847362667e-05, "loss": 0.0001, "step": 34582 }, { "epoch": 0.67, "learning_rate": 3.881632151336944e-05, "loss": 0.008, "step": 34584 }, { "epoch": 0.67, "learning_rate": 3.88156745531122e-05, "loss": 0.0005, "step": 34586 }, { "epoch": 0.67, "learning_rate": 3.8815027592854976e-05, "loss": 0.0103, "step": 34588 }, { "epoch": 0.67, "learning_rate": 3.881438063259774e-05, "loss": 0.0042, "step": 34590 }, { "epoch": 0.67, "learning_rate": 3.881373367234051e-05, "loss": 0.0, "step": 34592 }, { "epoch": 0.67, "learning_rate": 3.8813086712083276e-05, "loss": 0.0, "step": 34594 }, { "epoch": 0.67, "learning_rate": 3.8812439751826045e-05, "loss": 0.0002, "step": 34596 }, { "epoch": 0.67, "learning_rate": 3.881179279156882e-05, "loss": 0.008, "step": 34598 }, { "epoch": 0.67, "learning_rate": 3.8811145831311584e-05, "loss": 0.0, "step": 34600 }, { "epoch": 0.67, "learning_rate": 3.881049887105435e-05, "loss": 0.0002, "step": 34602 }, { "epoch": 0.67, "learning_rate": 3.880985191079712e-05, "loss": 0.0, "step": 34604 }, { "epoch": 0.67, "learning_rate": 3.880920495053989e-05, "loss": 0.0, "step": 34606 }, { "epoch": 0.67, "learning_rate": 3.880855799028266e-05, "loss": 0.0041, "step": 34608 }, { "epoch": 0.67, "learning_rate": 3.880791103002543e-05, "loss": 0.0, "step": 34610 }, { "epoch": 0.67, "learning_rate": 3.88072640697682e-05, "loss": 0.0038, "step": 34612 }, { "epoch": 0.67, "learning_rate": 3.880661710951096e-05, "loss": 0.0113, "step": 34614 }, { "epoch": 0.67, "learning_rate": 3.8805970149253736e-05, "loss": 0.0001, "step": 34616 }, { "epoch": 0.67, "learning_rate": 3.88053231889965e-05, "loss": 0.0002, "step": 34618 }, { "epoch": 0.67, "learning_rate": 3.8804676228739275e-05, "loss": 0.011, "step": 34620 }, { "epoch": 0.67, "learning_rate": 3.880402926848204e-05, "loss": 0.0034, "step": 34622 }, { "epoch": 0.67, "learning_rate": 3.8803382308224806e-05, "loss": 0.0, "step": 34624 }, { "epoch": 0.67, "learning_rate": 3.8802735347967575e-05, "loss": 0.0, "step": 34626 }, { "epoch": 0.67, "learning_rate": 3.8802088387710344e-05, "loss": 0.0093, "step": 34628 }, { "epoch": 0.67, "learning_rate": 3.880144142745311e-05, "loss": 0.0082, "step": 34630 }, { "epoch": 0.67, "learning_rate": 3.880079446719588e-05, "loss": 0.0049, "step": 34632 }, { "epoch": 0.67, "learning_rate": 3.880014750693865e-05, "loss": 0.0, "step": 34634 }, { "epoch": 0.67, "learning_rate": 3.8799500546681414e-05, "loss": 0.0, "step": 34636 }, { "epoch": 0.67, "learning_rate": 3.879885358642419e-05, "loss": 0.0003, "step": 34638 }, { "epoch": 0.67, "learning_rate": 3.879820662616696e-05, "loss": 0.0, "step": 34640 }, { "epoch": 0.67, "learning_rate": 3.879755966590973e-05, "loss": 0.0001, "step": 34642 }, { "epoch": 0.67, "learning_rate": 3.87969127056525e-05, "loss": 0.0, "step": 34644 }, { "epoch": 0.67, "learning_rate": 3.879626574539526e-05, "loss": 0.0, "step": 34646 }, { "epoch": 0.67, "learning_rate": 3.8795618785138035e-05, "loss": 0.0, "step": 34648 }, { "epoch": 0.67, "learning_rate": 3.87949718248808e-05, "loss": 0.0001, "step": 34650 }, { "epoch": 0.67, "learning_rate": 3.8794324864623566e-05, "loss": 0.0011, "step": 34652 }, { "epoch": 0.67, "learning_rate": 3.8793677904366335e-05, "loss": 0.0, "step": 34654 }, { "epoch": 0.67, "learning_rate": 3.8793030944109104e-05, "loss": 0.0001, "step": 34656 }, { "epoch": 0.67, "learning_rate": 3.8792383983851874e-05, "loss": 0.0, "step": 34658 }, { "epoch": 0.67, "learning_rate": 3.879173702359464e-05, "loss": 0.0, "step": 34660 }, { "epoch": 0.67, "learning_rate": 3.879109006333741e-05, "loss": 0.0, "step": 34662 }, { "epoch": 0.67, "learning_rate": 3.879044310308018e-05, "loss": 0.0, "step": 34664 }, { "epoch": 0.67, "learning_rate": 3.878979614282295e-05, "loss": 0.0, "step": 34666 }, { "epoch": 0.67, "learning_rate": 3.878914918256571e-05, "loss": 0.0, "step": 34668 }, { "epoch": 0.67, "learning_rate": 3.878850222230849e-05, "loss": 0.0092, "step": 34670 }, { "epoch": 0.67, "learning_rate": 3.878785526205125e-05, "loss": 0.0001, "step": 34672 }, { "epoch": 0.67, "learning_rate": 3.878720830179402e-05, "loss": 0.0212, "step": 34674 }, { "epoch": 0.67, "learning_rate": 3.8786561341536795e-05, "loss": 0.0001, "step": 34676 }, { "epoch": 0.67, "learning_rate": 3.878591438127956e-05, "loss": 0.0039, "step": 34678 }, { "epoch": 0.67, "learning_rate": 3.8785267421022333e-05, "loss": 0.0, "step": 34680 }, { "epoch": 0.67, "learning_rate": 3.8784620460765096e-05, "loss": 0.0017, "step": 34682 }, { "epoch": 0.67, "learning_rate": 3.8783973500507865e-05, "loss": 0.0087, "step": 34684 }, { "epoch": 0.67, "learning_rate": 3.8783326540250634e-05, "loss": 0.0521, "step": 34686 }, { "epoch": 0.67, "learning_rate": 3.87826795799934e-05, "loss": 0.0002, "step": 34688 }, { "epoch": 0.67, "learning_rate": 3.878203261973617e-05, "loss": 0.0001, "step": 34690 }, { "epoch": 0.67, "learning_rate": 3.878138565947894e-05, "loss": 0.0004, "step": 34692 }, { "epoch": 0.67, "learning_rate": 3.878073869922171e-05, "loss": 0.0002, "step": 34694 }, { "epoch": 0.67, "learning_rate": 3.878009173896447e-05, "loss": 0.0021, "step": 34696 }, { "epoch": 0.67, "learning_rate": 3.877944477870725e-05, "loss": 0.0, "step": 34698 }, { "epoch": 0.67, "learning_rate": 3.877879781845001e-05, "loss": 0.0022, "step": 34700 }, { "epoch": 0.67, "learning_rate": 3.877815085819279e-05, "loss": 0.0001, "step": 34702 }, { "epoch": 0.67, "learning_rate": 3.877750389793555e-05, "loss": 0.0, "step": 34704 }, { "epoch": 0.67, "learning_rate": 3.877685693767832e-05, "loss": 0.0009, "step": 34706 }, { "epoch": 0.67, "learning_rate": 3.8776209977421094e-05, "loss": 0.0001, "step": 34708 }, { "epoch": 0.67, "learning_rate": 3.8775563017163856e-05, "loss": 0.0004, "step": 34710 }, { "epoch": 0.67, "learning_rate": 3.8774916056906625e-05, "loss": 0.0005, "step": 34712 }, { "epoch": 0.67, "learning_rate": 3.8774269096649394e-05, "loss": 0.0, "step": 34714 }, { "epoch": 0.67, "learning_rate": 3.8773622136392163e-05, "loss": 0.0, "step": 34716 }, { "epoch": 0.67, "learning_rate": 3.8772975176134926e-05, "loss": 0.0001, "step": 34718 }, { "epoch": 0.67, "learning_rate": 3.87723282158777e-05, "loss": 0.0003, "step": 34720 }, { "epoch": 0.67, "learning_rate": 3.877168125562047e-05, "loss": 0.0006, "step": 34722 }, { "epoch": 0.67, "learning_rate": 3.877103429536324e-05, "loss": 0.0001, "step": 34724 }, { "epoch": 0.67, "learning_rate": 3.877038733510601e-05, "loss": 0.0046, "step": 34726 }, { "epoch": 0.67, "learning_rate": 3.876974037484877e-05, "loss": 0.0, "step": 34728 }, { "epoch": 0.67, "learning_rate": 3.876909341459155e-05, "loss": 0.0008, "step": 34730 }, { "epoch": 0.67, "learning_rate": 3.876844645433431e-05, "loss": 0.0009, "step": 34732 }, { "epoch": 0.67, "learning_rate": 3.876779949407708e-05, "loss": 0.0, "step": 34734 }, { "epoch": 0.67, "learning_rate": 3.876715253381985e-05, "loss": 0.0, "step": 34736 }, { "epoch": 0.67, "learning_rate": 3.8766505573562617e-05, "loss": 0.0007, "step": 34738 }, { "epoch": 0.67, "learning_rate": 3.876585861330539e-05, "loss": 0.0, "step": 34740 }, { "epoch": 0.67, "learning_rate": 3.8765211653048155e-05, "loss": 0.0002, "step": 34742 }, { "epoch": 0.67, "learning_rate": 3.8764564692790924e-05, "loss": 0.0015, "step": 34744 }, { "epoch": 0.67, "learning_rate": 3.876391773253369e-05, "loss": 0.0, "step": 34746 }, { "epoch": 0.67, "learning_rate": 3.876327077227646e-05, "loss": 0.0, "step": 34748 }, { "epoch": 0.67, "learning_rate": 3.8762623812019224e-05, "loss": 0.0, "step": 34750 }, { "epoch": 0.67, "learning_rate": 3.8761976851762e-05, "loss": 0.0083, "step": 34752 }, { "epoch": 0.67, "learning_rate": 3.876132989150477e-05, "loss": 0.0002, "step": 34754 }, { "epoch": 0.67, "learning_rate": 3.876068293124753e-05, "loss": 0.0043, "step": 34756 }, { "epoch": 0.67, "learning_rate": 3.876003597099031e-05, "loss": 0.0059, "step": 34758 }, { "epoch": 0.67, "learning_rate": 3.875938901073307e-05, "loss": 0.0, "step": 34760 }, { "epoch": 0.67, "learning_rate": 3.8758742050475846e-05, "loss": 0.0, "step": 34762 }, { "epoch": 0.67, "learning_rate": 3.875809509021861e-05, "loss": 0.0022, "step": 34764 }, { "epoch": 0.67, "learning_rate": 3.875744812996138e-05, "loss": 0.0034, "step": 34766 }, { "epoch": 0.67, "learning_rate": 3.8756801169704146e-05, "loss": 0.0002, "step": 34768 }, { "epoch": 0.67, "learning_rate": 3.8756154209446915e-05, "loss": 0.0006, "step": 34770 }, { "epoch": 0.67, "learning_rate": 3.8755507249189684e-05, "loss": 0.0022, "step": 34772 }, { "epoch": 0.67, "learning_rate": 3.875486028893245e-05, "loss": 0.0001, "step": 34774 }, { "epoch": 0.67, "learning_rate": 3.875421332867522e-05, "loss": 0.0, "step": 34776 }, { "epoch": 0.67, "learning_rate": 3.8753566368417985e-05, "loss": 0.0, "step": 34778 }, { "epoch": 0.68, "learning_rate": 3.875291940816076e-05, "loss": 0.0047, "step": 34780 }, { "epoch": 0.68, "learning_rate": 3.875227244790352e-05, "loss": 0.0, "step": 34782 }, { "epoch": 0.68, "learning_rate": 3.87516254876463e-05, "loss": 0.0001, "step": 34784 }, { "epoch": 0.68, "learning_rate": 3.875097852738907e-05, "loss": 0.0038, "step": 34786 }, { "epoch": 0.68, "learning_rate": 3.875033156713183e-05, "loss": 0.0001, "step": 34788 }, { "epoch": 0.68, "learning_rate": 3.8749684606874606e-05, "loss": 0.0028, "step": 34790 }, { "epoch": 0.68, "learning_rate": 3.874903764661737e-05, "loss": 0.0, "step": 34792 }, { "epoch": 0.68, "learning_rate": 3.874839068636014e-05, "loss": 0.0, "step": 34794 }, { "epoch": 0.68, "learning_rate": 3.8747743726102906e-05, "loss": 0.0019, "step": 34796 }, { "epoch": 0.68, "learning_rate": 3.8747096765845676e-05, "loss": 0.0069, "step": 34798 }, { "epoch": 0.68, "learning_rate": 3.8746449805588445e-05, "loss": 0.0027, "step": 34800 }, { "epoch": 0.68, "learning_rate": 3.8745802845331214e-05, "loss": 0.0, "step": 34802 }, { "epoch": 0.68, "learning_rate": 3.874515588507398e-05, "loss": 0.0, "step": 34804 }, { "epoch": 0.68, "learning_rate": 3.874450892481675e-05, "loss": 0.0, "step": 34806 }, { "epoch": 0.68, "learning_rate": 3.874386196455952e-05, "loss": 0.0, "step": 34808 }, { "epoch": 0.68, "learning_rate": 3.874321500430228e-05, "loss": 0.0005, "step": 34810 }, { "epoch": 0.68, "learning_rate": 3.874256804404506e-05, "loss": 0.0001, "step": 34812 }, { "epoch": 0.68, "learning_rate": 3.874192108378782e-05, "loss": 0.0, "step": 34814 }, { "epoch": 0.68, "learning_rate": 3.874127412353059e-05, "loss": 0.0001, "step": 34816 }, { "epoch": 0.68, "learning_rate": 3.8740627163273366e-05, "loss": 0.0, "step": 34818 }, { "epoch": 0.68, "learning_rate": 3.873998020301613e-05, "loss": 0.0, "step": 34820 }, { "epoch": 0.68, "learning_rate": 3.8739333242758905e-05, "loss": 0.0, "step": 34822 }, { "epoch": 0.68, "learning_rate": 3.873868628250167e-05, "loss": 0.0, "step": 34824 }, { "epoch": 0.68, "learning_rate": 3.8738039322244436e-05, "loss": 0.0003, "step": 34826 }, { "epoch": 0.68, "learning_rate": 3.8737392361987205e-05, "loss": 0.0, "step": 34828 }, { "epoch": 0.68, "learning_rate": 3.8736745401729974e-05, "loss": 0.0003, "step": 34830 }, { "epoch": 0.68, "learning_rate": 3.873609844147274e-05, "loss": 0.0003, "step": 34832 }, { "epoch": 0.68, "learning_rate": 3.873545148121551e-05, "loss": 0.0028, "step": 34834 }, { "epoch": 0.68, "learning_rate": 3.873480452095828e-05, "loss": 0.0001, "step": 34836 }, { "epoch": 0.68, "learning_rate": 3.8734157560701044e-05, "loss": 0.0, "step": 34838 }, { "epoch": 0.68, "learning_rate": 3.873351060044382e-05, "loss": 0.0348, "step": 34840 }, { "epoch": 0.68, "learning_rate": 3.873286364018658e-05, "loss": 0.0005, "step": 34842 }, { "epoch": 0.68, "learning_rate": 3.873221667992936e-05, "loss": 0.0, "step": 34844 }, { "epoch": 0.68, "learning_rate": 3.873156971967212e-05, "loss": 0.0004, "step": 34846 }, { "epoch": 0.68, "learning_rate": 3.873092275941489e-05, "loss": 0.0135, "step": 34848 }, { "epoch": 0.68, "learning_rate": 3.873027579915766e-05, "loss": 0.0038, "step": 34850 }, { "epoch": 0.68, "learning_rate": 3.872962883890043e-05, "loss": 0.0031, "step": 34852 }, { "epoch": 0.68, "learning_rate": 3.8728981878643196e-05, "loss": 0.0001, "step": 34854 }, { "epoch": 0.68, "learning_rate": 3.8728334918385965e-05, "loss": 0.0, "step": 34856 }, { "epoch": 0.68, "learning_rate": 3.8727687958128735e-05, "loss": 0.0019, "step": 34858 }, { "epoch": 0.68, "learning_rate": 3.87270409978715e-05, "loss": 0.0, "step": 34860 }, { "epoch": 0.68, "learning_rate": 3.872639403761427e-05, "loss": 0.0001, "step": 34862 }, { "epoch": 0.68, "learning_rate": 3.872574707735704e-05, "loss": 0.0001, "step": 34864 }, { "epoch": 0.68, "learning_rate": 3.872510011709981e-05, "loss": 0.0, "step": 34866 }, { "epoch": 0.68, "learning_rate": 3.872445315684258e-05, "loss": 0.0002, "step": 34868 }, { "epoch": 0.68, "learning_rate": 3.872380619658534e-05, "loss": 0.0004, "step": 34870 }, { "epoch": 0.68, "learning_rate": 3.872315923632812e-05, "loss": 0.01, "step": 34872 }, { "epoch": 0.68, "learning_rate": 3.872251227607088e-05, "loss": 0.0001, "step": 34874 }, { "epoch": 0.68, "learning_rate": 3.872186531581365e-05, "loss": 0.0004, "step": 34876 }, { "epoch": 0.68, "learning_rate": 3.872121835555642e-05, "loss": 0.0, "step": 34878 }, { "epoch": 0.68, "learning_rate": 3.872057139529919e-05, "loss": 0.0001, "step": 34880 }, { "epoch": 0.68, "learning_rate": 3.871992443504196e-05, "loss": 0.0029, "step": 34882 }, { "epoch": 0.68, "learning_rate": 3.8719277474784726e-05, "loss": 0.0007, "step": 34884 }, { "epoch": 0.68, "learning_rate": 3.8718630514527495e-05, "loss": 0.0, "step": 34886 }, { "epoch": 0.68, "learning_rate": 3.8717983554270264e-05, "loss": 0.0, "step": 34888 }, { "epoch": 0.68, "learning_rate": 3.871733659401303e-05, "loss": 0.0176, "step": 34890 }, { "epoch": 0.68, "learning_rate": 3.8716689633755795e-05, "loss": 0.0007, "step": 34892 }, { "epoch": 0.68, "learning_rate": 3.871604267349857e-05, "loss": 0.0079, "step": 34894 }, { "epoch": 0.68, "learning_rate": 3.8715395713241334e-05, "loss": 0.0001, "step": 34896 }, { "epoch": 0.68, "learning_rate": 3.87147487529841e-05, "loss": 0.0001, "step": 34898 }, { "epoch": 0.68, "learning_rate": 3.871410179272688e-05, "loss": 0.0094, "step": 34900 }, { "epoch": 0.68, "learning_rate": 3.871345483246964e-05, "loss": 0.0002, "step": 34902 }, { "epoch": 0.68, "learning_rate": 3.871280787221242e-05, "loss": 0.0, "step": 34904 }, { "epoch": 0.68, "learning_rate": 3.871216091195518e-05, "loss": 0.0, "step": 34906 }, { "epoch": 0.68, "learning_rate": 3.871151395169795e-05, "loss": 0.0001, "step": 34908 }, { "epoch": 0.68, "learning_rate": 3.871086699144072e-05, "loss": 0.0, "step": 34910 }, { "epoch": 0.68, "learning_rate": 3.8710220031183486e-05, "loss": 0.0, "step": 34912 }, { "epoch": 0.68, "learning_rate": 3.8709573070926255e-05, "loss": 0.0057, "step": 34914 }, { "epoch": 0.68, "learning_rate": 3.8708926110669024e-05, "loss": 0.0002, "step": 34916 }, { "epoch": 0.68, "learning_rate": 3.8708279150411794e-05, "loss": 0.0007, "step": 34918 }, { "epoch": 0.68, "learning_rate": 3.8707632190154556e-05, "loss": 0.0, "step": 34920 }, { "epoch": 0.68, "learning_rate": 3.870698522989733e-05, "loss": 0.0001, "step": 34922 }, { "epoch": 0.68, "learning_rate": 3.8706338269640094e-05, "loss": 0.0001, "step": 34924 }, { "epoch": 0.68, "learning_rate": 3.870569130938287e-05, "loss": 0.0003, "step": 34926 }, { "epoch": 0.68, "learning_rate": 3.870504434912563e-05, "loss": 0.0049, "step": 34928 }, { "epoch": 0.68, "learning_rate": 3.87043973888684e-05, "loss": 0.0532, "step": 34930 }, { "epoch": 0.68, "learning_rate": 3.870375042861118e-05, "loss": 0.0, "step": 34932 }, { "epoch": 0.68, "learning_rate": 3.870310346835394e-05, "loss": 0.0, "step": 34934 }, { "epoch": 0.68, "learning_rate": 3.870245650809671e-05, "loss": 0.0001, "step": 34936 }, { "epoch": 0.68, "learning_rate": 3.870180954783948e-05, "loss": 0.0036, "step": 34938 }, { "epoch": 0.68, "learning_rate": 3.870116258758225e-05, "loss": 0.0005, "step": 34940 }, { "epoch": 0.68, "learning_rate": 3.8700515627325016e-05, "loss": 0.0175, "step": 34942 }, { "epoch": 0.68, "learning_rate": 3.8699868667067785e-05, "loss": 0.0005, "step": 34944 }, { "epoch": 0.68, "learning_rate": 3.8699221706810554e-05, "loss": 0.0, "step": 34946 }, { "epoch": 0.68, "learning_rate": 3.869857474655332e-05, "loss": 0.0002, "step": 34948 }, { "epoch": 0.68, "learning_rate": 3.869792778629609e-05, "loss": 0.0003, "step": 34950 }, { "epoch": 0.68, "learning_rate": 3.8697280826038854e-05, "loss": 0.0002, "step": 34952 }, { "epoch": 0.68, "learning_rate": 3.869663386578163e-05, "loss": 0.0002, "step": 34954 }, { "epoch": 0.68, "learning_rate": 3.869598690552439e-05, "loss": 0.0001, "step": 34956 }, { "epoch": 0.68, "learning_rate": 3.869533994526716e-05, "loss": 0.0, "step": 34958 }, { "epoch": 0.68, "learning_rate": 3.869469298500993e-05, "loss": 0.0003, "step": 34960 }, { "epoch": 0.68, "learning_rate": 3.86940460247527e-05, "loss": 0.0, "step": 34962 }, { "epoch": 0.68, "learning_rate": 3.8693399064495476e-05, "loss": 0.0, "step": 34964 }, { "epoch": 0.68, "learning_rate": 3.869275210423824e-05, "loss": 0.0, "step": 34966 }, { "epoch": 0.68, "learning_rate": 3.869210514398101e-05, "loss": 0.0002, "step": 34968 }, { "epoch": 0.68, "learning_rate": 3.8691458183723776e-05, "loss": 0.0001, "step": 34970 }, { "epoch": 0.68, "learning_rate": 3.8690811223466545e-05, "loss": 0.0015, "step": 34972 }, { "epoch": 0.68, "learning_rate": 3.869016426320931e-05, "loss": 0.0, "step": 34974 }, { "epoch": 0.68, "learning_rate": 3.8689517302952083e-05, "loss": 0.0001, "step": 34976 }, { "epoch": 0.68, "learning_rate": 3.868887034269485e-05, "loss": 0.0007, "step": 34978 }, { "epoch": 0.68, "learning_rate": 3.8688223382437615e-05, "loss": 0.0, "step": 34980 }, { "epoch": 0.68, "learning_rate": 3.868757642218039e-05, "loss": 0.0001, "step": 34982 }, { "epoch": 0.68, "learning_rate": 3.868692946192315e-05, "loss": 0.0, "step": 34984 }, { "epoch": 0.68, "learning_rate": 3.868628250166593e-05, "loss": 0.0, "step": 34986 }, { "epoch": 0.68, "learning_rate": 3.868563554140869e-05, "loss": 0.0004, "step": 34988 }, { "epoch": 0.68, "learning_rate": 3.868498858115146e-05, "loss": 0.0, "step": 34990 }, { "epoch": 0.68, "learning_rate": 3.868434162089423e-05, "loss": 0.0001, "step": 34992 }, { "epoch": 0.68, "learning_rate": 3.8683694660637e-05, "loss": 0.0001, "step": 34994 }, { "epoch": 0.68, "learning_rate": 3.868304770037977e-05, "loss": 0.0, "step": 34996 }, { "epoch": 0.68, "learning_rate": 3.8682400740122537e-05, "loss": 0.0001, "step": 34998 }, { "epoch": 0.68, "learning_rate": 3.8681753779865306e-05, "loss": 0.0247, "step": 35000 }, { "epoch": 0.68, "learning_rate": 3.868110681960807e-05, "loss": 0.0, "step": 35002 }, { "epoch": 0.68, "learning_rate": 3.8680459859350844e-05, "loss": 0.0, "step": 35004 }, { "epoch": 0.68, "learning_rate": 3.8679812899093606e-05, "loss": 0.0, "step": 35006 }, { "epoch": 0.68, "learning_rate": 3.867916593883638e-05, "loss": 0.0007, "step": 35008 }, { "epoch": 0.68, "learning_rate": 3.867851897857915e-05, "loss": 0.0, "step": 35010 }, { "epoch": 0.68, "learning_rate": 3.8677872018321913e-05, "loss": 0.0002, "step": 35012 }, { "epoch": 0.68, "learning_rate": 3.867722505806469e-05, "loss": 0.0006, "step": 35014 }, { "epoch": 0.68, "learning_rate": 3.867657809780745e-05, "loss": 0.0046, "step": 35016 }, { "epoch": 0.68, "learning_rate": 3.867593113755022e-05, "loss": 0.0049, "step": 35018 }, { "epoch": 0.68, "learning_rate": 3.867528417729299e-05, "loss": 0.0039, "step": 35020 }, { "epoch": 0.68, "learning_rate": 3.867463721703576e-05, "loss": 0.0, "step": 35022 }, { "epoch": 0.68, "learning_rate": 3.867399025677853e-05, "loss": 0.0001, "step": 35024 }, { "epoch": 0.68, "learning_rate": 3.86733432965213e-05, "loss": 0.0, "step": 35026 }, { "epoch": 0.68, "learning_rate": 3.8672696336264066e-05, "loss": 0.0048, "step": 35028 }, { "epoch": 0.68, "learning_rate": 3.8672049376006835e-05, "loss": 0.0001, "step": 35030 }, { "epoch": 0.68, "learning_rate": 3.8671402415749604e-05, "loss": 0.0002, "step": 35032 }, { "epoch": 0.68, "learning_rate": 3.8670755455492367e-05, "loss": 0.0, "step": 35034 }, { "epoch": 0.68, "learning_rate": 3.867010849523514e-05, "loss": 0.0, "step": 35036 }, { "epoch": 0.68, "learning_rate": 3.8669461534977905e-05, "loss": 0.0, "step": 35038 }, { "epoch": 0.68, "learning_rate": 3.8668814574720674e-05, "loss": 0.0001, "step": 35040 }, { "epoch": 0.68, "learning_rate": 3.866816761446345e-05, "loss": 0.1017, "step": 35042 }, { "epoch": 0.68, "learning_rate": 3.866752065420621e-05, "loss": 0.0, "step": 35044 }, { "epoch": 0.68, "learning_rate": 3.866687369394899e-05, "loss": 0.0002, "step": 35046 }, { "epoch": 0.68, "learning_rate": 3.866622673369175e-05, "loss": 0.0001, "step": 35048 }, { "epoch": 0.68, "learning_rate": 3.866557977343452e-05, "loss": 0.0116, "step": 35050 }, { "epoch": 0.68, "learning_rate": 3.866493281317729e-05, "loss": 0.0024, "step": 35052 }, { "epoch": 0.68, "learning_rate": 3.866428585292006e-05, "loss": 0.0, "step": 35054 }, { "epoch": 0.68, "learning_rate": 3.8663638892662826e-05, "loss": 0.0001, "step": 35056 }, { "epoch": 0.68, "learning_rate": 3.8662991932405596e-05, "loss": 0.0, "step": 35058 }, { "epoch": 0.68, "learning_rate": 3.8662344972148365e-05, "loss": 0.008, "step": 35060 }, { "epoch": 0.68, "learning_rate": 3.866169801189113e-05, "loss": 0.0, "step": 35062 }, { "epoch": 0.68, "learning_rate": 3.86610510516339e-05, "loss": 0.0015, "step": 35064 }, { "epoch": 0.68, "learning_rate": 3.8660404091376665e-05, "loss": 0.0033, "step": 35066 }, { "epoch": 0.68, "learning_rate": 3.865975713111944e-05, "loss": 0.0, "step": 35068 }, { "epoch": 0.68, "learning_rate": 3.86591101708622e-05, "loss": 0.0001, "step": 35070 }, { "epoch": 0.68, "learning_rate": 3.865846321060497e-05, "loss": 0.0, "step": 35072 }, { "epoch": 0.68, "learning_rate": 3.865781625034774e-05, "loss": 0.0001, "step": 35074 }, { "epoch": 0.68, "learning_rate": 3.865716929009051e-05, "loss": 0.0001, "step": 35076 }, { "epoch": 0.68, "learning_rate": 3.865652232983328e-05, "loss": 0.0001, "step": 35078 }, { "epoch": 0.68, "learning_rate": 3.865587536957605e-05, "loss": 0.0, "step": 35080 }, { "epoch": 0.68, "learning_rate": 3.865522840931882e-05, "loss": 0.0, "step": 35082 }, { "epoch": 0.68, "learning_rate": 3.865458144906159e-05, "loss": 0.0, "step": 35084 }, { "epoch": 0.68, "learning_rate": 3.8653934488804356e-05, "loss": 0.0019, "step": 35086 }, { "epoch": 0.68, "learning_rate": 3.8653287528547125e-05, "loss": 0.0001, "step": 35088 }, { "epoch": 0.68, "learning_rate": 3.8652640568289894e-05, "loss": 0.1185, "step": 35090 }, { "epoch": 0.68, "learning_rate": 3.865199360803266e-05, "loss": 0.0001, "step": 35092 }, { "epoch": 0.68, "learning_rate": 3.8651346647775426e-05, "loss": 0.0, "step": 35094 }, { "epoch": 0.68, "learning_rate": 3.86506996875182e-05, "loss": 0.0001, "step": 35096 }, { "epoch": 0.68, "learning_rate": 3.8650052727260964e-05, "loss": 0.0, "step": 35098 }, { "epoch": 0.68, "learning_rate": 3.864940576700373e-05, "loss": 0.0061, "step": 35100 }, { "epoch": 0.68, "learning_rate": 3.86487588067465e-05, "loss": 0.0001, "step": 35102 }, { "epoch": 0.68, "learning_rate": 3.864811184648927e-05, "loss": 0.0084, "step": 35104 }, { "epoch": 0.68, "learning_rate": 3.864746488623204e-05, "loss": 0.0048, "step": 35106 }, { "epoch": 0.68, "learning_rate": 3.864681792597481e-05, "loss": 0.0001, "step": 35108 }, { "epoch": 0.68, "learning_rate": 3.864617096571758e-05, "loss": 0.0, "step": 35110 }, { "epoch": 0.68, "learning_rate": 3.864552400546035e-05, "loss": 0.0001, "step": 35112 }, { "epoch": 0.68, "learning_rate": 3.8644877045203116e-05, "loss": 0.0001, "step": 35114 }, { "epoch": 0.68, "learning_rate": 3.864423008494588e-05, "loss": 0.0091, "step": 35116 }, { "epoch": 0.68, "learning_rate": 3.8643583124688655e-05, "loss": 0.0005, "step": 35118 }, { "epoch": 0.68, "learning_rate": 3.8642936164431424e-05, "loss": 0.0, "step": 35120 }, { "epoch": 0.68, "learning_rate": 3.8642289204174186e-05, "loss": 0.0, "step": 35122 }, { "epoch": 0.68, "learning_rate": 3.864164224391696e-05, "loss": 0.0, "step": 35124 }, { "epoch": 0.68, "learning_rate": 3.8640995283659724e-05, "loss": 0.0, "step": 35126 }, { "epoch": 0.68, "learning_rate": 3.86403483234025e-05, "loss": 0.0, "step": 35128 }, { "epoch": 0.68, "learning_rate": 3.863970136314526e-05, "loss": 0.0, "step": 35130 }, { "epoch": 0.68, "learning_rate": 3.863905440288803e-05, "loss": 0.043, "step": 35132 }, { "epoch": 0.68, "learning_rate": 3.86384074426308e-05, "loss": 0.0037, "step": 35134 }, { "epoch": 0.68, "learning_rate": 3.863776048237357e-05, "loss": 0.0001, "step": 35136 }, { "epoch": 0.68, "learning_rate": 3.863711352211634e-05, "loss": 0.0, "step": 35138 }, { "epoch": 0.68, "learning_rate": 3.863646656185911e-05, "loss": 0.0146, "step": 35140 }, { "epoch": 0.68, "learning_rate": 3.863581960160188e-05, "loss": 0.0, "step": 35142 }, { "epoch": 0.68, "learning_rate": 3.863517264134464e-05, "loss": 0.0001, "step": 35144 }, { "epoch": 0.68, "learning_rate": 3.8634525681087415e-05, "loss": 0.0133, "step": 35146 }, { "epoch": 0.68, "learning_rate": 3.863387872083018e-05, "loss": 0.0001, "step": 35148 }, { "epoch": 0.68, "learning_rate": 3.863323176057295e-05, "loss": 0.0, "step": 35150 }, { "epoch": 0.68, "learning_rate": 3.8632584800315715e-05, "loss": 0.0, "step": 35152 }, { "epoch": 0.68, "learning_rate": 3.8631937840058485e-05, "loss": 0.0004, "step": 35154 }, { "epoch": 0.68, "learning_rate": 3.863129087980126e-05, "loss": 0.0019, "step": 35156 }, { "epoch": 0.68, "learning_rate": 3.863064391954402e-05, "loss": 0.0001, "step": 35158 }, { "epoch": 0.68, "learning_rate": 3.862999695928679e-05, "loss": 0.0005, "step": 35160 }, { "epoch": 0.68, "learning_rate": 3.862934999902956e-05, "loss": 0.0049, "step": 35162 }, { "epoch": 0.68, "learning_rate": 3.862870303877233e-05, "loss": 0.0039, "step": 35164 }, { "epoch": 0.68, "learning_rate": 3.86280560785151e-05, "loss": 0.0016, "step": 35166 }, { "epoch": 0.68, "learning_rate": 3.862740911825787e-05, "loss": 0.0, "step": 35168 }, { "epoch": 0.68, "learning_rate": 3.862676215800064e-05, "loss": 0.0, "step": 35170 }, { "epoch": 0.68, "learning_rate": 3.8626115197743406e-05, "loss": 0.0, "step": 35172 }, { "epoch": 0.68, "learning_rate": 3.8625468237486175e-05, "loss": 0.0016, "step": 35174 }, { "epoch": 0.68, "learning_rate": 3.862482127722894e-05, "loss": 0.0, "step": 35176 }, { "epoch": 0.68, "learning_rate": 3.8624174316971714e-05, "loss": 0.0027, "step": 35178 }, { "epoch": 0.68, "learning_rate": 3.8623527356714476e-05, "loss": 0.0131, "step": 35180 }, { "epoch": 0.68, "learning_rate": 3.8622880396457245e-05, "loss": 0.0002, "step": 35182 }, { "epoch": 0.68, "learning_rate": 3.8622233436200014e-05, "loss": 0.0001, "step": 35184 }, { "epoch": 0.68, "learning_rate": 3.862158647594278e-05, "loss": 0.0148, "step": 35186 }, { "epoch": 0.68, "learning_rate": 3.862093951568556e-05, "loss": 0.0001, "step": 35188 }, { "epoch": 0.68, "learning_rate": 3.862029255542832e-05, "loss": 0.0, "step": 35190 }, { "epoch": 0.68, "learning_rate": 3.861964559517109e-05, "loss": 0.0001, "step": 35192 }, { "epoch": 0.68, "learning_rate": 3.861899863491386e-05, "loss": 0.0028, "step": 35194 }, { "epoch": 0.68, "learning_rate": 3.861835167465663e-05, "loss": 0.0073, "step": 35196 }, { "epoch": 0.68, "learning_rate": 3.861770471439939e-05, "loss": 0.0135, "step": 35198 }, { "epoch": 0.68, "learning_rate": 3.861705775414217e-05, "loss": 0.0, "step": 35200 }, { "epoch": 0.68, "learning_rate": 3.8616410793884936e-05, "loss": 0.0366, "step": 35202 }, { "epoch": 0.68, "learning_rate": 3.86157638336277e-05, "loss": 0.014, "step": 35204 }, { "epoch": 0.68, "learning_rate": 3.8615116873370474e-05, "loss": 0.0, "step": 35206 }, { "epoch": 0.68, "learning_rate": 3.8614469913113236e-05, "loss": 0.0045, "step": 35208 }, { "epoch": 0.68, "learning_rate": 3.861382295285601e-05, "loss": 0.0033, "step": 35210 }, { "epoch": 0.68, "learning_rate": 3.8613175992598774e-05, "loss": 0.0046, "step": 35212 }, { "epoch": 0.68, "learning_rate": 3.8612529032341543e-05, "loss": 0.0, "step": 35214 }, { "epoch": 0.68, "learning_rate": 3.861188207208431e-05, "loss": 0.002, "step": 35216 }, { "epoch": 0.68, "learning_rate": 3.861123511182708e-05, "loss": 0.0004, "step": 35218 }, { "epoch": 0.68, "learning_rate": 3.861058815156985e-05, "loss": 0.0046, "step": 35220 }, { "epoch": 0.68, "learning_rate": 3.860994119131262e-05, "loss": 0.0009, "step": 35222 }, { "epoch": 0.68, "learning_rate": 3.860929423105539e-05, "loss": 0.0, "step": 35224 }, { "epoch": 0.68, "learning_rate": 3.860864727079816e-05, "loss": 0.0038, "step": 35226 }, { "epoch": 0.68, "learning_rate": 3.860800031054093e-05, "loss": 0.0001, "step": 35228 }, { "epoch": 0.68, "learning_rate": 3.860735335028369e-05, "loss": 0.0, "step": 35230 }, { "epoch": 0.68, "learning_rate": 3.8606706390026465e-05, "loss": 0.007, "step": 35232 }, { "epoch": 0.68, "learning_rate": 3.8606059429769234e-05, "loss": 0.0127, "step": 35234 }, { "epoch": 0.68, "learning_rate": 3.8605412469512e-05, "loss": 0.0001, "step": 35236 }, { "epoch": 0.68, "learning_rate": 3.860476550925477e-05, "loss": 0.0105, "step": 35238 }, { "epoch": 0.68, "learning_rate": 3.8604118548997535e-05, "loss": 0.0, "step": 35240 }, { "epoch": 0.68, "learning_rate": 3.8603471588740304e-05, "loss": 0.0, "step": 35242 }, { "epoch": 0.68, "learning_rate": 3.860282462848307e-05, "loss": 0.0, "step": 35244 }, { "epoch": 0.68, "learning_rate": 3.860217766822584e-05, "loss": 0.0013, "step": 35246 }, { "epoch": 0.68, "learning_rate": 3.860153070796861e-05, "loss": 0.0, "step": 35248 }, { "epoch": 0.68, "learning_rate": 3.860088374771138e-05, "loss": 0.0, "step": 35250 }, { "epoch": 0.68, "learning_rate": 3.860023678745415e-05, "loss": 0.0, "step": 35252 }, { "epoch": 0.68, "learning_rate": 3.859958982719692e-05, "loss": 0.0002, "step": 35254 }, { "epoch": 0.68, "learning_rate": 3.859894286693969e-05, "loss": 0.0008, "step": 35256 }, { "epoch": 0.68, "learning_rate": 3.859829590668245e-05, "loss": 0.0001, "step": 35258 }, { "epoch": 0.68, "learning_rate": 3.8597648946425226e-05, "loss": 0.0079, "step": 35260 }, { "epoch": 0.68, "learning_rate": 3.859700198616799e-05, "loss": 0.0011, "step": 35262 }, { "epoch": 0.68, "learning_rate": 3.859635502591076e-05, "loss": 0.0, "step": 35264 }, { "epoch": 0.68, "learning_rate": 3.859570806565353e-05, "loss": 0.0, "step": 35266 }, { "epoch": 0.68, "learning_rate": 3.8595061105396295e-05, "loss": 0.0001, "step": 35268 }, { "epoch": 0.68, "learning_rate": 3.859441414513907e-05, "loss": 0.0001, "step": 35270 }, { "epoch": 0.68, "learning_rate": 3.859376718488183e-05, "loss": 0.0087, "step": 35272 }, { "epoch": 0.68, "learning_rate": 3.85931202246246e-05, "loss": 0.0014, "step": 35274 }, { "epoch": 0.68, "learning_rate": 3.859247326436737e-05, "loss": 0.0, "step": 35276 }, { "epoch": 0.68, "learning_rate": 3.859182630411014e-05, "loss": 0.0038, "step": 35278 }, { "epoch": 0.68, "learning_rate": 3.859117934385291e-05, "loss": 0.0, "step": 35280 }, { "epoch": 0.68, "learning_rate": 3.859053238359568e-05, "loss": 0.0082, "step": 35282 }, { "epoch": 0.68, "learning_rate": 3.858988542333845e-05, "loss": 0.0005, "step": 35284 }, { "epoch": 0.68, "learning_rate": 3.858923846308121e-05, "loss": 0.0002, "step": 35286 }, { "epoch": 0.68, "learning_rate": 3.8588591502823986e-05, "loss": 0.0, "step": 35288 }, { "epoch": 0.68, "learning_rate": 3.858794454256675e-05, "loss": 0.0029, "step": 35290 }, { "epoch": 0.68, "learning_rate": 3.8587297582309524e-05, "loss": 0.0103, "step": 35292 }, { "epoch": 0.69, "learning_rate": 3.8586650622052287e-05, "loss": 0.0, "step": 35294 }, { "epoch": 0.69, "learning_rate": 3.8586003661795056e-05, "loss": 0.0075, "step": 35296 }, { "epoch": 0.69, "learning_rate": 3.858535670153783e-05, "loss": 0.0, "step": 35298 }, { "epoch": 0.69, "learning_rate": 3.8584709741280594e-05, "loss": 0.0, "step": 35300 }, { "epoch": 0.69, "learning_rate": 3.858406278102336e-05, "loss": 0.0035, "step": 35302 }, { "epoch": 0.69, "learning_rate": 3.858341582076613e-05, "loss": 0.0, "step": 35304 }, { "epoch": 0.69, "learning_rate": 3.85827688605089e-05, "loss": 0.0002, "step": 35306 }, { "epoch": 0.69, "learning_rate": 3.858212190025167e-05, "loss": 0.0098, "step": 35308 }, { "epoch": 0.69, "learning_rate": 3.858147493999444e-05, "loss": 0.0, "step": 35310 }, { "epoch": 0.69, "learning_rate": 3.858082797973721e-05, "loss": 0.0, "step": 35312 }, { "epoch": 0.69, "learning_rate": 3.858018101947998e-05, "loss": 0.0008, "step": 35314 }, { "epoch": 0.69, "learning_rate": 3.8579534059222746e-05, "loss": 0.0053, "step": 35316 }, { "epoch": 0.69, "learning_rate": 3.857888709896551e-05, "loss": 0.0003, "step": 35318 }, { "epoch": 0.69, "learning_rate": 3.8578240138708285e-05, "loss": 0.0, "step": 35320 }, { "epoch": 0.69, "learning_rate": 3.857759317845105e-05, "loss": 0.0001, "step": 35322 }, { "epoch": 0.69, "learning_rate": 3.8576946218193816e-05, "loss": 0.0001, "step": 35324 }, { "epoch": 0.69, "learning_rate": 3.8576299257936585e-05, "loss": 0.0042, "step": 35326 }, { "epoch": 0.69, "learning_rate": 3.8575652297679354e-05, "loss": 0.0001, "step": 35328 }, { "epoch": 0.69, "learning_rate": 3.857500533742212e-05, "loss": 0.0023, "step": 35330 }, { "epoch": 0.69, "learning_rate": 3.857435837716489e-05, "loss": 0.0117, "step": 35332 }, { "epoch": 0.69, "learning_rate": 3.857371141690766e-05, "loss": 0.0062, "step": 35334 }, { "epoch": 0.69, "learning_rate": 3.857306445665043e-05, "loss": 0.0004, "step": 35336 }, { "epoch": 0.69, "learning_rate": 3.85724174963932e-05, "loss": 0.0, "step": 35338 }, { "epoch": 0.69, "learning_rate": 3.857177053613596e-05, "loss": 0.0, "step": 35340 }, { "epoch": 0.69, "learning_rate": 3.857112357587874e-05, "loss": 0.0021, "step": 35342 }, { "epoch": 0.69, "learning_rate": 3.857047661562151e-05, "loss": 0.0059, "step": 35344 }, { "epoch": 0.69, "learning_rate": 3.856982965536427e-05, "loss": 0.0002, "step": 35346 }, { "epoch": 0.69, "learning_rate": 3.8569182695107045e-05, "loss": 0.0001, "step": 35348 }, { "epoch": 0.69, "learning_rate": 3.856853573484981e-05, "loss": 0.0, "step": 35350 }, { "epoch": 0.69, "learning_rate": 3.856788877459258e-05, "loss": 0.0, "step": 35352 }, { "epoch": 0.69, "learning_rate": 3.8567241814335346e-05, "loss": 0.0092, "step": 35354 }, { "epoch": 0.69, "learning_rate": 3.8566594854078115e-05, "loss": 0.0003, "step": 35356 }, { "epoch": 0.69, "learning_rate": 3.8565947893820884e-05, "loss": 0.0092, "step": 35358 }, { "epoch": 0.69, "learning_rate": 3.856530093356365e-05, "loss": 0.0061, "step": 35360 }, { "epoch": 0.69, "learning_rate": 3.856465397330642e-05, "loss": 0.0003, "step": 35362 }, { "epoch": 0.69, "learning_rate": 3.856400701304919e-05, "loss": 0.0004, "step": 35364 }, { "epoch": 0.69, "learning_rate": 3.856336005279196e-05, "loss": 0.0019, "step": 35366 }, { "epoch": 0.69, "learning_rate": 3.856271309253473e-05, "loss": 0.0013, "step": 35368 }, { "epoch": 0.69, "learning_rate": 3.85620661322775e-05, "loss": 0.0004, "step": 35370 }, { "epoch": 0.69, "learning_rate": 3.856141917202026e-05, "loss": 0.0002, "step": 35372 }, { "epoch": 0.69, "learning_rate": 3.8560772211763036e-05, "loss": 0.0001, "step": 35374 }, { "epoch": 0.69, "learning_rate": 3.85601252515058e-05, "loss": 0.0081, "step": 35376 }, { "epoch": 0.69, "learning_rate": 3.855947829124857e-05, "loss": 0.0, "step": 35378 }, { "epoch": 0.69, "learning_rate": 3.8558831330991344e-05, "loss": 0.0001, "step": 35380 }, { "epoch": 0.69, "learning_rate": 3.8558184370734106e-05, "loss": 0.0007, "step": 35382 }, { "epoch": 0.69, "learning_rate": 3.8557537410476875e-05, "loss": 0.0128, "step": 35384 }, { "epoch": 0.69, "learning_rate": 3.8556890450219644e-05, "loss": 0.0, "step": 35386 }, { "epoch": 0.69, "learning_rate": 3.855624348996241e-05, "loss": 0.0003, "step": 35388 }, { "epoch": 0.69, "learning_rate": 3.855559652970518e-05, "loss": 0.001, "step": 35390 }, { "epoch": 0.69, "learning_rate": 3.855494956944795e-05, "loss": 0.0, "step": 35392 }, { "epoch": 0.69, "learning_rate": 3.855430260919072e-05, "loss": 0.0016, "step": 35394 }, { "epoch": 0.69, "learning_rate": 3.855365564893349e-05, "loss": 0.0005, "step": 35396 }, { "epoch": 0.69, "learning_rate": 3.855300868867626e-05, "loss": 0.0001, "step": 35398 }, { "epoch": 0.69, "learning_rate": 3.855236172841902e-05, "loss": 0.0003, "step": 35400 }, { "epoch": 0.69, "learning_rate": 3.85517147681618e-05, "loss": 0.0001, "step": 35402 }, { "epoch": 0.69, "learning_rate": 3.855106780790456e-05, "loss": 0.0002, "step": 35404 }, { "epoch": 0.69, "learning_rate": 3.855042084764733e-05, "loss": 0.0001, "step": 35406 }, { "epoch": 0.69, "learning_rate": 3.85497738873901e-05, "loss": 0.0001, "step": 35408 }, { "epoch": 0.69, "learning_rate": 3.8549126927132866e-05, "loss": 0.0, "step": 35410 }, { "epoch": 0.69, "learning_rate": 3.854847996687564e-05, "loss": 0.0, "step": 35412 }, { "epoch": 0.69, "learning_rate": 3.8547833006618404e-05, "loss": 0.0077, "step": 35414 }, { "epoch": 0.69, "learning_rate": 3.8547186046361174e-05, "loss": 0.0, "step": 35416 }, { "epoch": 0.69, "learning_rate": 3.854653908610394e-05, "loss": 0.0015, "step": 35418 }, { "epoch": 0.69, "learning_rate": 3.854589212584671e-05, "loss": 0.0002, "step": 35420 }, { "epoch": 0.69, "learning_rate": 3.854524516558948e-05, "loss": 0.0001, "step": 35422 }, { "epoch": 0.69, "learning_rate": 3.854459820533225e-05, "loss": 0.017, "step": 35424 }, { "epoch": 0.69, "learning_rate": 3.854395124507502e-05, "loss": 0.0, "step": 35426 }, { "epoch": 0.69, "learning_rate": 3.854330428481778e-05, "loss": 0.0, "step": 35428 }, { "epoch": 0.69, "learning_rate": 3.854265732456056e-05, "loss": 0.0, "step": 35430 }, { "epoch": 0.69, "learning_rate": 3.854201036430332e-05, "loss": 0.0, "step": 35432 }, { "epoch": 0.69, "learning_rate": 3.8541363404046095e-05, "loss": 0.0, "step": 35434 }, { "epoch": 0.69, "learning_rate": 3.854071644378886e-05, "loss": 0.0, "step": 35436 }, { "epoch": 0.69, "learning_rate": 3.854006948353163e-05, "loss": 0.0114, "step": 35438 }, { "epoch": 0.69, "learning_rate": 3.8539422523274396e-05, "loss": 0.0013, "step": 35440 }, { "epoch": 0.69, "learning_rate": 3.8538775563017165e-05, "loss": 0.0, "step": 35442 }, { "epoch": 0.69, "learning_rate": 3.8538128602759934e-05, "loss": 0.0184, "step": 35444 }, { "epoch": 0.69, "learning_rate": 3.85374816425027e-05, "loss": 0.0, "step": 35446 }, { "epoch": 0.69, "learning_rate": 3.853683468224547e-05, "loss": 0.0, "step": 35448 }, { "epoch": 0.69, "learning_rate": 3.853618772198824e-05, "loss": 0.0, "step": 35450 }, { "epoch": 0.69, "learning_rate": 3.853554076173101e-05, "loss": 0.0032, "step": 35452 }, { "epoch": 0.69, "learning_rate": 3.853489380147377e-05, "loss": 0.0, "step": 35454 }, { "epoch": 0.69, "learning_rate": 3.853424684121655e-05, "loss": 0.0001, "step": 35456 }, { "epoch": 0.69, "learning_rate": 3.853359988095932e-05, "loss": 0.0002, "step": 35458 }, { "epoch": 0.69, "learning_rate": 3.853295292070208e-05, "loss": 0.0001, "step": 35460 }, { "epoch": 0.69, "learning_rate": 3.8532305960444856e-05, "loss": 0.0015, "step": 35462 }, { "epoch": 0.69, "learning_rate": 3.853165900018762e-05, "loss": 0.0, "step": 35464 }, { "epoch": 0.69, "learning_rate": 3.853101203993039e-05, "loss": 0.0, "step": 35466 }, { "epoch": 0.69, "learning_rate": 3.8530365079673156e-05, "loss": 0.0002, "step": 35468 }, { "epoch": 0.69, "learning_rate": 3.8529718119415925e-05, "loss": 0.0001, "step": 35470 }, { "epoch": 0.69, "learning_rate": 3.8529071159158694e-05, "loss": 0.0, "step": 35472 }, { "epoch": 0.69, "learning_rate": 3.8528424198901463e-05, "loss": 0.0, "step": 35474 }, { "epoch": 0.69, "learning_rate": 3.852777723864423e-05, "loss": 0.0005, "step": 35476 }, { "epoch": 0.69, "learning_rate": 3.8527130278387e-05, "loss": 0.0, "step": 35478 }, { "epoch": 0.69, "learning_rate": 3.852648331812977e-05, "loss": 0.0, "step": 35480 }, { "epoch": 0.69, "learning_rate": 3.852583635787253e-05, "loss": 0.0014, "step": 35482 }, { "epoch": 0.69, "learning_rate": 3.852518939761531e-05, "loss": 0.0, "step": 35484 }, { "epoch": 0.69, "learning_rate": 3.852454243735807e-05, "loss": 0.0002, "step": 35486 }, { "epoch": 0.69, "learning_rate": 3.852389547710084e-05, "loss": 0.0, "step": 35488 }, { "epoch": 0.69, "learning_rate": 3.8523248516843616e-05, "loss": 0.0011, "step": 35490 }, { "epoch": 0.69, "learning_rate": 3.852260155658638e-05, "loss": 0.0, "step": 35492 }, { "epoch": 0.69, "learning_rate": 3.8521954596329154e-05, "loss": 0.0016, "step": 35494 }, { "epoch": 0.69, "learning_rate": 3.852130763607192e-05, "loss": 0.0034, "step": 35496 }, { "epoch": 0.69, "learning_rate": 3.8520660675814686e-05, "loss": 0.0, "step": 35498 }, { "epoch": 0.69, "learning_rate": 3.8520013715557455e-05, "loss": 0.0, "step": 35500 }, { "epoch": 0.69, "learning_rate": 3.8519366755300224e-05, "loss": 0.0004, "step": 35502 }, { "epoch": 0.69, "learning_rate": 3.851871979504299e-05, "loss": 0.0001, "step": 35504 }, { "epoch": 0.69, "learning_rate": 3.851807283478576e-05, "loss": 0.0022, "step": 35506 }, { "epoch": 0.69, "learning_rate": 3.851742587452853e-05, "loss": 0.0, "step": 35508 }, { "epoch": 0.69, "learning_rate": 3.85167789142713e-05, "loss": 0.0, "step": 35510 }, { "epoch": 0.69, "learning_rate": 3.851613195401407e-05, "loss": 0.0009, "step": 35512 }, { "epoch": 0.69, "learning_rate": 3.851548499375683e-05, "loss": 0.0, "step": 35514 }, { "epoch": 0.69, "learning_rate": 3.851483803349961e-05, "loss": 0.0054, "step": 35516 }, { "epoch": 0.69, "learning_rate": 3.851419107324237e-05, "loss": 0.0021, "step": 35518 }, { "epoch": 0.69, "learning_rate": 3.851354411298514e-05, "loss": 0.0, "step": 35520 }, { "epoch": 0.69, "learning_rate": 3.8512897152727915e-05, "loss": 0.0033, "step": 35522 }, { "epoch": 0.69, "learning_rate": 3.851225019247068e-05, "loss": 0.0008, "step": 35524 }, { "epoch": 0.69, "learning_rate": 3.8511603232213446e-05, "loss": 0.0, "step": 35526 }, { "epoch": 0.69, "learning_rate": 3.8510956271956215e-05, "loss": 0.1554, "step": 35528 }, { "epoch": 0.69, "learning_rate": 3.8510309311698984e-05, "loss": 0.0001, "step": 35530 }, { "epoch": 0.69, "learning_rate": 3.850966235144175e-05, "loss": 0.0031, "step": 35532 }, { "epoch": 0.69, "learning_rate": 3.850901539118452e-05, "loss": 0.0004, "step": 35534 }, { "epoch": 0.69, "learning_rate": 3.850836843092729e-05, "loss": 0.0, "step": 35536 }, { "epoch": 0.69, "learning_rate": 3.850772147067006e-05, "loss": 0.0002, "step": 35538 }, { "epoch": 0.69, "learning_rate": 3.850707451041283e-05, "loss": 0.0005, "step": 35540 }, { "epoch": 0.69, "learning_rate": 3.850642755015559e-05, "loss": 0.0, "step": 35542 }, { "epoch": 0.69, "learning_rate": 3.850578058989837e-05, "loss": 0.0, "step": 35544 }, { "epoch": 0.69, "learning_rate": 3.850513362964113e-05, "loss": 0.0, "step": 35546 }, { "epoch": 0.69, "learning_rate": 3.85044866693839e-05, "loss": 0.0119, "step": 35548 }, { "epoch": 0.69, "learning_rate": 3.850383970912667e-05, "loss": 0.0003, "step": 35550 }, { "epoch": 0.69, "learning_rate": 3.850319274886944e-05, "loss": 0.0001, "step": 35552 }, { "epoch": 0.69, "learning_rate": 3.8502545788612207e-05, "loss": 0.0002, "step": 35554 }, { "epoch": 0.69, "learning_rate": 3.8501898828354976e-05, "loss": 0.0037, "step": 35556 }, { "epoch": 0.69, "learning_rate": 3.8501251868097745e-05, "loss": 0.0001, "step": 35558 }, { "epoch": 0.69, "learning_rate": 3.8500604907840514e-05, "loss": 0.0013, "step": 35560 }, { "epoch": 0.69, "learning_rate": 3.849995794758328e-05, "loss": 0.0, "step": 35562 }, { "epoch": 0.69, "learning_rate": 3.8499310987326045e-05, "loss": 0.0058, "step": 35564 }, { "epoch": 0.69, "learning_rate": 3.849866402706882e-05, "loss": 0.0065, "step": 35566 }, { "epoch": 0.69, "learning_rate": 3.849801706681159e-05, "loss": 0.0001, "step": 35568 }, { "epoch": 0.69, "learning_rate": 3.849737010655435e-05, "loss": 0.0096, "step": 35570 }, { "epoch": 0.69, "learning_rate": 3.849672314629713e-05, "loss": 0.0027, "step": 35572 }, { "epoch": 0.69, "learning_rate": 3.849607618603989e-05, "loss": 0.0011, "step": 35574 }, { "epoch": 0.69, "learning_rate": 3.8495429225782666e-05, "loss": 0.0001, "step": 35576 }, { "epoch": 0.69, "learning_rate": 3.849478226552543e-05, "loss": 0.0005, "step": 35578 }, { "epoch": 0.69, "learning_rate": 3.84941353052682e-05, "loss": 0.0101, "step": 35580 }, { "epoch": 0.69, "learning_rate": 3.849348834501097e-05, "loss": 0.0, "step": 35582 }, { "epoch": 0.69, "learning_rate": 3.8492841384753736e-05, "loss": 0.0, "step": 35584 }, { "epoch": 0.69, "learning_rate": 3.8492194424496505e-05, "loss": 0.0001, "step": 35586 }, { "epoch": 0.69, "learning_rate": 3.8491547464239274e-05, "loss": 0.0, "step": 35588 }, { "epoch": 0.69, "learning_rate": 3.849090050398204e-05, "loss": 0.027, "step": 35590 }, { "epoch": 0.69, "learning_rate": 3.849025354372481e-05, "loss": 0.0, "step": 35592 }, { "epoch": 0.69, "learning_rate": 3.848960658346758e-05, "loss": 0.0217, "step": 35594 }, { "epoch": 0.69, "learning_rate": 3.8488959623210344e-05, "loss": 0.0, "step": 35596 }, { "epoch": 0.69, "learning_rate": 3.848831266295312e-05, "loss": 0.0005, "step": 35598 }, { "epoch": 0.69, "learning_rate": 3.848766570269589e-05, "loss": 0.0025, "step": 35600 }, { "epoch": 0.69, "learning_rate": 3.848701874243865e-05, "loss": 0.0001, "step": 35602 }, { "epoch": 0.69, "learning_rate": 3.848637178218143e-05, "loss": 0.0, "step": 35604 }, { "epoch": 0.69, "learning_rate": 3.848572482192419e-05, "loss": 0.0058, "step": 35606 }, { "epoch": 0.69, "learning_rate": 3.848507786166696e-05, "loss": 0.0, "step": 35608 }, { "epoch": 0.69, "learning_rate": 3.848443090140973e-05, "loss": 0.0001, "step": 35610 }, { "epoch": 0.69, "learning_rate": 3.8483783941152496e-05, "loss": 0.0, "step": 35612 }, { "epoch": 0.69, "learning_rate": 3.8483136980895265e-05, "loss": 0.0095, "step": 35614 }, { "epoch": 0.69, "learning_rate": 3.8482490020638035e-05, "loss": 0.0001, "step": 35616 }, { "epoch": 0.69, "learning_rate": 3.8481843060380804e-05, "loss": 0.0, "step": 35618 }, { "epoch": 0.69, "learning_rate": 3.848119610012357e-05, "loss": 0.0002, "step": 35620 }, { "epoch": 0.69, "learning_rate": 3.848054913986634e-05, "loss": 0.0074, "step": 35622 }, { "epoch": 0.69, "learning_rate": 3.8479902179609104e-05, "loss": 0.028, "step": 35624 }, { "epoch": 0.69, "learning_rate": 3.847925521935188e-05, "loss": 0.0, "step": 35626 }, { "epoch": 0.69, "learning_rate": 3.847860825909464e-05, "loss": 0.0001, "step": 35628 }, { "epoch": 0.69, "learning_rate": 3.847796129883741e-05, "loss": 0.0, "step": 35630 }, { "epoch": 0.69, "learning_rate": 3.847731433858018e-05, "loss": 0.0007, "step": 35632 }, { "epoch": 0.69, "learning_rate": 3.847666737832295e-05, "loss": 0.0027, "step": 35634 }, { "epoch": 0.69, "learning_rate": 3.8476020418065725e-05, "loss": 0.0001, "step": 35636 }, { "epoch": 0.69, "learning_rate": 3.847537345780849e-05, "loss": 0.0001, "step": 35638 }, { "epoch": 0.69, "learning_rate": 3.847472649755126e-05, "loss": 0.0001, "step": 35640 }, { "epoch": 0.69, "learning_rate": 3.8474079537294026e-05, "loss": 0.0001, "step": 35642 }, { "epoch": 0.69, "learning_rate": 3.8473432577036795e-05, "loss": 0.0054, "step": 35644 }, { "epoch": 0.69, "learning_rate": 3.8472785616779564e-05, "loss": 0.001, "step": 35646 }, { "epoch": 0.69, "learning_rate": 3.847213865652233e-05, "loss": 0.0, "step": 35648 }, { "epoch": 0.69, "learning_rate": 3.84714916962651e-05, "loss": 0.0001, "step": 35650 }, { "epoch": 0.69, "learning_rate": 3.847084473600787e-05, "loss": 0.0002, "step": 35652 }, { "epoch": 0.69, "learning_rate": 3.847019777575064e-05, "loss": 0.0032, "step": 35654 }, { "epoch": 0.69, "learning_rate": 3.84695508154934e-05, "loss": 0.0, "step": 35656 }, { "epoch": 0.69, "learning_rate": 3.846890385523618e-05, "loss": 0.0001, "step": 35658 }, { "epoch": 0.69, "learning_rate": 3.846825689497894e-05, "loss": 0.0182, "step": 35660 }, { "epoch": 0.69, "learning_rate": 3.846760993472171e-05, "loss": 0.0, "step": 35662 }, { "epoch": 0.69, "learning_rate": 3.846696297446448e-05, "loss": 0.0, "step": 35664 }, { "epoch": 0.69, "learning_rate": 3.846631601420725e-05, "loss": 0.0001, "step": 35666 }, { "epoch": 0.69, "learning_rate": 3.846566905395002e-05, "loss": 0.0027, "step": 35668 }, { "epoch": 0.69, "learning_rate": 3.8465022093692786e-05, "loss": 0.0, "step": 35670 }, { "epoch": 0.69, "learning_rate": 3.8464375133435555e-05, "loss": 0.008, "step": 35672 }, { "epoch": 0.69, "learning_rate": 3.8463728173178324e-05, "loss": 0.0, "step": 35674 }, { "epoch": 0.69, "learning_rate": 3.8463081212921094e-05, "loss": 0.0, "step": 35676 }, { "epoch": 0.69, "learning_rate": 3.8462434252663856e-05, "loss": 0.0001, "step": 35678 }, { "epoch": 0.69, "learning_rate": 3.846178729240663e-05, "loss": 0.0027, "step": 35680 }, { "epoch": 0.69, "learning_rate": 3.84611403321494e-05, "loss": 0.0, "step": 35682 }, { "epoch": 0.69, "learning_rate": 3.846049337189216e-05, "loss": 0.0, "step": 35684 }, { "epoch": 0.69, "learning_rate": 3.845984641163494e-05, "loss": 0.0, "step": 35686 }, { "epoch": 0.69, "learning_rate": 3.84591994513777e-05, "loss": 0.0, "step": 35688 }, { "epoch": 0.69, "learning_rate": 3.845855249112047e-05, "loss": 0.0001, "step": 35690 }, { "epoch": 0.69, "learning_rate": 3.845790553086324e-05, "loss": 0.0, "step": 35692 }, { "epoch": 0.69, "learning_rate": 3.845725857060601e-05, "loss": 0.0001, "step": 35694 }, { "epoch": 0.69, "learning_rate": 3.845661161034878e-05, "loss": 0.0, "step": 35696 }, { "epoch": 0.69, "learning_rate": 3.845596465009155e-05, "loss": 0.0, "step": 35698 }, { "epoch": 0.69, "learning_rate": 3.8455317689834316e-05, "loss": 0.0002, "step": 35700 }, { "epoch": 0.69, "learning_rate": 3.8454670729577085e-05, "loss": 0.0063, "step": 35702 }, { "epoch": 0.69, "learning_rate": 3.8454023769319854e-05, "loss": 0.0092, "step": 35704 }, { "epoch": 0.69, "learning_rate": 3.8453376809062616e-05, "loss": 0.0, "step": 35706 }, { "epoch": 0.69, "learning_rate": 3.845272984880539e-05, "loss": 0.0, "step": 35708 }, { "epoch": 0.69, "learning_rate": 3.8452082888548154e-05, "loss": 0.0006, "step": 35710 }, { "epoch": 0.69, "learning_rate": 3.8451435928290924e-05, "loss": 0.0, "step": 35712 }, { "epoch": 0.69, "learning_rate": 3.84507889680337e-05, "loss": 0.005, "step": 35714 }, { "epoch": 0.69, "learning_rate": 3.845014200777646e-05, "loss": 0.0, "step": 35716 }, { "epoch": 0.69, "learning_rate": 3.844949504751924e-05, "loss": 0.0024, "step": 35718 }, { "epoch": 0.69, "learning_rate": 3.8448848087262e-05, "loss": 0.0001, "step": 35720 }, { "epoch": 0.69, "learning_rate": 3.844820112700477e-05, "loss": 0.0001, "step": 35722 }, { "epoch": 0.69, "learning_rate": 3.844755416674754e-05, "loss": 0.0002, "step": 35724 }, { "epoch": 0.69, "learning_rate": 3.844690720649031e-05, "loss": 0.0009, "step": 35726 }, { "epoch": 0.69, "learning_rate": 3.8446260246233076e-05, "loss": 0.0, "step": 35728 }, { "epoch": 0.69, "learning_rate": 3.8445613285975845e-05, "loss": 0.0002, "step": 35730 }, { "epoch": 0.69, "learning_rate": 3.8444966325718614e-05, "loss": 0.0001, "step": 35732 }, { "epoch": 0.69, "learning_rate": 3.8444319365461383e-05, "loss": 0.0, "step": 35734 }, { "epoch": 0.69, "learning_rate": 3.844367240520415e-05, "loss": 0.0003, "step": 35736 }, { "epoch": 0.69, "learning_rate": 3.8443025444946915e-05, "loss": 0.0001, "step": 35738 }, { "epoch": 0.69, "learning_rate": 3.844237848468969e-05, "loss": 0.0, "step": 35740 }, { "epoch": 0.69, "learning_rate": 3.844173152443245e-05, "loss": 0.0, "step": 35742 }, { "epoch": 0.69, "learning_rate": 3.844108456417522e-05, "loss": 0.0, "step": 35744 }, { "epoch": 0.69, "learning_rate": 3.8440437603918e-05, "loss": 0.0, "step": 35746 }, { "epoch": 0.69, "learning_rate": 3.843979064366076e-05, "loss": 0.0072, "step": 35748 }, { "epoch": 0.69, "learning_rate": 3.843914368340353e-05, "loss": 0.0002, "step": 35750 }, { "epoch": 0.69, "learning_rate": 3.84384967231463e-05, "loss": 0.0001, "step": 35752 }, { "epoch": 0.69, "learning_rate": 3.843784976288907e-05, "loss": 0.0, "step": 35754 }, { "epoch": 0.69, "learning_rate": 3.8437202802631837e-05, "loss": 0.0002, "step": 35756 }, { "epoch": 0.69, "learning_rate": 3.8436555842374606e-05, "loss": 0.0281, "step": 35758 }, { "epoch": 0.69, "learning_rate": 3.8435908882117375e-05, "loss": 0.0109, "step": 35760 }, { "epoch": 0.69, "learning_rate": 3.8435261921860144e-05, "loss": 0.0, "step": 35762 }, { "epoch": 0.69, "learning_rate": 3.843461496160291e-05, "loss": 0.0009, "step": 35764 }, { "epoch": 0.69, "learning_rate": 3.8433968001345675e-05, "loss": 0.0, "step": 35766 }, { "epoch": 0.69, "learning_rate": 3.843332104108845e-05, "loss": 0.0, "step": 35768 }, { "epoch": 0.69, "learning_rate": 3.8432674080831213e-05, "loss": 0.0, "step": 35770 }, { "epoch": 0.69, "learning_rate": 3.843202712057398e-05, "loss": 0.0003, "step": 35772 }, { "epoch": 0.69, "learning_rate": 3.843138016031675e-05, "loss": 0.0034, "step": 35774 }, { "epoch": 0.69, "learning_rate": 3.843073320005952e-05, "loss": 0.0049, "step": 35776 }, { "epoch": 0.69, "learning_rate": 3.8430086239802297e-05, "loss": 0.0036, "step": 35778 }, { "epoch": 0.69, "learning_rate": 3.842943927954506e-05, "loss": 0.0, "step": 35780 }, { "epoch": 0.69, "learning_rate": 3.842879231928783e-05, "loss": 0.0, "step": 35782 }, { "epoch": 0.69, "learning_rate": 3.84281453590306e-05, "loss": 0.001, "step": 35784 }, { "epoch": 0.69, "learning_rate": 3.8427498398773366e-05, "loss": 0.0, "step": 35786 }, { "epoch": 0.69, "learning_rate": 3.842685143851613e-05, "loss": 0.0, "step": 35788 }, { "epoch": 0.69, "learning_rate": 3.8426204478258904e-05, "loss": 0.0, "step": 35790 }, { "epoch": 0.69, "learning_rate": 3.842555751800167e-05, "loss": 0.0, "step": 35792 }, { "epoch": 0.69, "learning_rate": 3.842491055774444e-05, "loss": 0.0001, "step": 35794 }, { "epoch": 0.69, "learning_rate": 3.842426359748721e-05, "loss": 0.0001, "step": 35796 }, { "epoch": 0.69, "learning_rate": 3.8423616637229974e-05, "loss": 0.0032, "step": 35798 }, { "epoch": 0.69, "learning_rate": 3.842296967697275e-05, "loss": 0.0086, "step": 35800 }, { "epoch": 0.69, "learning_rate": 3.842232271671551e-05, "loss": 0.002, "step": 35802 }, { "epoch": 0.69, "learning_rate": 3.842167575645828e-05, "loss": 0.0036, "step": 35804 }, { "epoch": 0.69, "learning_rate": 3.842102879620105e-05, "loss": 0.0001, "step": 35806 }, { "epoch": 0.69, "learning_rate": 3.842038183594382e-05, "loss": 0.0048, "step": 35808 }, { "epoch": 0.7, "learning_rate": 3.841973487568659e-05, "loss": 0.0001, "step": 35810 }, { "epoch": 0.7, "learning_rate": 3.841908791542936e-05, "loss": 0.011, "step": 35812 }, { "epoch": 0.7, "learning_rate": 3.8418440955172126e-05, "loss": 0.0, "step": 35814 }, { "epoch": 0.7, "learning_rate": 3.8417793994914896e-05, "loss": 0.0, "step": 35816 }, { "epoch": 0.7, "learning_rate": 3.841747051478628e-05, "loss": 0.349, "step": 35818 }, { "epoch": 0.7, "learning_rate": 3.841682355452905e-05, "loss": 0.0, "step": 35820 }, { "epoch": 0.7, "learning_rate": 3.841617659427181e-05, "loss": 0.0, "step": 35822 }, { "epoch": 0.7, "learning_rate": 3.841552963401459e-05, "loss": 0.0, "step": 35824 }, { "epoch": 0.7, "learning_rate": 3.841488267375735e-05, "loss": 0.0001, "step": 35826 }, { "epoch": 0.7, "learning_rate": 3.8414235713500126e-05, "loss": 0.1346, "step": 35828 }, { "epoch": 0.7, "learning_rate": 3.841358875324289e-05, "loss": 0.0001, "step": 35830 }, { "epoch": 0.7, "learning_rate": 3.841294179298566e-05, "loss": 0.0001, "step": 35832 }, { "epoch": 0.7, "learning_rate": 3.8412294832728426e-05, "loss": 0.0, "step": 35834 }, { "epoch": 0.7, "learning_rate": 3.8411647872471195e-05, "loss": 0.0, "step": 35836 }, { "epoch": 0.7, "learning_rate": 3.8411000912213964e-05, "loss": 0.0087, "step": 35838 }, { "epoch": 0.7, "learning_rate": 3.841035395195673e-05, "loss": 0.0007, "step": 35840 }, { "epoch": 0.7, "learning_rate": 3.84097069916995e-05, "loss": 0.0012, "step": 35842 }, { "epoch": 0.7, "learning_rate": 3.8409060031442265e-05, "loss": 0.0, "step": 35844 }, { "epoch": 0.7, "learning_rate": 3.840841307118504e-05, "loss": 0.0, "step": 35846 }, { "epoch": 0.7, "learning_rate": 3.840776611092781e-05, "loss": 0.0003, "step": 35848 }, { "epoch": 0.7, "learning_rate": 3.840711915067058e-05, "loss": 0.0001, "step": 35850 }, { "epoch": 0.7, "learning_rate": 3.840647219041335e-05, "loss": 0.0, "step": 35852 }, { "epoch": 0.7, "learning_rate": 3.840582523015611e-05, "loss": 0.0055, "step": 35854 }, { "epoch": 0.7, "learning_rate": 3.8405178269898886e-05, "loss": 0.0, "step": 35856 }, { "epoch": 0.7, "learning_rate": 3.840453130964165e-05, "loss": 0.0001, "step": 35858 }, { "epoch": 0.7, "learning_rate": 3.840388434938442e-05, "loss": 0.0, "step": 35860 }, { "epoch": 0.7, "learning_rate": 3.8403237389127186e-05, "loss": 0.0, "step": 35862 }, { "epoch": 0.7, "learning_rate": 3.8402590428869956e-05, "loss": 0.0, "step": 35864 }, { "epoch": 0.7, "learning_rate": 3.8401943468612725e-05, "loss": 0.0005, "step": 35866 }, { "epoch": 0.7, "learning_rate": 3.8401296508355494e-05, "loss": 0.0, "step": 35868 }, { "epoch": 0.7, "learning_rate": 3.840064954809826e-05, "loss": 0.015, "step": 35870 }, { "epoch": 0.7, "learning_rate": 3.840000258784103e-05, "loss": 0.0002, "step": 35872 }, { "epoch": 0.7, "learning_rate": 3.83993556275838e-05, "loss": 0.0002, "step": 35874 }, { "epoch": 0.7, "learning_rate": 3.839870866732656e-05, "loss": 0.0035, "step": 35876 }, { "epoch": 0.7, "learning_rate": 3.839806170706934e-05, "loss": 0.0005, "step": 35878 }, { "epoch": 0.7, "learning_rate": 3.839741474681211e-05, "loss": 0.0001, "step": 35880 }, { "epoch": 0.7, "learning_rate": 3.839676778655487e-05, "loss": 0.0002, "step": 35882 }, { "epoch": 0.7, "learning_rate": 3.8396120826297646e-05, "loss": 0.0003, "step": 35884 }, { "epoch": 0.7, "learning_rate": 3.839547386604041e-05, "loss": 0.0008, "step": 35886 }, { "epoch": 0.7, "learning_rate": 3.8394826905783185e-05, "loss": 0.0, "step": 35888 }, { "epoch": 0.7, "learning_rate": 3.839417994552595e-05, "loss": 0.0014, "step": 35890 }, { "epoch": 0.7, "learning_rate": 3.8393532985268716e-05, "loss": 0.0, "step": 35892 }, { "epoch": 0.7, "learning_rate": 3.8392886025011485e-05, "loss": 0.0, "step": 35894 }, { "epoch": 0.7, "learning_rate": 3.8392239064754254e-05, "loss": 0.0, "step": 35896 }, { "epoch": 0.7, "learning_rate": 3.839159210449702e-05, "loss": 0.0, "step": 35898 }, { "epoch": 0.7, "learning_rate": 3.839094514423979e-05, "loss": 0.0, "step": 35900 }, { "epoch": 0.7, "learning_rate": 3.839029818398256e-05, "loss": 0.0001, "step": 35902 }, { "epoch": 0.7, "learning_rate": 3.8389651223725324e-05, "loss": 0.0, "step": 35904 }, { "epoch": 0.7, "learning_rate": 3.83890042634681e-05, "loss": 0.0024, "step": 35906 }, { "epoch": 0.7, "learning_rate": 3.838835730321086e-05, "loss": 0.0095, "step": 35908 }, { "epoch": 0.7, "learning_rate": 3.838771034295364e-05, "loss": 0.17, "step": 35910 }, { "epoch": 0.7, "learning_rate": 3.83870633826964e-05, "loss": 0.0003, "step": 35912 }, { "epoch": 0.7, "learning_rate": 3.838641642243917e-05, "loss": 0.0001, "step": 35914 }, { "epoch": 0.7, "learning_rate": 3.8385769462181945e-05, "loss": 0.0, "step": 35916 }, { "epoch": 0.7, "learning_rate": 3.838512250192471e-05, "loss": 0.0001, "step": 35918 }, { "epoch": 0.7, "learning_rate": 3.8384475541667476e-05, "loss": 0.0004, "step": 35920 }, { "epoch": 0.7, "learning_rate": 3.8383828581410245e-05, "loss": 0.0, "step": 35922 }, { "epoch": 0.7, "learning_rate": 3.8383181621153014e-05, "loss": 0.0, "step": 35924 }, { "epoch": 0.7, "learning_rate": 3.8382534660895784e-05, "loss": 0.0137, "step": 35926 }, { "epoch": 0.7, "learning_rate": 3.838188770063855e-05, "loss": 0.0, "step": 35928 }, { "epoch": 0.7, "learning_rate": 3.838124074038132e-05, "loss": 0.0, "step": 35930 }, { "epoch": 0.7, "learning_rate": 3.838059378012409e-05, "loss": 0.0, "step": 35932 }, { "epoch": 0.7, "learning_rate": 3.837994681986686e-05, "loss": 0.0009, "step": 35934 }, { "epoch": 0.7, "learning_rate": 3.837929985960962e-05, "loss": 0.0, "step": 35936 }, { "epoch": 0.7, "learning_rate": 3.83786528993524e-05, "loss": 0.0, "step": 35938 }, { "epoch": 0.7, "learning_rate": 3.837800593909516e-05, "loss": 0.0, "step": 35940 }, { "epoch": 0.7, "learning_rate": 3.837735897883793e-05, "loss": 0.0001, "step": 35942 }, { "epoch": 0.7, "learning_rate": 3.83767120185807e-05, "loss": 0.0058, "step": 35944 }, { "epoch": 0.7, "learning_rate": 3.837606505832347e-05, "loss": 0.0, "step": 35946 }, { "epoch": 0.7, "learning_rate": 3.8375418098066244e-05, "loss": 0.0, "step": 35948 }, { "epoch": 0.7, "learning_rate": 3.8374771137809006e-05, "loss": 0.0, "step": 35950 }, { "epoch": 0.7, "learning_rate": 3.8374124177551775e-05, "loss": 0.0001, "step": 35952 }, { "epoch": 0.7, "learning_rate": 3.8373477217294544e-05, "loss": 0.0, "step": 35954 }, { "epoch": 0.7, "learning_rate": 3.837283025703731e-05, "loss": 0.0061, "step": 35956 }, { "epoch": 0.7, "learning_rate": 3.8372183296780075e-05, "loss": 0.0001, "step": 35958 }, { "epoch": 0.7, "learning_rate": 3.837153633652285e-05, "loss": 0.0, "step": 35960 }, { "epoch": 0.7, "learning_rate": 3.837088937626562e-05, "loss": 0.0174, "step": 35962 }, { "epoch": 0.7, "learning_rate": 3.837024241600838e-05, "loss": 0.0002, "step": 35964 }, { "epoch": 0.7, "learning_rate": 3.836959545575116e-05, "loss": 0.0, "step": 35966 }, { "epoch": 0.7, "learning_rate": 3.836894849549392e-05, "loss": 0.0, "step": 35968 }, { "epoch": 0.7, "learning_rate": 3.83683015352367e-05, "loss": 0.0231, "step": 35970 }, { "epoch": 0.7, "learning_rate": 3.836765457497946e-05, "loss": 0.0028, "step": 35972 }, { "epoch": 0.7, "learning_rate": 3.836700761472223e-05, "loss": 0.0138, "step": 35974 }, { "epoch": 0.7, "learning_rate": 3.8366360654465e-05, "loss": 0.0, "step": 35976 }, { "epoch": 0.7, "learning_rate": 3.8365713694207766e-05, "loss": 0.0001, "step": 35978 }, { "epoch": 0.7, "learning_rate": 3.8365066733950535e-05, "loss": 0.0001, "step": 35980 }, { "epoch": 0.7, "learning_rate": 3.8364419773693304e-05, "loss": 0.0, "step": 35982 }, { "epoch": 0.7, "learning_rate": 3.8363772813436073e-05, "loss": 0.0, "step": 35984 }, { "epoch": 0.7, "learning_rate": 3.8363125853178836e-05, "loss": 0.0001, "step": 35986 }, { "epoch": 0.7, "learning_rate": 3.836247889292161e-05, "loss": 0.0, "step": 35988 }, { "epoch": 0.7, "learning_rate": 3.8361831932664374e-05, "loss": 0.0001, "step": 35990 }, { "epoch": 0.7, "learning_rate": 3.836118497240715e-05, "loss": 0.0, "step": 35992 }, { "epoch": 0.7, "learning_rate": 3.836053801214992e-05, "loss": 0.0008, "step": 35994 }, { "epoch": 0.7, "learning_rate": 3.835989105189268e-05, "loss": 0.0, "step": 35996 }, { "epoch": 0.7, "learning_rate": 3.835924409163546e-05, "loss": 0.0003, "step": 35998 }, { "epoch": 0.7, "learning_rate": 3.835859713137822e-05, "loss": 0.0001, "step": 36000 }, { "epoch": 0.7, "learning_rate": 3.835795017112099e-05, "loss": 0.0003, "step": 36002 }, { "epoch": 0.7, "learning_rate": 3.835730321086376e-05, "loss": 0.0, "step": 36004 }, { "epoch": 0.7, "learning_rate": 3.835665625060653e-05, "loss": 0.0005, "step": 36006 }, { "epoch": 0.7, "learning_rate": 3.8356009290349296e-05, "loss": 0.0005, "step": 36008 }, { "epoch": 0.7, "learning_rate": 3.8355362330092065e-05, "loss": 0.0001, "step": 36010 }, { "epoch": 0.7, "learning_rate": 3.8354715369834834e-05, "loss": 0.0, "step": 36012 }, { "epoch": 0.7, "learning_rate": 3.83540684095776e-05, "loss": 0.0, "step": 36014 }, { "epoch": 0.7, "learning_rate": 3.835342144932037e-05, "loss": 0.0212, "step": 36016 }, { "epoch": 0.7, "learning_rate": 3.8352774489063134e-05, "loss": 0.0, "step": 36018 }, { "epoch": 0.7, "learning_rate": 3.835212752880591e-05, "loss": 0.004, "step": 36020 }, { "epoch": 0.7, "learning_rate": 3.835148056854867e-05, "loss": 0.0004, "step": 36022 }, { "epoch": 0.7, "learning_rate": 3.835083360829144e-05, "loss": 0.006, "step": 36024 }, { "epoch": 0.7, "learning_rate": 3.835018664803422e-05, "loss": 0.0001, "step": 36026 }, { "epoch": 0.7, "learning_rate": 3.834953968777698e-05, "loss": 0.0031, "step": 36028 }, { "epoch": 0.7, "learning_rate": 3.8348892727519756e-05, "loss": 0.0011, "step": 36030 }, { "epoch": 0.7, "learning_rate": 3.834824576726252e-05, "loss": 0.0, "step": 36032 }, { "epoch": 0.7, "learning_rate": 3.834759880700529e-05, "loss": 0.0, "step": 36034 }, { "epoch": 0.7, "learning_rate": 3.8346951846748056e-05, "loss": 0.0036, "step": 36036 }, { "epoch": 0.7, "learning_rate": 3.8346304886490825e-05, "loss": 0.0, "step": 36038 }, { "epoch": 0.7, "learning_rate": 3.8345657926233594e-05, "loss": 0.0004, "step": 36040 }, { "epoch": 0.7, "learning_rate": 3.834501096597636e-05, "loss": 0.0, "step": 36042 }, { "epoch": 0.7, "learning_rate": 3.834436400571913e-05, "loss": 0.0, "step": 36044 }, { "epoch": 0.7, "learning_rate": 3.8343717045461895e-05, "loss": 0.001, "step": 36046 }, { "epoch": 0.7, "learning_rate": 3.834307008520467e-05, "loss": 0.0, "step": 36048 }, { "epoch": 0.7, "learning_rate": 3.834242312494743e-05, "loss": 0.0001, "step": 36050 }, { "epoch": 0.7, "learning_rate": 3.834177616469021e-05, "loss": 0.0022, "step": 36052 }, { "epoch": 0.7, "learning_rate": 3.834112920443297e-05, "loss": 0.0, "step": 36054 }, { "epoch": 0.7, "learning_rate": 3.834048224417574e-05, "loss": 0.0003, "step": 36056 }, { "epoch": 0.7, "learning_rate": 3.833983528391851e-05, "loss": 0.0006, "step": 36058 }, { "epoch": 0.7, "learning_rate": 3.833918832366128e-05, "loss": 0.0102, "step": 36060 }, { "epoch": 0.7, "learning_rate": 3.833854136340405e-05, "loss": 0.007, "step": 36062 }, { "epoch": 0.7, "learning_rate": 3.8337894403146817e-05, "loss": 0.0, "step": 36064 }, { "epoch": 0.7, "learning_rate": 3.8337247442889586e-05, "loss": 0.0003, "step": 36066 }, { "epoch": 0.7, "learning_rate": 3.833660048263235e-05, "loss": 0.0, "step": 36068 }, { "epoch": 0.7, "learning_rate": 3.8335953522375124e-05, "loss": 0.0006, "step": 36070 }, { "epoch": 0.7, "learning_rate": 3.833530656211789e-05, "loss": 0.0, "step": 36072 }, { "epoch": 0.7, "learning_rate": 3.833465960186066e-05, "loss": 0.0075, "step": 36074 }, { "epoch": 0.7, "learning_rate": 3.833401264160343e-05, "loss": 0.0006, "step": 36076 }, { "epoch": 0.7, "learning_rate": 3.833336568134619e-05, "loss": 0.0007, "step": 36078 }, { "epoch": 0.7, "learning_rate": 3.833271872108897e-05, "loss": 0.0127, "step": 36080 }, { "epoch": 0.7, "learning_rate": 3.833207176083173e-05, "loss": 0.0, "step": 36082 }, { "epoch": 0.7, "learning_rate": 3.83314248005745e-05, "loss": 0.0001, "step": 36084 }, { "epoch": 0.7, "learning_rate": 3.833077784031727e-05, "loss": 0.0001, "step": 36086 }, { "epoch": 0.7, "learning_rate": 3.833013088006004e-05, "loss": 0.0065, "step": 36088 }, { "epoch": 0.7, "learning_rate": 3.832948391980281e-05, "loss": 0.0, "step": 36090 }, { "epoch": 0.7, "learning_rate": 3.832883695954558e-05, "loss": 0.0001, "step": 36092 }, { "epoch": 0.7, "learning_rate": 3.8328189999288346e-05, "loss": 0.0, "step": 36094 }, { "epoch": 0.7, "learning_rate": 3.8327543039031115e-05, "loss": 0.0007, "step": 36096 }, { "epoch": 0.7, "learning_rate": 3.8326896078773884e-05, "loss": 0.0005, "step": 36098 }, { "epoch": 0.7, "learning_rate": 3.8326249118516646e-05, "loss": 0.0, "step": 36100 }, { "epoch": 0.7, "learning_rate": 3.832560215825942e-05, "loss": 0.0002, "step": 36102 }, { "epoch": 0.7, "learning_rate": 3.832495519800219e-05, "loss": 0.0003, "step": 36104 }, { "epoch": 0.7, "learning_rate": 3.8324308237744954e-05, "loss": 0.0083, "step": 36106 }, { "epoch": 0.7, "learning_rate": 3.832366127748773e-05, "loss": 0.0003, "step": 36108 }, { "epoch": 0.7, "learning_rate": 3.832301431723049e-05, "loss": 0.0, "step": 36110 }, { "epoch": 0.7, "learning_rate": 3.832236735697327e-05, "loss": 0.0, "step": 36112 }, { "epoch": 0.7, "learning_rate": 3.832172039671603e-05, "loss": 0.0, "step": 36114 }, { "epoch": 0.7, "learning_rate": 3.83210734364588e-05, "loss": 0.0, "step": 36116 }, { "epoch": 0.7, "learning_rate": 3.832042647620157e-05, "loss": 0.0003, "step": 36118 }, { "epoch": 0.7, "learning_rate": 3.831977951594434e-05, "loss": 0.0, "step": 36120 }, { "epoch": 0.7, "learning_rate": 3.8319132555687106e-05, "loss": 0.0016, "step": 36122 }, { "epoch": 0.7, "learning_rate": 3.8318485595429875e-05, "loss": 0.0026, "step": 36124 }, { "epoch": 0.7, "learning_rate": 3.8317838635172645e-05, "loss": 0.0024, "step": 36126 }, { "epoch": 0.7, "learning_rate": 3.831719167491541e-05, "loss": 0.0, "step": 36128 }, { "epoch": 0.7, "learning_rate": 3.831654471465818e-05, "loss": 0.0002, "step": 36130 }, { "epoch": 0.7, "learning_rate": 3.8315897754400945e-05, "loss": 0.0005, "step": 36132 }, { "epoch": 0.7, "learning_rate": 3.831525079414372e-05, "loss": 0.0, "step": 36134 }, { "epoch": 0.7, "learning_rate": 3.831460383388648e-05, "loss": 0.0001, "step": 36136 }, { "epoch": 0.7, "learning_rate": 3.831395687362925e-05, "loss": 0.0, "step": 36138 }, { "epoch": 0.7, "learning_rate": 3.831330991337203e-05, "loss": 0.0, "step": 36140 }, { "epoch": 0.7, "learning_rate": 3.831266295311479e-05, "loss": 0.014, "step": 36142 }, { "epoch": 0.7, "learning_rate": 3.831201599285756e-05, "loss": 0.0001, "step": 36144 }, { "epoch": 0.7, "learning_rate": 3.831136903260033e-05, "loss": 0.0001, "step": 36146 }, { "epoch": 0.7, "learning_rate": 3.83107220723431e-05, "loss": 0.0, "step": 36148 }, { "epoch": 0.7, "learning_rate": 3.831007511208587e-05, "loss": 0.0017, "step": 36150 }, { "epoch": 0.7, "learning_rate": 3.8309428151828636e-05, "loss": 0.0, "step": 36152 }, { "epoch": 0.7, "learning_rate": 3.8308781191571405e-05, "loss": 0.0086, "step": 36154 }, { "epoch": 0.7, "learning_rate": 3.8308134231314174e-05, "loss": 0.0, "step": 36156 }, { "epoch": 0.7, "learning_rate": 3.830748727105694e-05, "loss": 0.0001, "step": 36158 }, { "epoch": 0.7, "learning_rate": 3.8306840310799705e-05, "loss": 0.0, "step": 36160 }, { "epoch": 0.7, "learning_rate": 3.830619335054248e-05, "loss": 0.0003, "step": 36162 }, { "epoch": 0.7, "learning_rate": 3.8305546390285244e-05, "loss": 0.0, "step": 36164 }, { "epoch": 0.7, "learning_rate": 3.830489943002801e-05, "loss": 0.0001, "step": 36166 }, { "epoch": 0.7, "learning_rate": 3.830425246977078e-05, "loss": 0.0037, "step": 36168 }, { "epoch": 0.7, "learning_rate": 3.830360550951355e-05, "loss": 0.0001, "step": 36170 }, { "epoch": 0.7, "learning_rate": 3.830295854925633e-05, "loss": 0.0, "step": 36172 }, { "epoch": 0.7, "learning_rate": 3.830231158899909e-05, "loss": 0.0002, "step": 36174 }, { "epoch": 0.7, "learning_rate": 3.830166462874186e-05, "loss": 0.0, "step": 36176 }, { "epoch": 0.7, "learning_rate": 3.830101766848463e-05, "loss": 0.0, "step": 36178 }, { "epoch": 0.7, "learning_rate": 3.8300370708227396e-05, "loss": 0.0015, "step": 36180 }, { "epoch": 0.7, "learning_rate": 3.8299723747970165e-05, "loss": 0.0022, "step": 36182 }, { "epoch": 0.7, "learning_rate": 3.8299076787712934e-05, "loss": 0.0, "step": 36184 }, { "epoch": 0.7, "learning_rate": 3.8298429827455704e-05, "loss": 0.0001, "step": 36186 }, { "epoch": 0.7, "learning_rate": 3.8297782867198466e-05, "loss": 0.0117, "step": 36188 }, { "epoch": 0.7, "learning_rate": 3.829713590694124e-05, "loss": 0.0, "step": 36190 }, { "epoch": 0.7, "learning_rate": 3.8296488946684004e-05, "loss": 0.0003, "step": 36192 }, { "epoch": 0.7, "learning_rate": 3.829584198642678e-05, "loss": 0.0002, "step": 36194 }, { "epoch": 0.7, "learning_rate": 3.829519502616954e-05, "loss": 0.0, "step": 36196 }, { "epoch": 0.7, "learning_rate": 3.829454806591231e-05, "loss": 0.0002, "step": 36198 }, { "epoch": 0.7, "learning_rate": 3.829390110565508e-05, "loss": 0.0002, "step": 36200 }, { "epoch": 0.7, "learning_rate": 3.829325414539785e-05, "loss": 0.0001, "step": 36202 }, { "epoch": 0.7, "learning_rate": 3.829260718514062e-05, "loss": 0.0, "step": 36204 }, { "epoch": 0.7, "learning_rate": 3.829196022488339e-05, "loss": 0.0, "step": 36206 }, { "epoch": 0.7, "learning_rate": 3.829131326462616e-05, "loss": 0.0004, "step": 36208 }, { "epoch": 0.7, "learning_rate": 3.829066630436892e-05, "loss": 0.0001, "step": 36210 }, { "epoch": 0.7, "learning_rate": 3.8290019344111695e-05, "loss": 0.0, "step": 36212 }, { "epoch": 0.7, "learning_rate": 3.828937238385446e-05, "loss": 0.0006, "step": 36214 }, { "epoch": 0.7, "learning_rate": 3.828872542359723e-05, "loss": 0.0, "step": 36216 }, { "epoch": 0.7, "learning_rate": 3.828807846334e-05, "loss": 0.02, "step": 36218 }, { "epoch": 0.7, "learning_rate": 3.8287431503082764e-05, "loss": 0.0, "step": 36220 }, { "epoch": 0.7, "learning_rate": 3.828678454282554e-05, "loss": 0.0, "step": 36222 }, { "epoch": 0.7, "learning_rate": 3.82861375825683e-05, "loss": 0.0006, "step": 36224 }, { "epoch": 0.7, "learning_rate": 3.828549062231107e-05, "loss": 0.0025, "step": 36226 }, { "epoch": 0.7, "learning_rate": 3.828484366205384e-05, "loss": 0.0046, "step": 36228 }, { "epoch": 0.7, "learning_rate": 3.828419670179661e-05, "loss": 0.0001, "step": 36230 }, { "epoch": 0.7, "learning_rate": 3.828354974153938e-05, "loss": 0.0009, "step": 36232 }, { "epoch": 0.7, "learning_rate": 3.828290278128215e-05, "loss": 0.0006, "step": 36234 }, { "epoch": 0.7, "learning_rate": 3.828225582102492e-05, "loss": 0.0064, "step": 36236 }, { "epoch": 0.7, "learning_rate": 3.8281608860767686e-05, "loss": 0.0004, "step": 36238 }, { "epoch": 0.7, "learning_rate": 3.8280961900510455e-05, "loss": 0.0, "step": 36240 }, { "epoch": 0.7, "learning_rate": 3.828031494025322e-05, "loss": 0.0002, "step": 36242 }, { "epoch": 0.7, "learning_rate": 3.8279667979995993e-05, "loss": 0.0, "step": 36244 }, { "epoch": 0.7, "learning_rate": 3.8279021019738756e-05, "loss": 0.0087, "step": 36246 }, { "epoch": 0.7, "learning_rate": 3.8278374059481525e-05, "loss": 0.0001, "step": 36248 }, { "epoch": 0.7, "learning_rate": 3.82777270992243e-05, "loss": 0.0014, "step": 36250 }, { "epoch": 0.7, "learning_rate": 3.827708013896706e-05, "loss": 0.0001, "step": 36252 }, { "epoch": 0.7, "learning_rate": 3.827643317870984e-05, "loss": 0.0014, "step": 36254 }, { "epoch": 0.7, "learning_rate": 3.82757862184526e-05, "loss": 0.0001, "step": 36256 }, { "epoch": 0.7, "learning_rate": 3.827513925819537e-05, "loss": 0.0003, "step": 36258 }, { "epoch": 0.7, "learning_rate": 3.827449229793814e-05, "loss": 0.0008, "step": 36260 }, { "epoch": 0.7, "learning_rate": 3.827384533768091e-05, "loss": 0.0, "step": 36262 }, { "epoch": 0.7, "learning_rate": 3.827319837742368e-05, "loss": 0.0044, "step": 36264 }, { "epoch": 0.7, "learning_rate": 3.8272551417166447e-05, "loss": 0.0, "step": 36266 }, { "epoch": 0.7, "learning_rate": 3.8271904456909216e-05, "loss": 0.0, "step": 36268 }, { "epoch": 0.7, "learning_rate": 3.827125749665198e-05, "loss": 0.0009, "step": 36270 }, { "epoch": 0.7, "learning_rate": 3.8270610536394754e-05, "loss": 0.0001, "step": 36272 }, { "epoch": 0.7, "learning_rate": 3.8269963576137516e-05, "loss": 0.0075, "step": 36274 }, { "epoch": 0.7, "learning_rate": 3.826931661588029e-05, "loss": 0.0, "step": 36276 }, { "epoch": 0.7, "learning_rate": 3.8268669655623054e-05, "loss": 0.0002, "step": 36278 }, { "epoch": 0.7, "learning_rate": 3.8268022695365823e-05, "loss": 0.005, "step": 36280 }, { "epoch": 0.7, "learning_rate": 3.82673757351086e-05, "loss": 0.0, "step": 36282 }, { "epoch": 0.7, "learning_rate": 3.826672877485136e-05, "loss": 0.0, "step": 36284 }, { "epoch": 0.7, "learning_rate": 3.826608181459413e-05, "loss": 0.0027, "step": 36286 }, { "epoch": 0.7, "learning_rate": 3.82654348543369e-05, "loss": 0.0004, "step": 36288 }, { "epoch": 0.7, "learning_rate": 3.826478789407967e-05, "loss": 0.0005, "step": 36290 }, { "epoch": 0.7, "learning_rate": 3.826414093382243e-05, "loss": 0.0002, "step": 36292 }, { "epoch": 0.7, "learning_rate": 3.826349397356521e-05, "loss": 0.0001, "step": 36294 }, { "epoch": 0.7, "learning_rate": 3.8262847013307976e-05, "loss": 0.0048, "step": 36296 }, { "epoch": 0.7, "learning_rate": 3.8262200053050745e-05, "loss": 0.0, "step": 36298 }, { "epoch": 0.7, "learning_rate": 3.8261553092793514e-05, "loss": 0.0001, "step": 36300 }, { "epoch": 0.7, "learning_rate": 3.8260906132536277e-05, "loss": 0.0001, "step": 36302 }, { "epoch": 0.7, "learning_rate": 3.826025917227905e-05, "loss": 0.0, "step": 36304 }, { "epoch": 0.7, "learning_rate": 3.8259612212021815e-05, "loss": 0.0, "step": 36306 }, { "epoch": 0.7, "learning_rate": 3.8258965251764584e-05, "loss": 0.0, "step": 36308 }, { "epoch": 0.7, "learning_rate": 3.825831829150735e-05, "loss": 0.0001, "step": 36310 }, { "epoch": 0.7, "learning_rate": 3.825767133125012e-05, "loss": 0.002, "step": 36312 }, { "epoch": 0.7, "learning_rate": 3.825702437099289e-05, "loss": 0.0, "step": 36314 }, { "epoch": 0.7, "learning_rate": 3.825637741073566e-05, "loss": 0.0014, "step": 36316 }, { "epoch": 0.7, "learning_rate": 3.825573045047843e-05, "loss": 0.0041, "step": 36318 }, { "epoch": 0.7, "learning_rate": 3.82550834902212e-05, "loss": 0.0003, "step": 36320 }, { "epoch": 0.7, "learning_rate": 3.825443652996397e-05, "loss": 0.0003, "step": 36322 }, { "epoch": 0.71, "learning_rate": 3.825378956970673e-05, "loss": 0.0061, "step": 36324 }, { "epoch": 0.71, "learning_rate": 3.8253142609449506e-05, "loss": 0.003, "step": 36326 }, { "epoch": 0.71, "learning_rate": 3.8252495649192275e-05, "loss": 0.0, "step": 36328 }, { "epoch": 0.71, "learning_rate": 3.825184868893504e-05, "loss": 0.0002, "step": 36330 }, { "epoch": 0.71, "learning_rate": 3.825120172867781e-05, "loss": 0.0012, "step": 36332 }, { "epoch": 0.71, "learning_rate": 3.8250554768420575e-05, "loss": 0.0, "step": 36334 }, { "epoch": 0.71, "learning_rate": 3.824990780816335e-05, "loss": 0.0, "step": 36336 }, { "epoch": 0.71, "learning_rate": 3.824926084790611e-05, "loss": 0.0, "step": 36338 }, { "epoch": 0.71, "learning_rate": 3.824861388764888e-05, "loss": 0.0111, "step": 36340 }, { "epoch": 0.71, "learning_rate": 3.824796692739165e-05, "loss": 0.0001, "step": 36342 }, { "epoch": 0.71, "learning_rate": 3.824731996713442e-05, "loss": 0.0, "step": 36344 }, { "epoch": 0.71, "learning_rate": 3.824667300687719e-05, "loss": 0.0, "step": 36346 }, { "epoch": 0.71, "learning_rate": 3.824602604661996e-05, "loss": 0.0207, "step": 36348 }, { "epoch": 0.71, "learning_rate": 3.824537908636273e-05, "loss": 0.0005, "step": 36350 }, { "epoch": 0.71, "learning_rate": 3.824473212610549e-05, "loss": 0.0156, "step": 36352 }, { "epoch": 0.71, "learning_rate": 3.8244085165848266e-05, "loss": 0.0, "step": 36354 }, { "epoch": 0.71, "learning_rate": 3.824343820559103e-05, "loss": 0.0001, "step": 36356 }, { "epoch": 0.71, "learning_rate": 3.8242791245333804e-05, "loss": 0.0002, "step": 36358 }, { "epoch": 0.71, "learning_rate": 3.824214428507657e-05, "loss": 0.0001, "step": 36360 }, { "epoch": 0.71, "learning_rate": 3.8241497324819336e-05, "loss": 0.0001, "step": 36362 }, { "epoch": 0.71, "learning_rate": 3.824085036456211e-05, "loss": 0.0001, "step": 36364 }, { "epoch": 0.71, "learning_rate": 3.8240203404304874e-05, "loss": 0.0003, "step": 36366 }, { "epoch": 0.71, "learning_rate": 3.823955644404764e-05, "loss": 0.0003, "step": 36368 }, { "epoch": 0.71, "learning_rate": 3.823890948379041e-05, "loss": 0.0003, "step": 36370 }, { "epoch": 0.71, "learning_rate": 3.823826252353318e-05, "loss": 0.0014, "step": 36372 }, { "epoch": 0.71, "learning_rate": 3.823761556327595e-05, "loss": 0.0, "step": 36374 }, { "epoch": 0.71, "learning_rate": 3.823696860301872e-05, "loss": 0.0001, "step": 36376 }, { "epoch": 0.71, "learning_rate": 3.823632164276149e-05, "loss": 0.0037, "step": 36378 }, { "epoch": 0.71, "learning_rate": 3.823567468250426e-05, "loss": 0.0, "step": 36380 }, { "epoch": 0.71, "learning_rate": 3.8235027722247026e-05, "loss": 0.0013, "step": 36382 }, { "epoch": 0.71, "learning_rate": 3.823438076198979e-05, "loss": 0.0018, "step": 36384 }, { "epoch": 0.71, "learning_rate": 3.8233733801732565e-05, "loss": 0.0, "step": 36386 }, { "epoch": 0.71, "learning_rate": 3.823308684147533e-05, "loss": 0.0016, "step": 36388 }, { "epoch": 0.71, "learning_rate": 3.8232439881218096e-05, "loss": 0.0, "step": 36390 }, { "epoch": 0.71, "learning_rate": 3.8231792920960865e-05, "loss": 0.0, "step": 36392 }, { "epoch": 0.71, "learning_rate": 3.8231145960703634e-05, "loss": 0.0, "step": 36394 }, { "epoch": 0.71, "learning_rate": 3.823049900044641e-05, "loss": 0.0, "step": 36396 }, { "epoch": 0.71, "learning_rate": 3.822985204018917e-05, "loss": 0.0001, "step": 36398 }, { "epoch": 0.71, "learning_rate": 3.822920507993194e-05, "loss": 0.0001, "step": 36400 }, { "epoch": 0.71, "learning_rate": 3.822855811967471e-05, "loss": 0.0006, "step": 36402 }, { "epoch": 0.71, "learning_rate": 3.822791115941748e-05, "loss": 0.0014, "step": 36404 }, { "epoch": 0.71, "learning_rate": 3.822726419916025e-05, "loss": 0.0001, "step": 36406 }, { "epoch": 0.71, "learning_rate": 3.822661723890302e-05, "loss": 0.0, "step": 36408 }, { "epoch": 0.71, "learning_rate": 3.822597027864579e-05, "loss": 0.0004, "step": 36410 }, { "epoch": 0.71, "learning_rate": 3.822532331838855e-05, "loss": 0.0, "step": 36412 }, { "epoch": 0.71, "learning_rate": 3.8224676358131325e-05, "loss": 0.0, "step": 36414 }, { "epoch": 0.71, "learning_rate": 3.822402939787409e-05, "loss": 0.0017, "step": 36416 }, { "epoch": 0.71, "learning_rate": 3.822338243761686e-05, "loss": 0.0, "step": 36418 }, { "epoch": 0.71, "learning_rate": 3.8222735477359625e-05, "loss": 0.0, "step": 36420 }, { "epoch": 0.71, "learning_rate": 3.8222088517102395e-05, "loss": 0.0186, "step": 36422 }, { "epoch": 0.71, "learning_rate": 3.8221441556845164e-05, "loss": 0.0054, "step": 36424 }, { "epoch": 0.71, "learning_rate": 3.822079459658793e-05, "loss": 0.0037, "step": 36426 }, { "epoch": 0.71, "learning_rate": 3.82201476363307e-05, "loss": 0.0, "step": 36428 }, { "epoch": 0.71, "learning_rate": 3.821950067607347e-05, "loss": 0.0, "step": 36430 }, { "epoch": 0.71, "learning_rate": 3.821885371581624e-05, "loss": 0.0, "step": 36432 }, { "epoch": 0.71, "learning_rate": 3.8218206755559e-05, "loss": 0.0124, "step": 36434 }, { "epoch": 0.71, "learning_rate": 3.821755979530178e-05, "loss": 0.0001, "step": 36436 }, { "epoch": 0.71, "learning_rate": 3.821691283504454e-05, "loss": 0.0, "step": 36438 }, { "epoch": 0.71, "learning_rate": 3.8216265874787316e-05, "loss": 0.0001, "step": 36440 }, { "epoch": 0.71, "learning_rate": 3.8215618914530085e-05, "loss": 0.0005, "step": 36442 }, { "epoch": 0.71, "learning_rate": 3.821497195427285e-05, "loss": 0.0, "step": 36444 }, { "epoch": 0.71, "learning_rate": 3.8214324994015624e-05, "loss": 0.0052, "step": 36446 }, { "epoch": 0.71, "learning_rate": 3.8213678033758386e-05, "loss": 0.0001, "step": 36448 }, { "epoch": 0.71, "learning_rate": 3.8213031073501155e-05, "loss": 0.0, "step": 36450 }, { "epoch": 0.71, "learning_rate": 3.8212384113243924e-05, "loss": 0.0001, "step": 36452 }, { "epoch": 0.71, "learning_rate": 3.821173715298669e-05, "loss": 0.0031, "step": 36454 }, { "epoch": 0.71, "learning_rate": 3.821109019272946e-05, "loss": 0.0055, "step": 36456 }, { "epoch": 0.71, "learning_rate": 3.821044323247223e-05, "loss": 0.0, "step": 36458 }, { "epoch": 0.71, "learning_rate": 3.8209796272215e-05, "loss": 0.0, "step": 36460 }, { "epoch": 0.71, "learning_rate": 3.820914931195777e-05, "loss": 0.0, "step": 36462 }, { "epoch": 0.71, "learning_rate": 3.820850235170054e-05, "loss": 0.0001, "step": 36464 }, { "epoch": 0.71, "learning_rate": 3.82078553914433e-05, "loss": 0.0001, "step": 36466 }, { "epoch": 0.71, "learning_rate": 3.820720843118608e-05, "loss": 0.0005, "step": 36468 }, { "epoch": 0.71, "learning_rate": 3.820656147092884e-05, "loss": 0.0001, "step": 36470 }, { "epoch": 0.71, "learning_rate": 3.820591451067161e-05, "loss": 0.0001, "step": 36472 }, { "epoch": 0.71, "learning_rate": 3.8205267550414384e-05, "loss": 0.0001, "step": 36474 }, { "epoch": 0.71, "learning_rate": 3.8204620590157146e-05, "loss": 0.0008, "step": 36476 }, { "epoch": 0.71, "learning_rate": 3.820397362989992e-05, "loss": 0.0, "step": 36478 }, { "epoch": 0.71, "learning_rate": 3.8203326669642684e-05, "loss": 0.0, "step": 36480 }, { "epoch": 0.71, "learning_rate": 3.8202679709385454e-05, "loss": 0.0379, "step": 36482 }, { "epoch": 0.71, "learning_rate": 3.820203274912822e-05, "loss": 0.0, "step": 36484 }, { "epoch": 0.71, "learning_rate": 3.820138578887099e-05, "loss": 0.0001, "step": 36486 }, { "epoch": 0.71, "learning_rate": 3.820073882861376e-05, "loss": 0.0009, "step": 36488 }, { "epoch": 0.71, "learning_rate": 3.820009186835653e-05, "loss": 0.0001, "step": 36490 }, { "epoch": 0.71, "learning_rate": 3.81994449080993e-05, "loss": 0.0004, "step": 36492 }, { "epoch": 0.71, "learning_rate": 3.819879794784206e-05, "loss": 0.0005, "step": 36494 }, { "epoch": 0.71, "learning_rate": 3.819815098758484e-05, "loss": 0.0001, "step": 36496 }, { "epoch": 0.71, "learning_rate": 3.81975040273276e-05, "loss": 0.001, "step": 36498 }, { "epoch": 0.71, "learning_rate": 3.8196857067070375e-05, "loss": 0.0624, "step": 36500 }, { "epoch": 0.71, "learning_rate": 3.819621010681314e-05, "loss": 0.0001, "step": 36502 }, { "epoch": 0.71, "learning_rate": 3.819556314655591e-05, "loss": 0.0, "step": 36504 }, { "epoch": 0.71, "learning_rate": 3.819491618629868e-05, "loss": 0.0, "step": 36506 }, { "epoch": 0.71, "learning_rate": 3.8194269226041445e-05, "loss": 0.0091, "step": 36508 }, { "epoch": 0.71, "learning_rate": 3.8193622265784214e-05, "loss": 0.0002, "step": 36510 }, { "epoch": 0.71, "learning_rate": 3.819297530552698e-05, "loss": 0.0013, "step": 36512 }, { "epoch": 0.71, "learning_rate": 3.819232834526975e-05, "loss": 0.0, "step": 36514 }, { "epoch": 0.71, "learning_rate": 3.819168138501252e-05, "loss": 0.0013, "step": 36516 }, { "epoch": 0.71, "learning_rate": 3.819103442475529e-05, "loss": 0.0, "step": 36518 }, { "epoch": 0.71, "learning_rate": 3.819038746449806e-05, "loss": 0.0, "step": 36520 }, { "epoch": 0.71, "learning_rate": 3.818974050424083e-05, "loss": 0.0001, "step": 36522 }, { "epoch": 0.71, "learning_rate": 3.81890935439836e-05, "loss": 0.0002, "step": 36524 }, { "epoch": 0.71, "learning_rate": 3.818844658372636e-05, "loss": 0.0001, "step": 36526 }, { "epoch": 0.71, "learning_rate": 3.8187799623469136e-05, "loss": 0.0001, "step": 36528 }, { "epoch": 0.71, "learning_rate": 3.81871526632119e-05, "loss": 0.0, "step": 36530 }, { "epoch": 0.71, "learning_rate": 3.818650570295467e-05, "loss": 0.0, "step": 36532 }, { "epoch": 0.71, "learning_rate": 3.8185858742697436e-05, "loss": 0.0, "step": 36534 }, { "epoch": 0.71, "learning_rate": 3.8185211782440205e-05, "loss": 0.0134, "step": 36536 }, { "epoch": 0.71, "learning_rate": 3.818456482218298e-05, "loss": 0.0021, "step": 36538 }, { "epoch": 0.71, "learning_rate": 3.8183917861925743e-05, "loss": 0.0001, "step": 36540 }, { "epoch": 0.71, "learning_rate": 3.818327090166851e-05, "loss": 0.0, "step": 36542 }, { "epoch": 0.71, "learning_rate": 3.818262394141128e-05, "loss": 0.0, "step": 36544 }, { "epoch": 0.71, "learning_rate": 3.818197698115405e-05, "loss": 0.0, "step": 36546 }, { "epoch": 0.71, "learning_rate": 3.818133002089681e-05, "loss": 0.0003, "step": 36548 }, { "epoch": 0.71, "learning_rate": 3.818068306063959e-05, "loss": 0.0, "step": 36550 }, { "epoch": 0.71, "learning_rate": 3.818003610038236e-05, "loss": 0.0001, "step": 36552 }, { "epoch": 0.71, "learning_rate": 3.817938914012512e-05, "loss": 0.0, "step": 36554 }, { "epoch": 0.71, "learning_rate": 3.8178742179867896e-05, "loss": 0.0001, "step": 36556 }, { "epoch": 0.71, "learning_rate": 3.817809521961066e-05, "loss": 0.0, "step": 36558 }, { "epoch": 0.71, "learning_rate": 3.8177448259353434e-05, "loss": 0.0001, "step": 36560 }, { "epoch": 0.71, "learning_rate": 3.8176801299096197e-05, "loss": 0.0, "step": 36562 }, { "epoch": 0.71, "learning_rate": 3.8176154338838966e-05, "loss": 0.0, "step": 36564 }, { "epoch": 0.71, "learning_rate": 3.8175507378581735e-05, "loss": 0.0004, "step": 36566 }, { "epoch": 0.71, "learning_rate": 3.8174860418324504e-05, "loss": 0.0, "step": 36568 }, { "epoch": 0.71, "learning_rate": 3.817421345806727e-05, "loss": 0.0133, "step": 36570 }, { "epoch": 0.71, "learning_rate": 3.817356649781004e-05, "loss": 0.0002, "step": 36572 }, { "epoch": 0.71, "learning_rate": 3.817291953755281e-05, "loss": 0.0003, "step": 36574 }, { "epoch": 0.71, "learning_rate": 3.817227257729557e-05, "loss": 0.0002, "step": 36576 }, { "epoch": 0.71, "learning_rate": 3.817162561703835e-05, "loss": 0.0, "step": 36578 }, { "epoch": 0.71, "learning_rate": 3.817097865678111e-05, "loss": 0.0037, "step": 36580 }, { "epoch": 0.71, "learning_rate": 3.817033169652389e-05, "loss": 0.0126, "step": 36582 }, { "epoch": 0.71, "learning_rate": 3.8169684736266656e-05, "loss": 0.0, "step": 36584 }, { "epoch": 0.71, "learning_rate": 3.816903777600942e-05, "loss": 0.0003, "step": 36586 }, { "epoch": 0.71, "learning_rate": 3.8168390815752195e-05, "loss": 0.0001, "step": 36588 }, { "epoch": 0.71, "learning_rate": 3.816774385549496e-05, "loss": 0.0001, "step": 36590 }, { "epoch": 0.71, "learning_rate": 3.8167096895237726e-05, "loss": 0.0064, "step": 36592 }, { "epoch": 0.71, "learning_rate": 3.8166449934980495e-05, "loss": 0.0, "step": 36594 }, { "epoch": 0.71, "learning_rate": 3.8165802974723264e-05, "loss": 0.0042, "step": 36596 }, { "epoch": 0.71, "learning_rate": 3.816515601446603e-05, "loss": 0.0001, "step": 36598 }, { "epoch": 0.71, "learning_rate": 3.81645090542088e-05, "loss": 0.0, "step": 36600 }, { "epoch": 0.71, "learning_rate": 3.816386209395157e-05, "loss": 0.0, "step": 36602 }, { "epoch": 0.71, "learning_rate": 3.816321513369434e-05, "loss": 0.0002, "step": 36604 }, { "epoch": 0.71, "learning_rate": 3.816256817343711e-05, "loss": 0.0093, "step": 36606 }, { "epoch": 0.71, "learning_rate": 3.816192121317987e-05, "loss": 0.0048, "step": 36608 }, { "epoch": 0.71, "learning_rate": 3.816127425292265e-05, "loss": 0.0, "step": 36610 }, { "epoch": 0.71, "learning_rate": 3.816062729266541e-05, "loss": 0.0, "step": 36612 }, { "epoch": 0.71, "learning_rate": 3.815998033240818e-05, "loss": 0.0002, "step": 36614 }, { "epoch": 0.71, "learning_rate": 3.815933337215095e-05, "loss": 0.0, "step": 36616 }, { "epoch": 0.71, "learning_rate": 3.815868641189372e-05, "loss": 0.0, "step": 36618 }, { "epoch": 0.71, "learning_rate": 3.815803945163649e-05, "loss": 0.0, "step": 36620 }, { "epoch": 0.71, "learning_rate": 3.8157392491379256e-05, "loss": 0.0, "step": 36622 }, { "epoch": 0.71, "learning_rate": 3.8156745531122025e-05, "loss": 0.0, "step": 36624 }, { "epoch": 0.71, "learning_rate": 3.8156098570864794e-05, "loss": 0.0, "step": 36626 }, { "epoch": 0.71, "learning_rate": 3.815545161060756e-05, "loss": 0.0007, "step": 36628 }, { "epoch": 0.71, "learning_rate": 3.815480465035033e-05, "loss": 0.0, "step": 36630 }, { "epoch": 0.71, "learning_rate": 3.81541576900931e-05, "loss": 0.0, "step": 36632 }, { "epoch": 0.71, "learning_rate": 3.815351072983587e-05, "loss": 0.0, "step": 36634 }, { "epoch": 0.71, "learning_rate": 3.815286376957863e-05, "loss": 0.0013, "step": 36636 }, { "epoch": 0.71, "learning_rate": 3.815221680932141e-05, "loss": 0.0, "step": 36638 }, { "epoch": 0.71, "learning_rate": 3.815156984906417e-05, "loss": 0.0002, "step": 36640 }, { "epoch": 0.71, "learning_rate": 3.8150922888806946e-05, "loss": 0.0104, "step": 36642 }, { "epoch": 0.71, "learning_rate": 3.815027592854971e-05, "loss": 0.0, "step": 36644 }, { "epoch": 0.71, "learning_rate": 3.814962896829248e-05, "loss": 0.0003, "step": 36646 }, { "epoch": 0.71, "learning_rate": 3.814898200803525e-05, "loss": 0.0014, "step": 36648 }, { "epoch": 0.71, "learning_rate": 3.8148335047778016e-05, "loss": 0.0037, "step": 36650 }, { "epoch": 0.71, "learning_rate": 3.8147688087520785e-05, "loss": 0.0, "step": 36652 }, { "epoch": 0.71, "learning_rate": 3.8147041127263554e-05, "loss": 0.0026, "step": 36654 }, { "epoch": 0.71, "learning_rate": 3.814639416700632e-05, "loss": 0.0, "step": 36656 }, { "epoch": 0.71, "learning_rate": 3.814574720674909e-05, "loss": 0.0007, "step": 36658 }, { "epoch": 0.71, "learning_rate": 3.814510024649186e-05, "loss": 0.0, "step": 36660 }, { "epoch": 0.71, "learning_rate": 3.814445328623463e-05, "loss": 0.0072, "step": 36662 }, { "epoch": 0.71, "learning_rate": 3.81438063259774e-05, "loss": 0.0001, "step": 36664 }, { "epoch": 0.71, "learning_rate": 3.814315936572017e-05, "loss": 0.0, "step": 36666 }, { "epoch": 0.71, "learning_rate": 3.814251240546293e-05, "loss": 0.0009, "step": 36668 }, { "epoch": 0.71, "learning_rate": 3.814186544520571e-05, "loss": 0.0, "step": 36670 }, { "epoch": 0.71, "learning_rate": 3.814121848494847e-05, "loss": 0.0039, "step": 36672 }, { "epoch": 0.71, "learning_rate": 3.814057152469124e-05, "loss": 0.0001, "step": 36674 }, { "epoch": 0.71, "learning_rate": 3.813992456443401e-05, "loss": 0.001, "step": 36676 }, { "epoch": 0.71, "learning_rate": 3.8139277604176776e-05, "loss": 0.0, "step": 36678 }, { "epoch": 0.71, "learning_rate": 3.8138630643919545e-05, "loss": 0.0, "step": 36680 }, { "epoch": 0.71, "learning_rate": 3.8137983683662315e-05, "loss": 0.0, "step": 36682 }, { "epoch": 0.71, "learning_rate": 3.8137336723405084e-05, "loss": 0.0, "step": 36684 }, { "epoch": 0.71, "learning_rate": 3.813668976314785e-05, "loss": 0.0004, "step": 36686 }, { "epoch": 0.71, "learning_rate": 3.813604280289062e-05, "loss": 0.0004, "step": 36688 }, { "epoch": 0.71, "learning_rate": 3.8135395842633384e-05, "loss": 0.0, "step": 36690 }, { "epoch": 0.71, "learning_rate": 3.813474888237616e-05, "loss": 0.0, "step": 36692 }, { "epoch": 0.71, "learning_rate": 3.813410192211892e-05, "loss": 0.0029, "step": 36694 }, { "epoch": 0.71, "learning_rate": 3.813345496186169e-05, "loss": 0.0012, "step": 36696 }, { "epoch": 0.71, "learning_rate": 3.813280800160447e-05, "loss": 0.0001, "step": 36698 }, { "epoch": 0.71, "learning_rate": 3.813216104134723e-05, "loss": 0.0, "step": 36700 }, { "epoch": 0.71, "learning_rate": 3.8131514081090005e-05, "loss": 0.0, "step": 36702 }, { "epoch": 0.71, "learning_rate": 3.813086712083277e-05, "loss": 0.0018, "step": 36704 }, { "epoch": 0.71, "learning_rate": 3.813022016057554e-05, "loss": 0.0, "step": 36706 }, { "epoch": 0.71, "learning_rate": 3.8129573200318306e-05, "loss": 0.0002, "step": 36708 }, { "epoch": 0.71, "learning_rate": 3.8128926240061075e-05, "loss": 0.0, "step": 36710 }, { "epoch": 0.71, "learning_rate": 3.8128279279803844e-05, "loss": 0.0, "step": 36712 }, { "epoch": 0.71, "learning_rate": 3.812763231954661e-05, "loss": 0.0013, "step": 36714 }, { "epoch": 0.71, "learning_rate": 3.812698535928938e-05, "loss": 0.0, "step": 36716 }, { "epoch": 0.71, "learning_rate": 3.8126338399032144e-05, "loss": 0.0001, "step": 36718 }, { "epoch": 0.71, "learning_rate": 3.812569143877492e-05, "loss": 0.0001, "step": 36720 }, { "epoch": 0.71, "learning_rate": 3.812504447851768e-05, "loss": 0.0001, "step": 36722 }, { "epoch": 0.71, "learning_rate": 3.812439751826046e-05, "loss": 0.0015, "step": 36724 }, { "epoch": 0.71, "learning_rate": 3.812375055800322e-05, "loss": 0.0088, "step": 36726 }, { "epoch": 0.71, "learning_rate": 3.812310359774599e-05, "loss": 0.0081, "step": 36728 }, { "epoch": 0.71, "learning_rate": 3.8122456637488766e-05, "loss": 0.0, "step": 36730 }, { "epoch": 0.71, "learning_rate": 3.812180967723153e-05, "loss": 0.0011, "step": 36732 }, { "epoch": 0.71, "learning_rate": 3.81211627169743e-05, "loss": 0.0001, "step": 36734 }, { "epoch": 0.71, "learning_rate": 3.8120515756717066e-05, "loss": 0.0005, "step": 36736 }, { "epoch": 0.71, "learning_rate": 3.8119868796459835e-05, "loss": 0.0, "step": 36738 }, { "epoch": 0.71, "learning_rate": 3.8119221836202604e-05, "loss": 0.0001, "step": 36740 }, { "epoch": 0.71, "learning_rate": 3.8118574875945373e-05, "loss": 0.0001, "step": 36742 }, { "epoch": 0.71, "learning_rate": 3.811792791568814e-05, "loss": 0.001, "step": 36744 }, { "epoch": 0.71, "learning_rate": 3.811728095543091e-05, "loss": 0.0001, "step": 36746 }, { "epoch": 0.71, "learning_rate": 3.811663399517368e-05, "loss": 0.0009, "step": 36748 }, { "epoch": 0.71, "learning_rate": 3.811598703491644e-05, "loss": 0.0002, "step": 36750 }, { "epoch": 0.71, "learning_rate": 3.811534007465922e-05, "loss": 0.0165, "step": 36752 }, { "epoch": 0.71, "learning_rate": 3.811469311440198e-05, "loss": 0.0001, "step": 36754 }, { "epoch": 0.71, "learning_rate": 3.811404615414475e-05, "loss": 0.0003, "step": 36756 }, { "epoch": 0.71, "learning_rate": 3.811339919388752e-05, "loss": 0.0262, "step": 36758 }, { "epoch": 0.71, "learning_rate": 3.811275223363029e-05, "loss": 0.0006, "step": 36760 }, { "epoch": 0.71, "learning_rate": 3.8112105273373064e-05, "loss": 0.0012, "step": 36762 }, { "epoch": 0.71, "learning_rate": 3.811145831311583e-05, "loss": 0.0001, "step": 36764 }, { "epoch": 0.71, "learning_rate": 3.8110811352858596e-05, "loss": 0.0003, "step": 36766 }, { "epoch": 0.71, "learning_rate": 3.8110164392601365e-05, "loss": 0.0031, "step": 36768 }, { "epoch": 0.71, "learning_rate": 3.8109517432344134e-05, "loss": 0.0, "step": 36770 }, { "epoch": 0.71, "learning_rate": 3.8108870472086896e-05, "loss": 0.0032, "step": 36772 }, { "epoch": 0.71, "learning_rate": 3.810822351182967e-05, "loss": 0.0004, "step": 36774 }, { "epoch": 0.71, "learning_rate": 3.810757655157244e-05, "loss": 0.0036, "step": 36776 }, { "epoch": 0.71, "learning_rate": 3.8106929591315203e-05, "loss": 0.0086, "step": 36778 }, { "epoch": 0.71, "learning_rate": 3.810628263105798e-05, "loss": 0.007, "step": 36780 }, { "epoch": 0.71, "learning_rate": 3.810563567080074e-05, "loss": 0.0013, "step": 36782 }, { "epoch": 0.71, "learning_rate": 3.810498871054352e-05, "loss": 0.0, "step": 36784 }, { "epoch": 0.71, "learning_rate": 3.810434175028628e-05, "loss": 0.0, "step": 36786 }, { "epoch": 0.71, "learning_rate": 3.810369479002905e-05, "loss": 0.0, "step": 36788 }, { "epoch": 0.71, "learning_rate": 3.810304782977182e-05, "loss": 0.0, "step": 36790 }, { "epoch": 0.71, "learning_rate": 3.810240086951459e-05, "loss": 0.0003, "step": 36792 }, { "epoch": 0.71, "learning_rate": 3.8101753909257356e-05, "loss": 0.0003, "step": 36794 }, { "epoch": 0.71, "learning_rate": 3.8101106949000125e-05, "loss": 0.0122, "step": 36796 }, { "epoch": 0.71, "learning_rate": 3.8100459988742894e-05, "loss": 0.0001, "step": 36798 }, { "epoch": 0.71, "learning_rate": 3.809981302848566e-05, "loss": 0.0019, "step": 36800 }, { "epoch": 0.71, "learning_rate": 3.809916606822843e-05, "loss": 0.0, "step": 36802 }, { "epoch": 0.71, "learning_rate": 3.8098519107971195e-05, "loss": 0.0, "step": 36804 }, { "epoch": 0.71, "learning_rate": 3.809787214771397e-05, "loss": 0.0246, "step": 36806 }, { "epoch": 0.71, "learning_rate": 3.809722518745674e-05, "loss": 0.0, "step": 36808 }, { "epoch": 0.71, "learning_rate": 3.80965782271995e-05, "loss": 0.0074, "step": 36810 }, { "epoch": 0.71, "learning_rate": 3.809593126694228e-05, "loss": 0.0008, "step": 36812 }, { "epoch": 0.71, "learning_rate": 3.809528430668504e-05, "loss": 0.0026, "step": 36814 }, { "epoch": 0.71, "learning_rate": 3.809463734642781e-05, "loss": 0.0001, "step": 36816 }, { "epoch": 0.71, "learning_rate": 3.809399038617058e-05, "loss": 0.0054, "step": 36818 }, { "epoch": 0.71, "learning_rate": 3.809334342591335e-05, "loss": 0.0001, "step": 36820 }, { "epoch": 0.71, "learning_rate": 3.8092696465656117e-05, "loss": 0.0, "step": 36822 }, { "epoch": 0.71, "learning_rate": 3.8092049505398886e-05, "loss": 0.008, "step": 36824 }, { "epoch": 0.71, "learning_rate": 3.8091402545141655e-05, "loss": 0.0006, "step": 36826 }, { "epoch": 0.71, "learning_rate": 3.8090755584884424e-05, "loss": 0.0002, "step": 36828 }, { "epoch": 0.71, "learning_rate": 3.809010862462719e-05, "loss": 0.0117, "step": 36830 }, { "epoch": 0.71, "learning_rate": 3.8089461664369955e-05, "loss": 0.0002, "step": 36832 }, { "epoch": 0.71, "learning_rate": 3.808881470411273e-05, "loss": 0.0047, "step": 36834 }, { "epoch": 0.71, "learning_rate": 3.808816774385549e-05, "loss": 0.004, "step": 36836 }, { "epoch": 0.71, "learning_rate": 3.808752078359826e-05, "loss": 0.0044, "step": 36838 }, { "epoch": 0.72, "learning_rate": 3.808687382334104e-05, "loss": 0.0002, "step": 36840 }, { "epoch": 0.72, "learning_rate": 3.80862268630838e-05, "loss": 0.0075, "step": 36842 }, { "epoch": 0.72, "learning_rate": 3.8085579902826576e-05, "loss": 0.0, "step": 36844 }, { "epoch": 0.72, "learning_rate": 3.808493294256934e-05, "loss": 0.0, "step": 36846 }, { "epoch": 0.72, "learning_rate": 3.808428598231211e-05, "loss": 0.0001, "step": 36848 }, { "epoch": 0.72, "learning_rate": 3.808363902205488e-05, "loss": 0.0001, "step": 36850 }, { "epoch": 0.72, "learning_rate": 3.8082992061797646e-05, "loss": 0.0, "step": 36852 }, { "epoch": 0.72, "learning_rate": 3.8082345101540415e-05, "loss": 0.0101, "step": 36854 }, { "epoch": 0.72, "learning_rate": 3.8081698141283184e-05, "loss": 0.0002, "step": 36856 }, { "epoch": 0.72, "learning_rate": 3.808105118102595e-05, "loss": 0.0002, "step": 36858 }, { "epoch": 0.72, "learning_rate": 3.8080404220768716e-05, "loss": 0.0001, "step": 36860 }, { "epoch": 0.72, "learning_rate": 3.807975726051149e-05, "loss": 0.0001, "step": 36862 }, { "epoch": 0.72, "learning_rate": 3.8079110300254254e-05, "loss": 0.0, "step": 36864 }, { "epoch": 0.72, "learning_rate": 3.807846333999703e-05, "loss": 0.0048, "step": 36866 }, { "epoch": 0.72, "learning_rate": 3.807781637973979e-05, "loss": 0.0007, "step": 36868 }, { "epoch": 0.72, "learning_rate": 3.807716941948256e-05, "loss": 0.0, "step": 36870 }, { "epoch": 0.72, "learning_rate": 3.807652245922533e-05, "loss": 0.0003, "step": 36872 }, { "epoch": 0.72, "learning_rate": 3.80758754989681e-05, "loss": 0.0009, "step": 36874 }, { "epoch": 0.72, "learning_rate": 3.807522853871087e-05, "loss": 0.0104, "step": 36876 }, { "epoch": 0.72, "learning_rate": 3.807458157845364e-05, "loss": 0.0007, "step": 36878 }, { "epoch": 0.72, "learning_rate": 3.8073934618196406e-05, "loss": 0.0003, "step": 36880 }, { "epoch": 0.72, "learning_rate": 3.8073287657939176e-05, "loss": 0.002, "step": 36882 }, { "epoch": 0.72, "learning_rate": 3.8072640697681945e-05, "loss": 0.0001, "step": 36884 }, { "epoch": 0.72, "learning_rate": 3.8071993737424714e-05, "loss": 0.0, "step": 36886 }, { "epoch": 0.72, "learning_rate": 3.807134677716748e-05, "loss": 0.0024, "step": 36888 }, { "epoch": 0.72, "learning_rate": 3.807069981691025e-05, "loss": 0.0, "step": 36890 }, { "epoch": 0.72, "learning_rate": 3.8070052856653014e-05, "loss": 0.0, "step": 36892 }, { "epoch": 0.72, "learning_rate": 3.806940589639579e-05, "loss": 0.0, "step": 36894 }, { "epoch": 0.72, "learning_rate": 3.806875893613855e-05, "loss": 0.0, "step": 36896 }, { "epoch": 0.72, "learning_rate": 3.806811197588132e-05, "loss": 0.0002, "step": 36898 }, { "epoch": 0.72, "learning_rate": 3.806746501562409e-05, "loss": 0.0018, "step": 36900 }, { "epoch": 0.72, "learning_rate": 3.806681805536686e-05, "loss": 0.0005, "step": 36902 }, { "epoch": 0.72, "learning_rate": 3.806617109510963e-05, "loss": 0.0169, "step": 36904 }, { "epoch": 0.72, "learning_rate": 3.80655241348524e-05, "loss": 0.0, "step": 36906 }, { "epoch": 0.72, "learning_rate": 3.806487717459517e-05, "loss": 0.0001, "step": 36908 }, { "epoch": 0.72, "learning_rate": 3.8064230214337936e-05, "loss": 0.0049, "step": 36910 }, { "epoch": 0.72, "learning_rate": 3.8063583254080705e-05, "loss": 0.0014, "step": 36912 }, { "epoch": 0.72, "learning_rate": 3.806293629382347e-05, "loss": 0.0003, "step": 36914 }, { "epoch": 0.72, "learning_rate": 3.806228933356624e-05, "loss": 0.0, "step": 36916 }, { "epoch": 0.72, "learning_rate": 3.8061642373309005e-05, "loss": 0.0, "step": 36918 }, { "epoch": 0.72, "learning_rate": 3.8060995413051775e-05, "loss": 0.0005, "step": 36920 }, { "epoch": 0.72, "learning_rate": 3.806034845279455e-05, "loss": 0.0, "step": 36922 }, { "epoch": 0.72, "learning_rate": 3.805970149253731e-05, "loss": 0.0016, "step": 36924 }, { "epoch": 0.72, "learning_rate": 3.805905453228009e-05, "loss": 0.0, "step": 36926 }, { "epoch": 0.72, "learning_rate": 3.805840757202285e-05, "loss": 0.0005, "step": 36928 }, { "epoch": 0.72, "learning_rate": 3.805776061176562e-05, "loss": 0.0, "step": 36930 }, { "epoch": 0.72, "learning_rate": 3.805711365150839e-05, "loss": 0.0, "step": 36932 }, { "epoch": 0.72, "learning_rate": 3.805646669125116e-05, "loss": 0.0, "step": 36934 }, { "epoch": 0.72, "learning_rate": 3.805581973099393e-05, "loss": 0.0001, "step": 36936 }, { "epoch": 0.72, "learning_rate": 3.8055172770736696e-05, "loss": 0.0, "step": 36938 }, { "epoch": 0.72, "learning_rate": 3.8054525810479465e-05, "loss": 0.0007, "step": 36940 }, { "epoch": 0.72, "learning_rate": 3.8053878850222234e-05, "loss": 0.0053, "step": 36942 }, { "epoch": 0.72, "learning_rate": 3.8053231889965004e-05, "loss": 0.0, "step": 36944 }, { "epoch": 0.72, "learning_rate": 3.8052584929707766e-05, "loss": 0.0001, "step": 36946 }, { "epoch": 0.72, "learning_rate": 3.805193796945054e-05, "loss": 0.0, "step": 36948 }, { "epoch": 0.72, "learning_rate": 3.8051291009193304e-05, "loss": 0.0084, "step": 36950 }, { "epoch": 0.72, "learning_rate": 3.805064404893607e-05, "loss": 0.0017, "step": 36952 }, { "epoch": 0.72, "learning_rate": 3.804999708867885e-05, "loss": 0.0, "step": 36954 }, { "epoch": 0.72, "learning_rate": 3.804935012842161e-05, "loss": 0.0, "step": 36956 }, { "epoch": 0.72, "learning_rate": 3.804870316816438e-05, "loss": 0.0118, "step": 36958 }, { "epoch": 0.72, "learning_rate": 3.804805620790715e-05, "loss": 0.0, "step": 36960 }, { "epoch": 0.72, "learning_rate": 3.804740924764992e-05, "loss": 0.0002, "step": 36962 }, { "epoch": 0.72, "learning_rate": 3.804676228739269e-05, "loss": 0.0004, "step": 36964 }, { "epoch": 0.72, "learning_rate": 3.804611532713546e-05, "loss": 0.0159, "step": 36966 }, { "epoch": 0.72, "learning_rate": 3.8045468366878226e-05, "loss": 0.0069, "step": 36968 }, { "epoch": 0.72, "learning_rate": 3.8044821406620995e-05, "loss": 0.0006, "step": 36970 }, { "epoch": 0.72, "learning_rate": 3.8044174446363764e-05, "loss": 0.0002, "step": 36972 }, { "epoch": 0.72, "learning_rate": 3.8043527486106526e-05, "loss": 0.0001, "step": 36974 }, { "epoch": 0.72, "learning_rate": 3.80428805258493e-05, "loss": 0.0007, "step": 36976 }, { "epoch": 0.72, "learning_rate": 3.8042233565592064e-05, "loss": 0.0089, "step": 36978 }, { "epoch": 0.72, "learning_rate": 3.8041586605334834e-05, "loss": 0.0068, "step": 36980 }, { "epoch": 0.72, "learning_rate": 3.80409396450776e-05, "loss": 0.0001, "step": 36982 }, { "epoch": 0.72, "learning_rate": 3.804029268482037e-05, "loss": 0.001, "step": 36984 }, { "epoch": 0.72, "learning_rate": 3.803964572456315e-05, "loss": 0.0013, "step": 36986 }, { "epoch": 0.72, "learning_rate": 3.803899876430591e-05, "loss": 0.0001, "step": 36988 }, { "epoch": 0.72, "learning_rate": 3.803835180404868e-05, "loss": 0.0086, "step": 36990 }, { "epoch": 0.72, "learning_rate": 3.803770484379145e-05, "loss": 0.0092, "step": 36992 }, { "epoch": 0.72, "learning_rate": 3.803705788353422e-05, "loss": 0.0, "step": 36994 }, { "epoch": 0.72, "learning_rate": 3.803641092327698e-05, "loss": 0.0, "step": 36996 }, { "epoch": 0.72, "learning_rate": 3.8035763963019755e-05, "loss": 0.0, "step": 36998 }, { "epoch": 0.72, "learning_rate": 3.8035117002762524e-05, "loss": 0.0, "step": 37000 }, { "epoch": 0.72, "learning_rate": 3.803447004250529e-05, "loss": 0.0, "step": 37002 }, { "epoch": 0.72, "learning_rate": 3.803382308224806e-05, "loss": 0.0, "step": 37004 }, { "epoch": 0.72, "learning_rate": 3.8033176121990825e-05, "loss": 0.0, "step": 37006 }, { "epoch": 0.72, "learning_rate": 3.80325291617336e-05, "loss": 0.0, "step": 37008 }, { "epoch": 0.72, "learning_rate": 3.8032205681604985e-05, "loss": 0.1417, "step": 37010 }, { "epoch": 0.72, "learning_rate": 3.803155872134775e-05, "loss": 0.0, "step": 37012 }, { "epoch": 0.72, "learning_rate": 3.8030911761090523e-05, "loss": 0.0002, "step": 37014 }, { "epoch": 0.72, "learning_rate": 3.8030264800833286e-05, "loss": 0.004, "step": 37016 }, { "epoch": 0.72, "learning_rate": 3.8029617840576055e-05, "loss": 0.0007, "step": 37018 }, { "epoch": 0.72, "learning_rate": 3.8028970880318824e-05, "loss": 0.0044, "step": 37020 }, { "epoch": 0.72, "learning_rate": 3.802832392006159e-05, "loss": 0.0002, "step": 37022 }, { "epoch": 0.72, "learning_rate": 3.802767695980436e-05, "loss": 0.0087, "step": 37024 }, { "epoch": 0.72, "learning_rate": 3.802702999954713e-05, "loss": 0.0001, "step": 37026 }, { "epoch": 0.72, "learning_rate": 3.80263830392899e-05, "loss": 0.0001, "step": 37028 }, { "epoch": 0.72, "learning_rate": 3.802573607903266e-05, "loss": 0.0, "step": 37030 }, { "epoch": 0.72, "learning_rate": 3.802508911877544e-05, "loss": 0.0001, "step": 37032 }, { "epoch": 0.72, "learning_rate": 3.80244421585182e-05, "loss": 0.0006, "step": 37034 }, { "epoch": 0.72, "learning_rate": 3.8023795198260977e-05, "loss": 0.0341, "step": 37036 }, { "epoch": 0.72, "learning_rate": 3.802314823800374e-05, "loss": 0.0004, "step": 37038 }, { "epoch": 0.72, "learning_rate": 3.802250127774651e-05, "loss": 0.0004, "step": 37040 }, { "epoch": 0.72, "learning_rate": 3.8021854317489284e-05, "loss": 0.0004, "step": 37042 }, { "epoch": 0.72, "learning_rate": 3.8021207357232046e-05, "loss": 0.023, "step": 37044 }, { "epoch": 0.72, "learning_rate": 3.8020560396974815e-05, "loss": 0.0, "step": 37046 }, { "epoch": 0.72, "learning_rate": 3.8019913436717584e-05, "loss": 0.0, "step": 37048 }, { "epoch": 0.72, "learning_rate": 3.8019266476460353e-05, "loss": 0.0002, "step": 37050 }, { "epoch": 0.72, "learning_rate": 3.8018619516203116e-05, "loss": 0.0022, "step": 37052 }, { "epoch": 0.72, "learning_rate": 3.801797255594589e-05, "loss": 0.0003, "step": 37054 }, { "epoch": 0.72, "learning_rate": 3.801732559568866e-05, "loss": 0.0, "step": 37056 }, { "epoch": 0.72, "learning_rate": 3.801667863543143e-05, "loss": 0.0007, "step": 37058 }, { "epoch": 0.72, "learning_rate": 3.80160316751742e-05, "loss": 0.0003, "step": 37060 }, { "epoch": 0.72, "learning_rate": 3.801538471491696e-05, "loss": 0.0, "step": 37062 }, { "epoch": 0.72, "learning_rate": 3.801473775465974e-05, "loss": 0.0, "step": 37064 }, { "epoch": 0.72, "learning_rate": 3.80140907944025e-05, "loss": 0.0, "step": 37066 }, { "epoch": 0.72, "learning_rate": 3.801344383414527e-05, "loss": 0.0, "step": 37068 }, { "epoch": 0.72, "learning_rate": 3.801279687388804e-05, "loss": 0.0, "step": 37070 }, { "epoch": 0.72, "learning_rate": 3.8012149913630807e-05, "loss": 0.0011, "step": 37072 }, { "epoch": 0.72, "learning_rate": 3.8011502953373576e-05, "loss": 0.0001, "step": 37074 }, { "epoch": 0.72, "learning_rate": 3.8010855993116345e-05, "loss": 0.0, "step": 37076 }, { "epoch": 0.72, "learning_rate": 3.8010209032859114e-05, "loss": 0.0, "step": 37078 }, { "epoch": 0.72, "learning_rate": 3.800956207260188e-05, "loss": 0.0, "step": 37080 }, { "epoch": 0.72, "learning_rate": 3.800891511234465e-05, "loss": 0.0, "step": 37082 }, { "epoch": 0.72, "learning_rate": 3.8008268152087414e-05, "loss": 0.0, "step": 37084 }, { "epoch": 0.72, "learning_rate": 3.800762119183019e-05, "loss": 0.0001, "step": 37086 }, { "epoch": 0.72, "learning_rate": 3.800697423157296e-05, "loss": 0.0004, "step": 37088 }, { "epoch": 0.72, "learning_rate": 3.800632727131572e-05, "loss": 0.0109, "step": 37090 }, { "epoch": 0.72, "learning_rate": 3.80056803110585e-05, "loss": 0.0, "step": 37092 }, { "epoch": 0.72, "learning_rate": 3.800503335080126e-05, "loss": 0.0, "step": 37094 }, { "epoch": 0.72, "learning_rate": 3.8004386390544036e-05, "loss": 0.0002, "step": 37096 }, { "epoch": 0.72, "learning_rate": 3.80037394302868e-05, "loss": 0.0, "step": 37098 }, { "epoch": 0.72, "learning_rate": 3.800309247002957e-05, "loss": 0.0003, "step": 37100 }, { "epoch": 0.72, "learning_rate": 3.8002445509772336e-05, "loss": 0.0002, "step": 37102 }, { "epoch": 0.72, "learning_rate": 3.8001798549515105e-05, "loss": 0.0001, "step": 37104 }, { "epoch": 0.72, "learning_rate": 3.8001151589257874e-05, "loss": 0.0039, "step": 37106 }, { "epoch": 0.72, "learning_rate": 3.800050462900064e-05, "loss": 0.0, "step": 37108 }, { "epoch": 0.72, "learning_rate": 3.799985766874341e-05, "loss": 0.0, "step": 37110 }, { "epoch": 0.72, "learning_rate": 3.7999210708486175e-05, "loss": 0.0, "step": 37112 }, { "epoch": 0.72, "learning_rate": 3.799856374822895e-05, "loss": 0.0, "step": 37114 }, { "epoch": 0.72, "learning_rate": 3.799791678797171e-05, "loss": 0.0, "step": 37116 }, { "epoch": 0.72, "learning_rate": 3.799726982771449e-05, "loss": 0.0004, "step": 37118 }, { "epoch": 0.72, "learning_rate": 3.799662286745726e-05, "loss": 0.0, "step": 37120 }, { "epoch": 0.72, "learning_rate": 3.799597590720002e-05, "loss": 0.0067, "step": 37122 }, { "epoch": 0.72, "learning_rate": 3.7995328946942796e-05, "loss": 0.0001, "step": 37124 }, { "epoch": 0.72, "learning_rate": 3.799468198668556e-05, "loss": 0.0, "step": 37126 }, { "epoch": 0.72, "learning_rate": 3.799403502642833e-05, "loss": 0.0109, "step": 37128 }, { "epoch": 0.72, "learning_rate": 3.7993388066171096e-05, "loss": 0.0, "step": 37130 }, { "epoch": 0.72, "learning_rate": 3.7992741105913866e-05, "loss": 0.0001, "step": 37132 }, { "epoch": 0.72, "learning_rate": 3.7992094145656635e-05, "loss": 0.0, "step": 37134 }, { "epoch": 0.72, "learning_rate": 3.7991447185399404e-05, "loss": 0.0002, "step": 37136 }, { "epoch": 0.72, "learning_rate": 3.799080022514217e-05, "loss": 0.0, "step": 37138 }, { "epoch": 0.72, "learning_rate": 3.799015326488494e-05, "loss": 0.0054, "step": 37140 }, { "epoch": 0.72, "learning_rate": 3.798950630462771e-05, "loss": 0.0063, "step": 37142 }, { "epoch": 0.72, "learning_rate": 3.798885934437047e-05, "loss": 0.0076, "step": 37144 }, { "epoch": 0.72, "learning_rate": 3.798821238411325e-05, "loss": 0.0, "step": 37146 }, { "epoch": 0.72, "learning_rate": 3.798756542385601e-05, "loss": 0.0001, "step": 37148 }, { "epoch": 0.72, "learning_rate": 3.798691846359878e-05, "loss": 0.0001, "step": 37150 }, { "epoch": 0.72, "learning_rate": 3.798627150334155e-05, "loss": 0.0, "step": 37152 }, { "epoch": 0.72, "learning_rate": 3.798562454308432e-05, "loss": 0.0001, "step": 37154 }, { "epoch": 0.72, "learning_rate": 3.7984977582827095e-05, "loss": 0.0, "step": 37156 }, { "epoch": 0.72, "learning_rate": 3.798433062256986e-05, "loss": 0.0001, "step": 37158 }, { "epoch": 0.72, "learning_rate": 3.7983683662312626e-05, "loss": 0.0, "step": 37160 }, { "epoch": 0.72, "learning_rate": 3.7983036702055395e-05, "loss": 0.0, "step": 37162 }, { "epoch": 0.72, "learning_rate": 3.7982389741798164e-05, "loss": 0.0011, "step": 37164 }, { "epoch": 0.72, "learning_rate": 3.798174278154093e-05, "loss": 0.0003, "step": 37166 }, { "epoch": 0.72, "learning_rate": 3.79810958212837e-05, "loss": 0.0001, "step": 37168 }, { "epoch": 0.72, "learning_rate": 3.798044886102647e-05, "loss": 0.036, "step": 37170 }, { "epoch": 0.72, "learning_rate": 3.7979801900769234e-05, "loss": 0.0, "step": 37172 }, { "epoch": 0.72, "learning_rate": 3.797915494051201e-05, "loss": 0.0, "step": 37174 }, { "epoch": 0.72, "learning_rate": 3.797850798025477e-05, "loss": 0.0, "step": 37176 }, { "epoch": 0.72, "learning_rate": 3.797786101999755e-05, "loss": 0.0, "step": 37178 }, { "epoch": 0.72, "learning_rate": 3.797721405974031e-05, "loss": 0.0, "step": 37180 }, { "epoch": 0.72, "learning_rate": 3.797656709948308e-05, "loss": 0.0, "step": 37182 }, { "epoch": 0.72, "learning_rate": 3.797592013922585e-05, "loss": 0.0001, "step": 37184 }, { "epoch": 0.72, "learning_rate": 3.797527317896862e-05, "loss": 0.0028, "step": 37186 }, { "epoch": 0.72, "learning_rate": 3.7974626218711386e-05, "loss": 0.0, "step": 37188 }, { "epoch": 0.72, "learning_rate": 3.7973979258454155e-05, "loss": 0.0, "step": 37190 }, { "epoch": 0.72, "learning_rate": 3.7973332298196925e-05, "loss": 0.0005, "step": 37192 }, { "epoch": 0.72, "learning_rate": 3.797268533793969e-05, "loss": 0.0, "step": 37194 }, { "epoch": 0.72, "learning_rate": 3.797203837768246e-05, "loss": 0.0001, "step": 37196 }, { "epoch": 0.72, "learning_rate": 3.7971391417425225e-05, "loss": 0.0, "step": 37198 }, { "epoch": 0.72, "learning_rate": 3.7970744457168e-05, "loss": 0.0, "step": 37200 }, { "epoch": 0.72, "learning_rate": 3.797009749691077e-05, "loss": 0.0082, "step": 37202 }, { "epoch": 0.72, "learning_rate": 3.796945053665353e-05, "loss": 0.0, "step": 37204 }, { "epoch": 0.72, "learning_rate": 3.796880357639631e-05, "loss": 0.0, "step": 37206 }, { "epoch": 0.72, "learning_rate": 3.796815661613907e-05, "loss": 0.001, "step": 37208 }, { "epoch": 0.72, "learning_rate": 3.796750965588184e-05, "loss": 0.0004, "step": 37210 }, { "epoch": 0.72, "learning_rate": 3.796686269562461e-05, "loss": 0.0084, "step": 37212 }, { "epoch": 0.72, "learning_rate": 3.796621573536738e-05, "loss": 0.0001, "step": 37214 }, { "epoch": 0.72, "learning_rate": 3.796556877511015e-05, "loss": 0.0011, "step": 37216 }, { "epoch": 0.72, "learning_rate": 3.7964921814852916e-05, "loss": 0.0022, "step": 37218 }, { "epoch": 0.72, "learning_rate": 3.7964274854595685e-05, "loss": 0.0, "step": 37220 }, { "epoch": 0.72, "learning_rate": 3.7963627894338454e-05, "loss": 0.0002, "step": 37222 }, { "epoch": 0.72, "learning_rate": 3.796298093408122e-05, "loss": 0.0014, "step": 37224 }, { "epoch": 0.72, "learning_rate": 3.7962333973823985e-05, "loss": 0.0024, "step": 37226 }, { "epoch": 0.72, "learning_rate": 3.796168701356676e-05, "loss": 0.0001, "step": 37228 }, { "epoch": 0.72, "learning_rate": 3.7961040053309524e-05, "loss": 0.0179, "step": 37230 }, { "epoch": 0.72, "learning_rate": 3.796039309305229e-05, "loss": 0.0, "step": 37232 }, { "epoch": 0.72, "learning_rate": 3.795974613279507e-05, "loss": 0.0, "step": 37234 }, { "epoch": 0.72, "learning_rate": 3.795909917253783e-05, "loss": 0.0, "step": 37236 }, { "epoch": 0.72, "learning_rate": 3.795845221228061e-05, "loss": 0.0002, "step": 37238 }, { "epoch": 0.72, "learning_rate": 3.795780525202337e-05, "loss": 0.0001, "step": 37240 }, { "epoch": 0.72, "learning_rate": 3.795715829176614e-05, "loss": 0.0001, "step": 37242 }, { "epoch": 0.72, "learning_rate": 3.795651133150891e-05, "loss": 0.0006, "step": 37244 }, { "epoch": 0.72, "learning_rate": 3.7955864371251676e-05, "loss": 0.0001, "step": 37246 }, { "epoch": 0.72, "learning_rate": 3.7955217410994445e-05, "loss": 0.0, "step": 37248 }, { "epoch": 0.72, "learning_rate": 3.7954570450737214e-05, "loss": 0.0001, "step": 37250 }, { "epoch": 0.72, "learning_rate": 3.7953923490479983e-05, "loss": 0.0, "step": 37252 }, { "epoch": 0.72, "learning_rate": 3.7953276530222746e-05, "loss": 0.0003, "step": 37254 }, { "epoch": 0.72, "learning_rate": 3.795262956996552e-05, "loss": 0.0011, "step": 37256 }, { "epoch": 0.72, "learning_rate": 3.7951982609708284e-05, "loss": 0.0031, "step": 37258 }, { "epoch": 0.72, "learning_rate": 3.795133564945106e-05, "loss": 0.0059, "step": 37260 }, { "epoch": 0.72, "learning_rate": 3.795068868919382e-05, "loss": 0.0001, "step": 37262 }, { "epoch": 0.72, "learning_rate": 3.795004172893659e-05, "loss": 0.0, "step": 37264 }, { "epoch": 0.72, "learning_rate": 3.794939476867937e-05, "loss": 0.001, "step": 37266 }, { "epoch": 0.72, "learning_rate": 3.794874780842213e-05, "loss": 0.0003, "step": 37268 }, { "epoch": 0.72, "learning_rate": 3.79481008481649e-05, "loss": 0.0014, "step": 37270 }, { "epoch": 0.72, "learning_rate": 3.794745388790767e-05, "loss": 0.0, "step": 37272 }, { "epoch": 0.72, "learning_rate": 3.794680692765044e-05, "loss": 0.0564, "step": 37274 }, { "epoch": 0.72, "learning_rate": 3.79461599673932e-05, "loss": 0.0001, "step": 37276 }, { "epoch": 0.72, "learning_rate": 3.7945513007135975e-05, "loss": 0.0001, "step": 37278 }, { "epoch": 0.72, "learning_rate": 3.7944866046878744e-05, "loss": 0.0, "step": 37280 }, { "epoch": 0.72, "learning_rate": 3.794421908662151e-05, "loss": 0.0004, "step": 37282 }, { "epoch": 0.72, "learning_rate": 3.794357212636428e-05, "loss": 0.0001, "step": 37284 }, { "epoch": 0.72, "learning_rate": 3.7942925166107044e-05, "loss": 0.0007, "step": 37286 }, { "epoch": 0.72, "learning_rate": 3.794227820584982e-05, "loss": 0.0, "step": 37288 }, { "epoch": 0.72, "learning_rate": 3.794163124559258e-05, "loss": 0.0, "step": 37290 }, { "epoch": 0.72, "learning_rate": 3.794098428533535e-05, "loss": 0.0, "step": 37292 }, { "epoch": 0.72, "learning_rate": 3.794033732507812e-05, "loss": 0.0, "step": 37294 }, { "epoch": 0.72, "learning_rate": 3.793969036482089e-05, "loss": 0.0, "step": 37296 }, { "epoch": 0.72, "learning_rate": 3.793904340456366e-05, "loss": 0.0002, "step": 37298 }, { "epoch": 0.72, "learning_rate": 3.793839644430643e-05, "loss": 0.0003, "step": 37300 }, { "epoch": 0.72, "learning_rate": 3.79377494840492e-05, "loss": 0.0005, "step": 37302 }, { "epoch": 0.72, "learning_rate": 3.7937102523791966e-05, "loss": 0.0027, "step": 37304 }, { "epoch": 0.72, "learning_rate": 3.7936455563534735e-05, "loss": 0.0, "step": 37306 }, { "epoch": 0.72, "learning_rate": 3.79358086032775e-05, "loss": 0.0027, "step": 37308 }, { "epoch": 0.72, "learning_rate": 3.793516164302027e-05, "loss": 0.0034, "step": 37310 }, { "epoch": 0.72, "learning_rate": 3.793451468276304e-05, "loss": 0.0007, "step": 37312 }, { "epoch": 0.72, "learning_rate": 3.7933867722505805e-05, "loss": 0.0, "step": 37314 }, { "epoch": 0.72, "learning_rate": 3.793322076224858e-05, "loss": 0.0004, "step": 37316 }, { "epoch": 0.72, "learning_rate": 3.793257380199134e-05, "loss": 0.0002, "step": 37318 }, { "epoch": 0.72, "learning_rate": 3.793192684173412e-05, "loss": 0.0131, "step": 37320 }, { "epoch": 0.72, "learning_rate": 3.793127988147688e-05, "loss": 0.0, "step": 37322 }, { "epoch": 0.72, "learning_rate": 3.793063292121965e-05, "loss": 0.0, "step": 37324 }, { "epoch": 0.72, "learning_rate": 3.792998596096242e-05, "loss": 0.0001, "step": 37326 }, { "epoch": 0.72, "learning_rate": 3.792933900070519e-05, "loss": 0.0254, "step": 37328 }, { "epoch": 0.72, "learning_rate": 3.792869204044796e-05, "loss": 0.0114, "step": 37330 }, { "epoch": 0.72, "learning_rate": 3.7928045080190727e-05, "loss": 0.0041, "step": 37332 }, { "epoch": 0.72, "learning_rate": 3.7927398119933496e-05, "loss": 0.0182, "step": 37334 }, { "epoch": 0.72, "learning_rate": 3.792675115967626e-05, "loss": 0.0, "step": 37336 }, { "epoch": 0.72, "learning_rate": 3.7926104199419034e-05, "loss": 0.0, "step": 37338 }, { "epoch": 0.72, "learning_rate": 3.7925457239161796e-05, "loss": 0.0, "step": 37340 }, { "epoch": 0.72, "learning_rate": 3.792481027890457e-05, "loss": 0.0009, "step": 37342 }, { "epoch": 0.72, "learning_rate": 3.792416331864734e-05, "loss": 0.0009, "step": 37344 }, { "epoch": 0.72, "learning_rate": 3.79235163583901e-05, "loss": 0.0012, "step": 37346 }, { "epoch": 0.72, "learning_rate": 3.792286939813288e-05, "loss": 0.0, "step": 37348 }, { "epoch": 0.72, "learning_rate": 3.792222243787564e-05, "loss": 0.0003, "step": 37350 }, { "epoch": 0.72, "learning_rate": 3.792157547761841e-05, "loss": 0.005, "step": 37352 }, { "epoch": 0.72, "learning_rate": 3.792092851736118e-05, "loss": 0.0001, "step": 37354 }, { "epoch": 0.73, "learning_rate": 3.792028155710395e-05, "loss": 0.0001, "step": 37356 }, { "epoch": 0.73, "learning_rate": 3.791963459684672e-05, "loss": 0.0, "step": 37358 }, { "epoch": 0.73, "learning_rate": 3.791898763658949e-05, "loss": 0.0, "step": 37360 }, { "epoch": 0.73, "learning_rate": 3.7918340676332256e-05, "loss": 0.0, "step": 37362 }, { "epoch": 0.73, "learning_rate": 3.7917693716075025e-05, "loss": 0.0001, "step": 37364 }, { "epoch": 0.73, "learning_rate": 3.7917046755817794e-05, "loss": 0.024, "step": 37366 }, { "epoch": 0.73, "learning_rate": 3.7916399795560556e-05, "loss": 0.0, "step": 37368 }, { "epoch": 0.73, "learning_rate": 3.791575283530333e-05, "loss": 0.0057, "step": 37370 }, { "epoch": 0.73, "learning_rate": 3.7915105875046095e-05, "loss": 0.0001, "step": 37372 }, { "epoch": 0.73, "learning_rate": 3.7914458914788864e-05, "loss": 0.0004, "step": 37374 }, { "epoch": 0.73, "learning_rate": 3.791381195453163e-05, "loss": 0.0, "step": 37376 }, { "epoch": 0.73, "learning_rate": 3.79131649942744e-05, "loss": 0.0001, "step": 37378 }, { "epoch": 0.73, "learning_rate": 3.791251803401718e-05, "loss": 0.0001, "step": 37380 }, { "epoch": 0.73, "learning_rate": 3.791187107375994e-05, "loss": 0.0006, "step": 37382 }, { "epoch": 0.73, "learning_rate": 3.791122411350271e-05, "loss": 0.0, "step": 37384 }, { "epoch": 0.73, "learning_rate": 3.791057715324548e-05, "loss": 0.0005, "step": 37386 }, { "epoch": 0.73, "learning_rate": 3.790993019298825e-05, "loss": 0.0, "step": 37388 }, { "epoch": 0.73, "learning_rate": 3.7909283232731016e-05, "loss": 0.0032, "step": 37390 }, { "epoch": 0.73, "learning_rate": 3.7908636272473786e-05, "loss": 0.0, "step": 37392 }, { "epoch": 0.73, "learning_rate": 3.7907989312216555e-05, "loss": 0.0001, "step": 37394 }, { "epoch": 0.73, "learning_rate": 3.790734235195932e-05, "loss": 0.0094, "step": 37396 }, { "epoch": 0.73, "learning_rate": 3.790669539170209e-05, "loss": 0.0003, "step": 37398 }, { "epoch": 0.73, "learning_rate": 3.7906048431444855e-05, "loss": 0.0001, "step": 37400 }, { "epoch": 0.73, "learning_rate": 3.790540147118763e-05, "loss": 0.0, "step": 37402 }, { "epoch": 0.73, "learning_rate": 3.790475451093039e-05, "loss": 0.0001, "step": 37404 }, { "epoch": 0.73, "learning_rate": 3.790410755067316e-05, "loss": 0.0047, "step": 37406 }, { "epoch": 0.73, "learning_rate": 3.790346059041593e-05, "loss": 0.0008, "step": 37408 }, { "epoch": 0.73, "learning_rate": 3.79028136301587e-05, "loss": 0.0, "step": 37410 }, { "epoch": 0.73, "learning_rate": 3.790216666990147e-05, "loss": 0.0049, "step": 37412 }, { "epoch": 0.73, "learning_rate": 3.790151970964424e-05, "loss": 0.0029, "step": 37414 }, { "epoch": 0.73, "learning_rate": 3.790087274938701e-05, "loss": 0.0, "step": 37416 }, { "epoch": 0.73, "learning_rate": 3.790022578912977e-05, "loss": 0.0, "step": 37418 }, { "epoch": 0.73, "learning_rate": 3.7899578828872546e-05, "loss": 0.0001, "step": 37420 }, { "epoch": 0.73, "learning_rate": 3.7898931868615315e-05, "loss": 0.0275, "step": 37422 }, { "epoch": 0.73, "learning_rate": 3.7898284908358084e-05, "loss": 0.0006, "step": 37424 }, { "epoch": 0.73, "learning_rate": 3.789763794810085e-05, "loss": 0.0, "step": 37426 }, { "epoch": 0.73, "learning_rate": 3.7896990987843615e-05, "loss": 0.0151, "step": 37428 }, { "epoch": 0.73, "learning_rate": 3.789634402758639e-05, "loss": 0.0, "step": 37430 }, { "epoch": 0.73, "learning_rate": 3.7895697067329154e-05, "loss": 0.0001, "step": 37432 }, { "epoch": 0.73, "learning_rate": 3.789505010707192e-05, "loss": 0.0005, "step": 37434 }, { "epoch": 0.73, "learning_rate": 3.789440314681469e-05, "loss": 0.0, "step": 37436 }, { "epoch": 0.73, "learning_rate": 3.789375618655746e-05, "loss": 0.0, "step": 37438 }, { "epoch": 0.73, "learning_rate": 3.789310922630023e-05, "loss": 0.0142, "step": 37440 }, { "epoch": 0.73, "learning_rate": 3.7892462266043e-05, "loss": 0.0049, "step": 37442 }, { "epoch": 0.73, "learning_rate": 3.789181530578577e-05, "loss": 0.0, "step": 37444 }, { "epoch": 0.73, "learning_rate": 3.789116834552854e-05, "loss": 0.0103, "step": 37446 }, { "epoch": 0.73, "learning_rate": 3.7890521385271306e-05, "loss": 0.0001, "step": 37448 }, { "epoch": 0.73, "learning_rate": 3.788987442501407e-05, "loss": 0.0001, "step": 37450 }, { "epoch": 0.73, "learning_rate": 3.7889227464756844e-05, "loss": 0.0004, "step": 37452 }, { "epoch": 0.73, "learning_rate": 3.788858050449961e-05, "loss": 0.0, "step": 37454 }, { "epoch": 0.73, "learning_rate": 3.7887933544242376e-05, "loss": 0.002, "step": 37456 }, { "epoch": 0.73, "learning_rate": 3.788728658398515e-05, "loss": 0.0027, "step": 37458 }, { "epoch": 0.73, "learning_rate": 3.7886639623727914e-05, "loss": 0.0001, "step": 37460 }, { "epoch": 0.73, "learning_rate": 3.788599266347069e-05, "loss": 0.0, "step": 37462 }, { "epoch": 0.73, "learning_rate": 3.788534570321345e-05, "loss": 0.0, "step": 37464 }, { "epoch": 0.73, "learning_rate": 3.788469874295622e-05, "loss": 0.0001, "step": 37466 }, { "epoch": 0.73, "learning_rate": 3.788405178269899e-05, "loss": 0.0001, "step": 37468 }, { "epoch": 0.73, "learning_rate": 3.788340482244176e-05, "loss": 0.0, "step": 37470 }, { "epoch": 0.73, "learning_rate": 3.788275786218453e-05, "loss": 0.0, "step": 37472 }, { "epoch": 0.73, "learning_rate": 3.78821109019273e-05, "loss": 0.0006, "step": 37474 }, { "epoch": 0.73, "learning_rate": 3.788146394167007e-05, "loss": 0.0, "step": 37476 }, { "epoch": 0.73, "learning_rate": 3.788081698141283e-05, "loss": 0.0262, "step": 37478 }, { "epoch": 0.73, "learning_rate": 3.7880170021155605e-05, "loss": 0.0002, "step": 37480 }, { "epoch": 0.73, "learning_rate": 3.787952306089837e-05, "loss": 0.0074, "step": 37482 }, { "epoch": 0.73, "learning_rate": 3.787887610064114e-05, "loss": 0.0002, "step": 37484 }, { "epoch": 0.73, "learning_rate": 3.7878229140383905e-05, "loss": 0.0, "step": 37486 }, { "epoch": 0.73, "learning_rate": 3.7877582180126674e-05, "loss": 0.0036, "step": 37488 }, { "epoch": 0.73, "learning_rate": 3.787693521986945e-05, "loss": 0.0, "step": 37490 }, { "epoch": 0.73, "learning_rate": 3.787628825961221e-05, "loss": 0.0058, "step": 37492 }, { "epoch": 0.73, "learning_rate": 3.787564129935498e-05, "loss": 0.0003, "step": 37494 }, { "epoch": 0.73, "learning_rate": 3.787499433909775e-05, "loss": 0.0001, "step": 37496 }, { "epoch": 0.73, "learning_rate": 3.787434737884052e-05, "loss": 0.0013, "step": 37498 }, { "epoch": 0.73, "learning_rate": 3.787370041858328e-05, "loss": 0.0344, "step": 37500 }, { "epoch": 0.73, "learning_rate": 3.787305345832606e-05, "loss": 0.0132, "step": 37502 }, { "epoch": 0.73, "learning_rate": 3.787240649806883e-05, "loss": 0.0005, "step": 37504 }, { "epoch": 0.73, "learning_rate": 3.7871759537811596e-05, "loss": 0.0002, "step": 37506 }, { "epoch": 0.73, "learning_rate": 3.7871112577554365e-05, "loss": 0.0002, "step": 37508 }, { "epoch": 0.73, "learning_rate": 3.787046561729713e-05, "loss": 0.0003, "step": 37510 }, { "epoch": 0.73, "learning_rate": 3.7869818657039903e-05, "loss": 0.0288, "step": 37512 }, { "epoch": 0.73, "learning_rate": 3.7869171696782666e-05, "loss": 0.0, "step": 37514 }, { "epoch": 0.73, "learning_rate": 3.7868524736525435e-05, "loss": 0.0073, "step": 37516 }, { "epoch": 0.73, "learning_rate": 3.7867877776268204e-05, "loss": 0.0, "step": 37518 }, { "epoch": 0.73, "learning_rate": 3.786723081601097e-05, "loss": 0.0, "step": 37520 }, { "epoch": 0.73, "learning_rate": 3.786658385575375e-05, "loss": 0.0001, "step": 37522 }, { "epoch": 0.73, "learning_rate": 3.786593689549651e-05, "loss": 0.0058, "step": 37524 }, { "epoch": 0.73, "learning_rate": 3.786528993523928e-05, "loss": 0.0001, "step": 37526 }, { "epoch": 0.73, "learning_rate": 3.786464297498205e-05, "loss": 0.0, "step": 37528 }, { "epoch": 0.73, "learning_rate": 3.786399601472482e-05, "loss": 0.0001, "step": 37530 }, { "epoch": 0.73, "learning_rate": 3.786334905446758e-05, "loss": 0.0003, "step": 37532 }, { "epoch": 0.73, "learning_rate": 3.786270209421036e-05, "loss": 0.0, "step": 37534 }, { "epoch": 0.73, "learning_rate": 3.7862055133953126e-05, "loss": 0.0033, "step": 37536 }, { "epoch": 0.73, "learning_rate": 3.786140817369589e-05, "loss": 0.0296, "step": 37538 }, { "epoch": 0.73, "learning_rate": 3.7860761213438664e-05, "loss": 0.0, "step": 37540 }, { "epoch": 0.73, "learning_rate": 3.7860114253181426e-05, "loss": 0.0002, "step": 37542 }, { "epoch": 0.73, "learning_rate": 3.78594672929242e-05, "loss": 0.0, "step": 37544 }, { "epoch": 0.73, "learning_rate": 3.7858820332666964e-05, "loss": 0.0001, "step": 37546 }, { "epoch": 0.73, "learning_rate": 3.7858173372409733e-05, "loss": 0.0009, "step": 37548 }, { "epoch": 0.73, "learning_rate": 3.78575264121525e-05, "loss": 0.0398, "step": 37550 }, { "epoch": 0.73, "learning_rate": 3.785687945189527e-05, "loss": 0.0, "step": 37552 }, { "epoch": 0.73, "learning_rate": 3.785623249163804e-05, "loss": 0.0, "step": 37554 }, { "epoch": 0.73, "learning_rate": 3.785558553138081e-05, "loss": 0.0, "step": 37556 }, { "epoch": 0.73, "learning_rate": 3.785493857112358e-05, "loss": 0.0001, "step": 37558 }, { "epoch": 0.73, "learning_rate": 3.785429161086634e-05, "loss": 0.0185, "step": 37560 }, { "epoch": 0.73, "learning_rate": 3.785364465060912e-05, "loss": 0.0, "step": 37562 }, { "epoch": 0.73, "learning_rate": 3.785299769035188e-05, "loss": 0.0002, "step": 37564 }, { "epoch": 0.73, "learning_rate": 3.7852350730094655e-05, "loss": 0.0, "step": 37566 }, { "epoch": 0.73, "learning_rate": 3.7851703769837424e-05, "loss": 0.0001, "step": 37568 }, { "epoch": 0.73, "learning_rate": 3.7851056809580187e-05, "loss": 0.0, "step": 37570 }, { "epoch": 0.73, "learning_rate": 3.785040984932296e-05, "loss": 0.0081, "step": 37572 }, { "epoch": 0.73, "learning_rate": 3.7849762889065725e-05, "loss": 0.0, "step": 37574 }, { "epoch": 0.73, "learning_rate": 3.7849115928808494e-05, "loss": 0.0001, "step": 37576 }, { "epoch": 0.73, "learning_rate": 3.784846896855126e-05, "loss": 0.0001, "step": 37578 }, { "epoch": 0.73, "learning_rate": 3.784782200829403e-05, "loss": 0.0006, "step": 37580 }, { "epoch": 0.73, "learning_rate": 3.78471750480368e-05, "loss": 0.0, "step": 37582 }, { "epoch": 0.73, "learning_rate": 3.784652808777957e-05, "loss": 0.0006, "step": 37584 }, { "epoch": 0.73, "learning_rate": 3.784588112752234e-05, "loss": 0.0029, "step": 37586 }, { "epoch": 0.73, "learning_rate": 3.784523416726511e-05, "loss": 0.003, "step": 37588 }, { "epoch": 0.73, "learning_rate": 3.784458720700788e-05, "loss": 0.0, "step": 37590 }, { "epoch": 0.73, "learning_rate": 3.784394024675064e-05, "loss": 0.0, "step": 37592 }, { "epoch": 0.73, "learning_rate": 3.7843293286493416e-05, "loss": 0.0, "step": 37594 }, { "epoch": 0.73, "learning_rate": 3.784264632623618e-05, "loss": 0.0002, "step": 37596 }, { "epoch": 0.73, "learning_rate": 3.784199936597895e-05, "loss": 0.003, "step": 37598 }, { "epoch": 0.73, "learning_rate": 3.784135240572172e-05, "loss": 0.0027, "step": 37600 }, { "epoch": 0.73, "learning_rate": 3.7840705445464485e-05, "loss": 0.003, "step": 37602 }, { "epoch": 0.73, "learning_rate": 3.784005848520726e-05, "loss": 0.0009, "step": 37604 }, { "epoch": 0.73, "learning_rate": 3.783941152495002e-05, "loss": 0.0001, "step": 37606 }, { "epoch": 0.73, "learning_rate": 3.783876456469279e-05, "loss": 0.0012, "step": 37608 }, { "epoch": 0.73, "learning_rate": 3.783811760443556e-05, "loss": 0.0002, "step": 37610 }, { "epoch": 0.73, "learning_rate": 3.783747064417833e-05, "loss": 0.0056, "step": 37612 }, { "epoch": 0.73, "learning_rate": 3.78368236839211e-05, "loss": 0.0001, "step": 37614 }, { "epoch": 0.73, "learning_rate": 3.783617672366387e-05, "loss": 0.0409, "step": 37616 }, { "epoch": 0.73, "learning_rate": 3.783552976340664e-05, "loss": 0.0, "step": 37618 }, { "epoch": 0.73, "learning_rate": 3.78348828031494e-05, "loss": 0.0001, "step": 37620 }, { "epoch": 0.73, "learning_rate": 3.7834235842892176e-05, "loss": 0.0, "step": 37622 }, { "epoch": 0.73, "learning_rate": 3.783358888263494e-05, "loss": 0.0, "step": 37624 }, { "epoch": 0.73, "learning_rate": 3.7832941922377714e-05, "loss": 0.0072, "step": 37626 }, { "epoch": 0.73, "learning_rate": 3.7832294962120476e-05, "loss": 0.0, "step": 37628 }, { "epoch": 0.73, "learning_rate": 3.7831648001863246e-05, "loss": 0.0011, "step": 37630 }, { "epoch": 0.73, "learning_rate": 3.7831001041606015e-05, "loss": 0.0001, "step": 37632 }, { "epoch": 0.73, "learning_rate": 3.7830354081348784e-05, "loss": 0.0001, "step": 37634 }, { "epoch": 0.73, "learning_rate": 3.782970712109155e-05, "loss": 0.0002, "step": 37636 }, { "epoch": 0.73, "learning_rate": 3.782906016083432e-05, "loss": 0.0, "step": 37638 }, { "epoch": 0.73, "learning_rate": 3.782841320057709e-05, "loss": 0.0028, "step": 37640 }, { "epoch": 0.73, "learning_rate": 3.782776624031985e-05, "loss": 0.0003, "step": 37642 }, { "epoch": 0.73, "learning_rate": 3.782711928006263e-05, "loss": 0.0, "step": 37644 }, { "epoch": 0.73, "learning_rate": 3.78264723198054e-05, "loss": 0.0009, "step": 37646 }, { "epoch": 0.73, "learning_rate": 3.782582535954817e-05, "loss": 0.0016, "step": 37648 }, { "epoch": 0.73, "learning_rate": 3.7825178399290936e-05, "loss": 0.0028, "step": 37650 }, { "epoch": 0.73, "learning_rate": 3.78245314390337e-05, "loss": 0.0063, "step": 37652 }, { "epoch": 0.73, "learning_rate": 3.7823884478776475e-05, "loss": 0.0, "step": 37654 }, { "epoch": 0.73, "learning_rate": 3.782323751851924e-05, "loss": 0.0022, "step": 37656 }, { "epoch": 0.73, "learning_rate": 3.7822590558262006e-05, "loss": 0.0, "step": 37658 }, { "epoch": 0.73, "learning_rate": 3.7821943598004775e-05, "loss": 0.0062, "step": 37660 }, { "epoch": 0.73, "learning_rate": 3.7821296637747544e-05, "loss": 0.0, "step": 37662 }, { "epoch": 0.73, "learning_rate": 3.782064967749031e-05, "loss": 0.0001, "step": 37664 }, { "epoch": 0.73, "learning_rate": 3.782000271723308e-05, "loss": 0.0003, "step": 37666 }, { "epoch": 0.73, "learning_rate": 3.781935575697585e-05, "loss": 0.0001, "step": 37668 }, { "epoch": 0.73, "learning_rate": 3.781870879671862e-05, "loss": 0.0, "step": 37670 }, { "epoch": 0.73, "learning_rate": 3.781806183646139e-05, "loss": 0.0004, "step": 37672 }, { "epoch": 0.73, "learning_rate": 3.781741487620415e-05, "loss": 0.0, "step": 37674 }, { "epoch": 0.73, "learning_rate": 3.781676791594693e-05, "loss": 0.0007, "step": 37676 }, { "epoch": 0.73, "learning_rate": 3.781612095568969e-05, "loss": 0.0346, "step": 37678 }, { "epoch": 0.73, "learning_rate": 3.781547399543246e-05, "loss": 0.0073, "step": 37680 }, { "epoch": 0.73, "learning_rate": 3.7814827035175235e-05, "loss": 0.0012, "step": 37682 }, { "epoch": 0.73, "learning_rate": 3.7814180074918e-05, "loss": 0.0, "step": 37684 }, { "epoch": 0.73, "learning_rate": 3.781353311466077e-05, "loss": 0.0007, "step": 37686 }, { "epoch": 0.73, "learning_rate": 3.7812886154403535e-05, "loss": 0.0015, "step": 37688 }, { "epoch": 0.73, "learning_rate": 3.7812239194146305e-05, "loss": 0.0001, "step": 37690 }, { "epoch": 0.73, "learning_rate": 3.7811592233889074e-05, "loss": 0.0, "step": 37692 }, { "epoch": 0.73, "learning_rate": 3.781094527363184e-05, "loss": 0.0, "step": 37694 }, { "epoch": 0.73, "learning_rate": 3.781029831337461e-05, "loss": 0.0001, "step": 37696 }, { "epoch": 0.73, "learning_rate": 3.780965135311738e-05, "loss": 0.0, "step": 37698 }, { "epoch": 0.73, "learning_rate": 3.780900439286015e-05, "loss": 0.0116, "step": 37700 }, { "epoch": 0.73, "learning_rate": 3.780835743260291e-05, "loss": 0.0001, "step": 37702 }, { "epoch": 0.73, "learning_rate": 3.780771047234569e-05, "loss": 0.0001, "step": 37704 }, { "epoch": 0.73, "learning_rate": 3.780706351208845e-05, "loss": 0.0, "step": 37706 }, { "epoch": 0.73, "learning_rate": 3.7806416551831226e-05, "loss": 0.0, "step": 37708 }, { "epoch": 0.73, "learning_rate": 3.780576959157399e-05, "loss": 0.0, "step": 37710 }, { "epoch": 0.73, "learning_rate": 3.780512263131676e-05, "loss": 0.0011, "step": 37712 }, { "epoch": 0.73, "learning_rate": 3.7804475671059534e-05, "loss": 0.0003, "step": 37714 }, { "epoch": 0.73, "learning_rate": 3.7803828710802296e-05, "loss": 0.0006, "step": 37716 }, { "epoch": 0.73, "learning_rate": 3.7803181750545065e-05, "loss": 0.0002, "step": 37718 }, { "epoch": 0.73, "learning_rate": 3.7802534790287834e-05, "loss": 0.0019, "step": 37720 }, { "epoch": 0.73, "learning_rate": 3.78018878300306e-05, "loss": 0.0, "step": 37722 }, { "epoch": 0.73, "learning_rate": 3.780124086977337e-05, "loss": 0.0, "step": 37724 }, { "epoch": 0.73, "learning_rate": 3.780059390951614e-05, "loss": 0.0, "step": 37726 }, { "epoch": 0.73, "learning_rate": 3.779994694925891e-05, "loss": 0.0086, "step": 37728 }, { "epoch": 0.73, "learning_rate": 3.779929998900168e-05, "loss": 0.0114, "step": 37730 }, { "epoch": 0.73, "learning_rate": 3.779865302874445e-05, "loss": 0.0172, "step": 37732 }, { "epoch": 0.73, "learning_rate": 3.779800606848721e-05, "loss": 0.0, "step": 37734 }, { "epoch": 0.73, "learning_rate": 3.779735910822999e-05, "loss": 0.0004, "step": 37736 }, { "epoch": 0.73, "learning_rate": 3.779671214797275e-05, "loss": 0.0, "step": 37738 }, { "epoch": 0.73, "learning_rate": 3.779606518771552e-05, "loss": 0.0003, "step": 37740 }, { "epoch": 0.73, "learning_rate": 3.779541822745829e-05, "loss": 0.0, "step": 37742 }, { "epoch": 0.73, "learning_rate": 3.7794771267201056e-05, "loss": 0.0001, "step": 37744 }, { "epoch": 0.73, "learning_rate": 3.779412430694383e-05, "loss": 0.0, "step": 37746 }, { "epoch": 0.73, "learning_rate": 3.7793477346686594e-05, "loss": 0.0, "step": 37748 }, { "epoch": 0.73, "learning_rate": 3.7792830386429364e-05, "loss": 0.0001, "step": 37750 }, { "epoch": 0.73, "learning_rate": 3.779218342617213e-05, "loss": 0.0116, "step": 37752 }, { "epoch": 0.73, "learning_rate": 3.77915364659149e-05, "loss": 0.0001, "step": 37754 }, { "epoch": 0.73, "learning_rate": 3.7790889505657664e-05, "loss": 0.0, "step": 37756 }, { "epoch": 0.73, "learning_rate": 3.779024254540044e-05, "loss": 0.0, "step": 37758 }, { "epoch": 0.73, "learning_rate": 3.778959558514321e-05, "loss": 0.0001, "step": 37760 }, { "epoch": 0.73, "learning_rate": 3.778894862488597e-05, "loss": 0.0055, "step": 37762 }, { "epoch": 0.73, "learning_rate": 3.778830166462875e-05, "loss": 0.0001, "step": 37764 }, { "epoch": 0.73, "learning_rate": 3.778765470437151e-05, "loss": 0.0, "step": 37766 }, { "epoch": 0.73, "learning_rate": 3.7787007744114285e-05, "loss": 0.0, "step": 37768 }, { "epoch": 0.73, "learning_rate": 3.778636078385705e-05, "loss": 0.0, "step": 37770 }, { "epoch": 0.73, "learning_rate": 3.778571382359982e-05, "loss": 0.0, "step": 37772 }, { "epoch": 0.73, "learning_rate": 3.7785066863342586e-05, "loss": 0.0006, "step": 37774 }, { "epoch": 0.73, "learning_rate": 3.7784419903085355e-05, "loss": 0.0011, "step": 37776 }, { "epoch": 0.73, "learning_rate": 3.7783772942828124e-05, "loss": 0.0001, "step": 37778 }, { "epoch": 0.73, "learning_rate": 3.778312598257089e-05, "loss": 0.0027, "step": 37780 }, { "epoch": 0.73, "learning_rate": 3.778247902231366e-05, "loss": 0.0, "step": 37782 }, { "epoch": 0.73, "learning_rate": 3.7781832062056424e-05, "loss": 0.0004, "step": 37784 }, { "epoch": 0.73, "learning_rate": 3.77811851017992e-05, "loss": 0.0018, "step": 37786 }, { "epoch": 0.73, "learning_rate": 3.778053814154196e-05, "loss": 0.0, "step": 37788 }, { "epoch": 0.73, "learning_rate": 3.777989118128474e-05, "loss": 0.0, "step": 37790 }, { "epoch": 0.73, "learning_rate": 3.777924422102751e-05, "loss": 0.0003, "step": 37792 }, { "epoch": 0.73, "learning_rate": 3.777859726077027e-05, "loss": 0.0, "step": 37794 }, { "epoch": 0.73, "learning_rate": 3.7777950300513046e-05, "loss": 0.0, "step": 37796 }, { "epoch": 0.73, "learning_rate": 3.777730334025581e-05, "loss": 0.0124, "step": 37798 }, { "epoch": 0.73, "learning_rate": 3.777665637999858e-05, "loss": 0.019, "step": 37800 }, { "epoch": 0.73, "learning_rate": 3.7776009419741346e-05, "loss": 0.0, "step": 37802 }, { "epoch": 0.73, "learning_rate": 3.7775362459484115e-05, "loss": 0.0, "step": 37804 }, { "epoch": 0.73, "learning_rate": 3.7774715499226884e-05, "loss": 0.0047, "step": 37806 }, { "epoch": 0.73, "learning_rate": 3.7774068538969653e-05, "loss": 0.0001, "step": 37808 }, { "epoch": 0.73, "learning_rate": 3.777342157871242e-05, "loss": 0.0129, "step": 37810 }, { "epoch": 0.73, "learning_rate": 3.777277461845519e-05, "loss": 0.0005, "step": 37812 }, { "epoch": 0.73, "learning_rate": 3.777212765819796e-05, "loss": 0.0, "step": 37814 }, { "epoch": 0.73, "learning_rate": 3.777148069794072e-05, "loss": 0.0, "step": 37816 }, { "epoch": 0.73, "learning_rate": 3.77708337376835e-05, "loss": 0.0, "step": 37818 }, { "epoch": 0.73, "learning_rate": 3.777018677742626e-05, "loss": 0.0291, "step": 37820 }, { "epoch": 0.73, "learning_rate": 3.776953981716903e-05, "loss": 0.0, "step": 37822 }, { "epoch": 0.73, "learning_rate": 3.7768892856911806e-05, "loss": 0.0001, "step": 37824 }, { "epoch": 0.73, "learning_rate": 3.776824589665457e-05, "loss": 0.0005, "step": 37826 }, { "epoch": 0.73, "learning_rate": 3.7767598936397344e-05, "loss": 0.0, "step": 37828 }, { "epoch": 0.73, "learning_rate": 3.7766951976140107e-05, "loss": 0.004, "step": 37830 }, { "epoch": 0.73, "learning_rate": 3.7766305015882876e-05, "loss": 0.0002, "step": 37832 }, { "epoch": 0.73, "learning_rate": 3.7765658055625645e-05, "loss": 0.0, "step": 37834 }, { "epoch": 0.73, "learning_rate": 3.7765011095368414e-05, "loss": 0.0001, "step": 37836 }, { "epoch": 0.73, "learning_rate": 3.776436413511118e-05, "loss": 0.0001, "step": 37838 }, { "epoch": 0.73, "learning_rate": 3.776371717485395e-05, "loss": 0.0, "step": 37840 }, { "epoch": 0.73, "learning_rate": 3.776307021459672e-05, "loss": 0.0, "step": 37842 }, { "epoch": 0.73, "learning_rate": 3.776242325433948e-05, "loss": 0.0002, "step": 37844 }, { "epoch": 0.73, "learning_rate": 3.776177629408226e-05, "loss": 0.0006, "step": 37846 }, { "epoch": 0.73, "learning_rate": 3.776112933382502e-05, "loss": 0.0003, "step": 37848 }, { "epoch": 0.73, "learning_rate": 3.77604823735678e-05, "loss": 0.0, "step": 37850 }, { "epoch": 0.73, "learning_rate": 3.775983541331056e-05, "loss": 0.0001, "step": 37852 }, { "epoch": 0.73, "learning_rate": 3.775918845305333e-05, "loss": 0.0055, "step": 37854 }, { "epoch": 0.73, "learning_rate": 3.77585414927961e-05, "loss": 0.0231, "step": 37856 }, { "epoch": 0.73, "learning_rate": 3.775789453253887e-05, "loss": 0.0, "step": 37858 }, { "epoch": 0.73, "learning_rate": 3.7757247572281636e-05, "loss": 0.0009, "step": 37860 }, { "epoch": 0.73, "learning_rate": 3.7756600612024405e-05, "loss": 0.0024, "step": 37862 }, { "epoch": 0.73, "learning_rate": 3.7755953651767174e-05, "loss": 0.0, "step": 37864 }, { "epoch": 0.73, "learning_rate": 3.7755306691509937e-05, "loss": 0.0, "step": 37866 }, { "epoch": 0.73, "learning_rate": 3.775465973125271e-05, "loss": 0.0, "step": 37868 }, { "epoch": 0.74, "learning_rate": 3.775401277099548e-05, "loss": 0.0, "step": 37870 }, { "epoch": 0.74, "learning_rate": 3.775336581073825e-05, "loss": 0.0226, "step": 37872 }, { "epoch": 0.74, "learning_rate": 3.775271885048102e-05, "loss": 0.0145, "step": 37874 }, { "epoch": 0.74, "learning_rate": 3.775207189022378e-05, "loss": 0.016, "step": 37876 }, { "epoch": 0.74, "learning_rate": 3.775142492996656e-05, "loss": 0.0001, "step": 37878 }, { "epoch": 0.74, "learning_rate": 3.775077796970932e-05, "loss": 0.0, "step": 37880 }, { "epoch": 0.74, "learning_rate": 3.775013100945209e-05, "loss": 0.0161, "step": 37882 }, { "epoch": 0.74, "learning_rate": 3.774948404919486e-05, "loss": 0.0001, "step": 37884 }, { "epoch": 0.74, "learning_rate": 3.774883708893763e-05, "loss": 0.0056, "step": 37886 }, { "epoch": 0.74, "learning_rate": 3.7748190128680396e-05, "loss": 0.0, "step": 37888 }, { "epoch": 0.74, "learning_rate": 3.7747543168423166e-05, "loss": 0.0001, "step": 37890 }, { "epoch": 0.74, "learning_rate": 3.7746896208165935e-05, "loss": 0.0, "step": 37892 }, { "epoch": 0.74, "learning_rate": 3.7746249247908704e-05, "loss": 0.0, "step": 37894 }, { "epoch": 0.74, "learning_rate": 3.774560228765147e-05, "loss": 0.0001, "step": 37896 }, { "epoch": 0.74, "learning_rate": 3.7744955327394235e-05, "loss": 0.0036, "step": 37898 }, { "epoch": 0.74, "learning_rate": 3.774430836713701e-05, "loss": 0.0, "step": 37900 }, { "epoch": 0.74, "learning_rate": 3.774366140687978e-05, "loss": 0.0131, "step": 37902 }, { "epoch": 0.74, "learning_rate": 3.774301444662254e-05, "loss": 0.003, "step": 37904 }, { "epoch": 0.74, "learning_rate": 3.774236748636532e-05, "loss": 0.0001, "step": 37906 }, { "epoch": 0.74, "learning_rate": 3.774172052610808e-05, "loss": 0.0008, "step": 37908 }, { "epoch": 0.74, "learning_rate": 3.7741073565850856e-05, "loss": 0.0001, "step": 37910 }, { "epoch": 0.74, "learning_rate": 3.774042660559362e-05, "loss": 0.0001, "step": 37912 }, { "epoch": 0.74, "learning_rate": 3.773977964533639e-05, "loss": 0.0044, "step": 37914 }, { "epoch": 0.74, "learning_rate": 3.773913268507916e-05, "loss": 0.0014, "step": 37916 }, { "epoch": 0.74, "learning_rate": 3.7738485724821926e-05, "loss": 0.0, "step": 37918 }, { "epoch": 0.74, "learning_rate": 3.7737838764564695e-05, "loss": 0.0016, "step": 37920 }, { "epoch": 0.74, "learning_rate": 3.7737191804307464e-05, "loss": 0.0, "step": 37922 }, { "epoch": 0.74, "learning_rate": 3.773654484405023e-05, "loss": 0.0, "step": 37924 }, { "epoch": 0.74, "learning_rate": 3.7735897883792996e-05, "loss": 0.0, "step": 37926 }, { "epoch": 0.74, "learning_rate": 3.773525092353577e-05, "loss": 0.0006, "step": 37928 }, { "epoch": 0.74, "learning_rate": 3.7734603963278534e-05, "loss": 0.0001, "step": 37930 }, { "epoch": 0.74, "learning_rate": 3.773395700302131e-05, "loss": 0.0006, "step": 37932 }, { "epoch": 0.74, "learning_rate": 3.773331004276407e-05, "loss": 0.0, "step": 37934 }, { "epoch": 0.74, "learning_rate": 3.773266308250684e-05, "loss": 0.0, "step": 37936 }, { "epoch": 0.74, "learning_rate": 3.773201612224962e-05, "loss": 0.1134, "step": 37938 }, { "epoch": 0.74, "learning_rate": 3.773136916199238e-05, "loss": 0.0001, "step": 37940 }, { "epoch": 0.74, "learning_rate": 3.773072220173515e-05, "loss": 0.0, "step": 37942 }, { "epoch": 0.74, "learning_rate": 3.773007524147792e-05, "loss": 0.001, "step": 37944 }, { "epoch": 0.74, "learning_rate": 3.7729428281220686e-05, "loss": 0.0102, "step": 37946 }, { "epoch": 0.74, "learning_rate": 3.7728781320963455e-05, "loss": 0.0, "step": 37948 }, { "epoch": 0.74, "learning_rate": 3.7728134360706225e-05, "loss": 0.0, "step": 37950 }, { "epoch": 0.74, "learning_rate": 3.7727487400448994e-05, "loss": 0.0, "step": 37952 }, { "epoch": 0.74, "learning_rate": 3.772684044019176e-05, "loss": 0.0017, "step": 37954 }, { "epoch": 0.74, "learning_rate": 3.772619347993453e-05, "loss": 0.0003, "step": 37956 }, { "epoch": 0.74, "learning_rate": 3.7725546519677294e-05, "loss": 0.0049, "step": 37958 }, { "epoch": 0.74, "learning_rate": 3.772489955942007e-05, "loss": 0.0, "step": 37960 }, { "epoch": 0.74, "learning_rate": 3.772425259916283e-05, "loss": 0.0009, "step": 37962 }, { "epoch": 0.74, "learning_rate": 3.77236056389056e-05, "loss": 0.0015, "step": 37964 }, { "epoch": 0.74, "learning_rate": 3.772295867864837e-05, "loss": 0.0, "step": 37966 }, { "epoch": 0.74, "learning_rate": 3.772231171839114e-05, "loss": 0.0, "step": 37968 }, { "epoch": 0.74, "learning_rate": 3.7721664758133915e-05, "loss": 0.0002, "step": 37970 }, { "epoch": 0.74, "learning_rate": 3.772101779787668e-05, "loss": 0.0003, "step": 37972 }, { "epoch": 0.74, "learning_rate": 3.772037083761945e-05, "loss": 0.0, "step": 37974 }, { "epoch": 0.74, "learning_rate": 3.7719723877362216e-05, "loss": 0.0056, "step": 37976 }, { "epoch": 0.74, "learning_rate": 3.7719076917104985e-05, "loss": 0.0, "step": 37978 }, { "epoch": 0.74, "learning_rate": 3.771842995684775e-05, "loss": 0.0029, "step": 37980 }, { "epoch": 0.74, "learning_rate": 3.771778299659052e-05, "loss": 0.0003, "step": 37982 }, { "epoch": 0.74, "learning_rate": 3.771713603633329e-05, "loss": 0.0111, "step": 37984 }, { "epoch": 0.74, "learning_rate": 3.7716489076076054e-05, "loss": 0.0001, "step": 37986 }, { "epoch": 0.74, "learning_rate": 3.771584211581883e-05, "loss": 0.0, "step": 37988 }, { "epoch": 0.74, "learning_rate": 3.771519515556159e-05, "loss": 0.0001, "step": 37990 }, { "epoch": 0.74, "learning_rate": 3.771454819530437e-05, "loss": 0.0, "step": 37992 }, { "epoch": 0.74, "learning_rate": 3.771390123504713e-05, "loss": 0.0145, "step": 37994 }, { "epoch": 0.74, "learning_rate": 3.77132542747899e-05, "loss": 0.0, "step": 37996 }, { "epoch": 0.74, "learning_rate": 3.771260731453267e-05, "loss": 0.0003, "step": 37998 }, { "epoch": 0.74, "learning_rate": 3.771196035427544e-05, "loss": 0.0, "step": 38000 }, { "epoch": 0.74, "learning_rate": 3.771131339401821e-05, "loss": 0.0303, "step": 38002 }, { "epoch": 0.74, "learning_rate": 3.7710666433760976e-05, "loss": 0.0005, "step": 38004 }, { "epoch": 0.74, "learning_rate": 3.7710019473503745e-05, "loss": 0.0106, "step": 38006 }, { "epoch": 0.74, "learning_rate": 3.770937251324651e-05, "loss": 0.0162, "step": 38008 }, { "epoch": 0.74, "learning_rate": 3.7708725552989284e-05, "loss": 0.0021, "step": 38010 }, { "epoch": 0.74, "learning_rate": 3.7708078592732046e-05, "loss": 0.0, "step": 38012 }, { "epoch": 0.74, "learning_rate": 3.770743163247482e-05, "loss": 0.0035, "step": 38014 }, { "epoch": 0.74, "learning_rate": 3.770678467221759e-05, "loss": 0.0002, "step": 38016 }, { "epoch": 0.74, "learning_rate": 3.770613771196035e-05, "loss": 0.0, "step": 38018 }, { "epoch": 0.74, "learning_rate": 3.770549075170313e-05, "loss": 0.0009, "step": 38020 }, { "epoch": 0.74, "learning_rate": 3.770484379144589e-05, "loss": 0.0013, "step": 38022 }, { "epoch": 0.74, "learning_rate": 3.770419683118866e-05, "loss": 0.0001, "step": 38024 }, { "epoch": 0.74, "learning_rate": 3.770354987093143e-05, "loss": 0.0, "step": 38026 }, { "epoch": 0.74, "learning_rate": 3.77029029106742e-05, "loss": 0.0102, "step": 38028 }, { "epoch": 0.74, "learning_rate": 3.770225595041697e-05, "loss": 0.0017, "step": 38030 }, { "epoch": 0.74, "learning_rate": 3.770160899015974e-05, "loss": 0.0002, "step": 38032 }, { "epoch": 0.74, "learning_rate": 3.7700962029902506e-05, "loss": 0.0, "step": 38034 }, { "epoch": 0.74, "learning_rate": 3.7700315069645275e-05, "loss": 0.0033, "step": 38036 }, { "epoch": 0.74, "learning_rate": 3.7699668109388044e-05, "loss": 0.0001, "step": 38038 }, { "epoch": 0.74, "learning_rate": 3.7699021149130806e-05, "loss": 0.0, "step": 38040 }, { "epoch": 0.74, "learning_rate": 3.769837418887358e-05, "loss": 0.0, "step": 38042 }, { "epoch": 0.74, "learning_rate": 3.7697727228616344e-05, "loss": 0.0, "step": 38044 }, { "epoch": 0.74, "learning_rate": 3.7697080268359113e-05, "loss": 0.0119, "step": 38046 }, { "epoch": 0.74, "learning_rate": 3.769643330810189e-05, "loss": 0.0, "step": 38048 }, { "epoch": 0.74, "learning_rate": 3.769578634784465e-05, "loss": 0.0, "step": 38050 }, { "epoch": 0.74, "learning_rate": 3.769513938758743e-05, "loss": 0.0001, "step": 38052 }, { "epoch": 0.74, "learning_rate": 3.769449242733019e-05, "loss": 0.0001, "step": 38054 }, { "epoch": 0.74, "learning_rate": 3.769384546707296e-05, "loss": 0.0, "step": 38056 }, { "epoch": 0.74, "learning_rate": 3.769319850681573e-05, "loss": 0.0001, "step": 38058 }, { "epoch": 0.74, "learning_rate": 3.76925515465585e-05, "loss": 0.0008, "step": 38060 }, { "epoch": 0.74, "learning_rate": 3.7691904586301266e-05, "loss": 0.0, "step": 38062 }, { "epoch": 0.74, "learning_rate": 3.7691257626044035e-05, "loss": 0.0001, "step": 38064 }, { "epoch": 0.74, "learning_rate": 3.7690610665786804e-05, "loss": 0.0002, "step": 38066 }, { "epoch": 0.74, "learning_rate": 3.768996370552957e-05, "loss": 0.0001, "step": 38068 }, { "epoch": 0.74, "learning_rate": 3.768931674527234e-05, "loss": 0.0041, "step": 38070 }, { "epoch": 0.74, "learning_rate": 3.7688669785015105e-05, "loss": 0.0, "step": 38072 }, { "epoch": 0.74, "learning_rate": 3.768802282475788e-05, "loss": 0.0147, "step": 38074 }, { "epoch": 0.74, "learning_rate": 3.768737586450064e-05, "loss": 0.0001, "step": 38076 }, { "epoch": 0.74, "learning_rate": 3.768672890424341e-05, "loss": 0.0003, "step": 38078 }, { "epoch": 0.74, "learning_rate": 3.768608194398619e-05, "loss": 0.0001, "step": 38080 }, { "epoch": 0.74, "learning_rate": 3.768543498372895e-05, "loss": 0.0013, "step": 38082 }, { "epoch": 0.74, "learning_rate": 3.768478802347172e-05, "loss": 0.0051, "step": 38084 }, { "epoch": 0.74, "learning_rate": 3.768414106321449e-05, "loss": 0.004, "step": 38086 }, { "epoch": 0.74, "learning_rate": 3.768349410295726e-05, "loss": 0.0168, "step": 38088 }, { "epoch": 0.74, "learning_rate": 3.7682847142700027e-05, "loss": 0.0, "step": 38090 }, { "epoch": 0.74, "learning_rate": 3.7682200182442796e-05, "loss": 0.0004, "step": 38092 }, { "epoch": 0.74, "learning_rate": 3.7681553222185565e-05, "loss": 0.0027, "step": 38094 }, { "epoch": 0.74, "learning_rate": 3.7680906261928334e-05, "loss": 0.0, "step": 38096 }, { "epoch": 0.74, "learning_rate": 3.76802593016711e-05, "loss": 0.0, "step": 38098 }, { "epoch": 0.74, "learning_rate": 3.7679612341413865e-05, "loss": 0.0023, "step": 38100 }, { "epoch": 0.74, "learning_rate": 3.767896538115664e-05, "loss": 0.0, "step": 38102 }, { "epoch": 0.74, "learning_rate": 3.76783184208994e-05, "loss": 0.0004, "step": 38104 }, { "epoch": 0.74, "learning_rate": 3.767767146064217e-05, "loss": 0.0094, "step": 38106 }, { "epoch": 0.74, "learning_rate": 3.767702450038494e-05, "loss": 0.0001, "step": 38108 }, { "epoch": 0.74, "learning_rate": 3.767637754012771e-05, "loss": 0.0, "step": 38110 }, { "epoch": 0.74, "learning_rate": 3.767573057987048e-05, "loss": 0.0046, "step": 38112 }, { "epoch": 0.74, "learning_rate": 3.767508361961325e-05, "loss": 0.0114, "step": 38114 }, { "epoch": 0.74, "learning_rate": 3.767443665935602e-05, "loss": 0.0, "step": 38116 }, { "epoch": 0.74, "learning_rate": 3.767378969909879e-05, "loss": 0.0, "step": 38118 }, { "epoch": 0.74, "learning_rate": 3.7673142738841556e-05, "loss": 0.0276, "step": 38120 }, { "epoch": 0.74, "learning_rate": 3.767249577858432e-05, "loss": 0.0, "step": 38122 }, { "epoch": 0.74, "learning_rate": 3.7671848818327094e-05, "loss": 0.0001, "step": 38124 }, { "epoch": 0.74, "learning_rate": 3.767120185806986e-05, "loss": 0.0001, "step": 38126 }, { "epoch": 0.74, "learning_rate": 3.7670554897812626e-05, "loss": 0.0, "step": 38128 }, { "epoch": 0.74, "learning_rate": 3.76699079375554e-05, "loss": 0.0, "step": 38130 }, { "epoch": 0.74, "learning_rate": 3.7669260977298164e-05, "loss": 0.0001, "step": 38132 }, { "epoch": 0.74, "learning_rate": 3.766861401704094e-05, "loss": 0.0, "step": 38134 }, { "epoch": 0.74, "learning_rate": 3.76679670567837e-05, "loss": 0.003, "step": 38136 }, { "epoch": 0.74, "learning_rate": 3.766732009652647e-05, "loss": 0.0182, "step": 38138 }, { "epoch": 0.74, "learning_rate": 3.766667313626924e-05, "loss": 0.0, "step": 38140 }, { "epoch": 0.74, "learning_rate": 3.766602617601201e-05, "loss": 0.0, "step": 38142 }, { "epoch": 0.74, "learning_rate": 3.766537921575478e-05, "loss": 0.0228, "step": 38144 }, { "epoch": 0.74, "learning_rate": 3.766473225549755e-05, "loss": 0.0, "step": 38146 }, { "epoch": 0.74, "learning_rate": 3.7664085295240316e-05, "loss": 0.0, "step": 38148 }, { "epoch": 0.74, "learning_rate": 3.766343833498308e-05, "loss": 0.0001, "step": 38150 }, { "epoch": 0.74, "learning_rate": 3.7662791374725855e-05, "loss": 0.0, "step": 38152 }, { "epoch": 0.74, "learning_rate": 3.766214441446862e-05, "loss": 0.0, "step": 38154 }, { "epoch": 0.74, "learning_rate": 3.766149745421139e-05, "loss": 0.0, "step": 38156 }, { "epoch": 0.74, "learning_rate": 3.7660850493954155e-05, "loss": 0.0001, "step": 38158 }, { "epoch": 0.74, "learning_rate": 3.7660203533696924e-05, "loss": 0.0123, "step": 38160 }, { "epoch": 0.74, "learning_rate": 3.76595565734397e-05, "loss": 0.0, "step": 38162 }, { "epoch": 0.74, "learning_rate": 3.765890961318246e-05, "loss": 0.0043, "step": 38164 }, { "epoch": 0.74, "learning_rate": 3.765826265292523e-05, "loss": 0.0108, "step": 38166 }, { "epoch": 0.74, "learning_rate": 3.7657615692668e-05, "loss": 0.015, "step": 38168 }, { "epoch": 0.74, "learning_rate": 3.765696873241077e-05, "loss": 0.0, "step": 38170 }, { "epoch": 0.74, "learning_rate": 3.765632177215354e-05, "loss": 0.0005, "step": 38172 }, { "epoch": 0.74, "learning_rate": 3.765567481189631e-05, "loss": 0.0001, "step": 38174 }, { "epoch": 0.74, "learning_rate": 3.765502785163908e-05, "loss": 0.0003, "step": 38176 }, { "epoch": 0.74, "learning_rate": 3.7654380891381846e-05, "loss": 0.0, "step": 38178 }, { "epoch": 0.74, "learning_rate": 3.7653733931124615e-05, "loss": 0.0049, "step": 38180 }, { "epoch": 0.74, "learning_rate": 3.765308697086738e-05, "loss": 0.0, "step": 38182 }, { "epoch": 0.74, "learning_rate": 3.765244001061015e-05, "loss": 0.0, "step": 38184 }, { "epoch": 0.74, "learning_rate": 3.7651793050352915e-05, "loss": 0.0, "step": 38186 }, { "epoch": 0.74, "learning_rate": 3.7651146090095685e-05, "loss": 0.0012, "step": 38188 }, { "epoch": 0.74, "learning_rate": 3.7650499129838454e-05, "loss": 0.0141, "step": 38190 }, { "epoch": 0.74, "learning_rate": 3.764985216958122e-05, "loss": 0.0, "step": 38192 }, { "epoch": 0.74, "learning_rate": 3.7649205209324e-05, "loss": 0.0, "step": 38194 }, { "epoch": 0.74, "learning_rate": 3.764855824906676e-05, "loss": 0.0, "step": 38196 }, { "epoch": 0.74, "learning_rate": 3.764791128880953e-05, "loss": 0.0058, "step": 38198 }, { "epoch": 0.74, "learning_rate": 3.76472643285523e-05, "loss": 0.0003, "step": 38200 }, { "epoch": 0.74, "learning_rate": 3.764661736829507e-05, "loss": 0.002, "step": 38202 }, { "epoch": 0.74, "learning_rate": 3.764597040803784e-05, "loss": 0.0013, "step": 38204 }, { "epoch": 0.74, "learning_rate": 3.7645323447780606e-05, "loss": 0.0027, "step": 38206 }, { "epoch": 0.74, "learning_rate": 3.7644676487523375e-05, "loss": 0.0, "step": 38208 }, { "epoch": 0.74, "learning_rate": 3.764402952726614e-05, "loss": 0.0128, "step": 38210 }, { "epoch": 0.74, "learning_rate": 3.7643382567008914e-05, "loss": 0.0, "step": 38212 }, { "epoch": 0.74, "learning_rate": 3.7642735606751676e-05, "loss": 0.0232, "step": 38214 }, { "epoch": 0.74, "learning_rate": 3.764208864649445e-05, "loss": 0.0049, "step": 38216 }, { "epoch": 0.74, "learning_rate": 3.7641441686237214e-05, "loss": 0.0, "step": 38218 }, { "epoch": 0.74, "learning_rate": 3.764079472597998e-05, "loss": 0.0001, "step": 38220 }, { "epoch": 0.74, "learning_rate": 3.764014776572275e-05, "loss": 0.0001, "step": 38222 }, { "epoch": 0.74, "learning_rate": 3.763950080546552e-05, "loss": 0.0047, "step": 38224 }, { "epoch": 0.74, "learning_rate": 3.763885384520829e-05, "loss": 0.0003, "step": 38226 }, { "epoch": 0.74, "learning_rate": 3.763820688495106e-05, "loss": 0.0, "step": 38228 }, { "epoch": 0.74, "learning_rate": 3.763755992469383e-05, "loss": 0.0043, "step": 38230 }, { "epoch": 0.74, "learning_rate": 3.76369129644366e-05, "loss": 0.0137, "step": 38232 }, { "epoch": 0.74, "learning_rate": 3.763626600417937e-05, "loss": 0.0, "step": 38234 }, { "epoch": 0.74, "learning_rate": 3.763561904392213e-05, "loss": 0.0157, "step": 38236 }, { "epoch": 0.74, "learning_rate": 3.7634972083664905e-05, "loss": 0.0248, "step": 38238 }, { "epoch": 0.74, "learning_rate": 3.7634325123407674e-05, "loss": 0.0, "step": 38240 }, { "epoch": 0.74, "learning_rate": 3.7633678163150436e-05, "loss": 0.0001, "step": 38242 }, { "epoch": 0.74, "learning_rate": 3.763303120289321e-05, "loss": 0.0, "step": 38244 }, { "epoch": 0.74, "learning_rate": 3.7632384242635974e-05, "loss": 0.0, "step": 38246 }, { "epoch": 0.74, "learning_rate": 3.7631737282378744e-05, "loss": 0.0001, "step": 38248 }, { "epoch": 0.74, "learning_rate": 3.763109032212151e-05, "loss": 0.0, "step": 38250 }, { "epoch": 0.74, "learning_rate": 3.763044336186428e-05, "loss": 0.0, "step": 38252 }, { "epoch": 0.74, "learning_rate": 3.762979640160705e-05, "loss": 0.0004, "step": 38254 }, { "epoch": 0.74, "learning_rate": 3.762914944134982e-05, "loss": 0.0066, "step": 38256 }, { "epoch": 0.74, "learning_rate": 3.762850248109259e-05, "loss": 0.0001, "step": 38258 }, { "epoch": 0.74, "learning_rate": 3.762785552083536e-05, "loss": 0.0, "step": 38260 }, { "epoch": 0.74, "learning_rate": 3.762720856057813e-05, "loss": 0.0, "step": 38262 }, { "epoch": 0.74, "learning_rate": 3.762656160032089e-05, "loss": 0.0044, "step": 38264 }, { "epoch": 0.74, "learning_rate": 3.7625914640063665e-05, "loss": 0.0, "step": 38266 }, { "epoch": 0.74, "learning_rate": 3.762526767980643e-05, "loss": 0.0, "step": 38268 }, { "epoch": 0.74, "learning_rate": 3.76246207195492e-05, "loss": 0.0001, "step": 38270 }, { "epoch": 0.74, "learning_rate": 3.762397375929197e-05, "loss": 0.0013, "step": 38272 }, { "epoch": 0.74, "learning_rate": 3.7623326799034735e-05, "loss": 0.0, "step": 38274 }, { "epoch": 0.74, "learning_rate": 3.762267983877751e-05, "loss": 0.0, "step": 38276 }, { "epoch": 0.74, "learning_rate": 3.762203287852027e-05, "loss": 0.0005, "step": 38278 }, { "epoch": 0.74, "learning_rate": 3.762138591826304e-05, "loss": 0.0005, "step": 38280 }, { "epoch": 0.74, "learning_rate": 3.762073895800581e-05, "loss": 0.0, "step": 38282 }, { "epoch": 0.74, "learning_rate": 3.762009199774858e-05, "loss": 0.0001, "step": 38284 }, { "epoch": 0.74, "learning_rate": 3.761944503749135e-05, "loss": 0.0001, "step": 38286 }, { "epoch": 0.74, "learning_rate": 3.761879807723412e-05, "loss": 0.0001, "step": 38288 }, { "epoch": 0.74, "learning_rate": 3.761815111697689e-05, "loss": 0.0, "step": 38290 }, { "epoch": 0.74, "learning_rate": 3.761750415671965e-05, "loss": 0.0054, "step": 38292 }, { "epoch": 0.74, "learning_rate": 3.7616857196462426e-05, "loss": 0.0015, "step": 38294 }, { "epoch": 0.74, "learning_rate": 3.761621023620519e-05, "loss": 0.0, "step": 38296 }, { "epoch": 0.74, "learning_rate": 3.7615563275947964e-05, "loss": 0.0, "step": 38298 }, { "epoch": 0.74, "learning_rate": 3.7614916315690726e-05, "loss": 0.0002, "step": 38300 }, { "epoch": 0.74, "learning_rate": 3.7614269355433495e-05, "loss": 0.0081, "step": 38302 }, { "epoch": 0.74, "learning_rate": 3.761362239517627e-05, "loss": 0.0008, "step": 38304 }, { "epoch": 0.74, "learning_rate": 3.7612975434919033e-05, "loss": 0.0001, "step": 38306 }, { "epoch": 0.74, "learning_rate": 3.76123284746618e-05, "loss": 0.0046, "step": 38308 }, { "epoch": 0.74, "learning_rate": 3.761168151440457e-05, "loss": 0.0001, "step": 38310 }, { "epoch": 0.74, "learning_rate": 3.761103455414734e-05, "loss": 0.0, "step": 38312 }, { "epoch": 0.74, "learning_rate": 3.761038759389011e-05, "loss": 0.0, "step": 38314 }, { "epoch": 0.74, "learning_rate": 3.760974063363288e-05, "loss": 0.0, "step": 38316 }, { "epoch": 0.74, "learning_rate": 3.760909367337565e-05, "loss": 0.0, "step": 38318 }, { "epoch": 0.74, "learning_rate": 3.760844671311842e-05, "loss": 0.0042, "step": 38320 }, { "epoch": 0.74, "learning_rate": 3.7607799752861186e-05, "loss": 0.0005, "step": 38322 }, { "epoch": 0.74, "learning_rate": 3.760715279260395e-05, "loss": 0.0001, "step": 38324 }, { "epoch": 0.74, "learning_rate": 3.7606505832346724e-05, "loss": 0.0035, "step": 38326 }, { "epoch": 0.74, "learning_rate": 3.7605858872089487e-05, "loss": 0.0086, "step": 38328 }, { "epoch": 0.74, "learning_rate": 3.7605211911832256e-05, "loss": 0.0007, "step": 38330 }, { "epoch": 0.74, "learning_rate": 3.7604564951575025e-05, "loss": 0.0071, "step": 38332 }, { "epoch": 0.74, "learning_rate": 3.7603917991317794e-05, "loss": 0.0, "step": 38334 }, { "epoch": 0.74, "learning_rate": 3.760327103106056e-05, "loss": 0.0001, "step": 38336 }, { "epoch": 0.74, "learning_rate": 3.760262407080333e-05, "loss": 0.0, "step": 38338 }, { "epoch": 0.74, "learning_rate": 3.76019771105461e-05, "loss": 0.0006, "step": 38340 }, { "epoch": 0.74, "learning_rate": 3.760133015028887e-05, "loss": 0.0, "step": 38342 }, { "epoch": 0.74, "learning_rate": 3.760068319003164e-05, "loss": 0.0, "step": 38344 }, { "epoch": 0.74, "learning_rate": 3.76000362297744e-05, "loss": 0.0, "step": 38346 }, { "epoch": 0.74, "learning_rate": 3.759938926951718e-05, "loss": 0.0, "step": 38348 }, { "epoch": 0.74, "learning_rate": 3.7598742309259947e-05, "loss": 0.0016, "step": 38350 }, { "epoch": 0.74, "learning_rate": 3.759809534900271e-05, "loss": 0.0, "step": 38352 }, { "epoch": 0.74, "learning_rate": 3.7597448388745485e-05, "loss": 0.0001, "step": 38354 }, { "epoch": 0.74, "learning_rate": 3.759680142848825e-05, "loss": 0.0002, "step": 38356 }, { "epoch": 0.74, "learning_rate": 3.759615446823102e-05, "loss": 0.0001, "step": 38358 }, { "epoch": 0.74, "learning_rate": 3.7595507507973785e-05, "loss": 0.0003, "step": 38360 }, { "epoch": 0.74, "learning_rate": 3.7594860547716554e-05, "loss": 0.0001, "step": 38362 }, { "epoch": 0.74, "learning_rate": 3.759421358745932e-05, "loss": 0.0001, "step": 38364 }, { "epoch": 0.74, "learning_rate": 3.759356662720209e-05, "loss": 0.0002, "step": 38366 }, { "epoch": 0.74, "learning_rate": 3.759291966694486e-05, "loss": 0.0, "step": 38368 }, { "epoch": 0.74, "learning_rate": 3.759227270668763e-05, "loss": 0.0129, "step": 38370 }, { "epoch": 0.74, "learning_rate": 3.75916257464304e-05, "loss": 0.0001, "step": 38372 }, { "epoch": 0.74, "learning_rate": 3.759097878617317e-05, "loss": 0.0003, "step": 38374 }, { "epoch": 0.74, "learning_rate": 3.759033182591594e-05, "loss": 0.0001, "step": 38376 }, { "epoch": 0.74, "learning_rate": 3.75896848656587e-05, "loss": 0.0, "step": 38378 }, { "epoch": 0.74, "learning_rate": 3.7589037905401476e-05, "loss": 0.0001, "step": 38380 }, { "epoch": 0.74, "learning_rate": 3.7588390945144245e-05, "loss": 0.0001, "step": 38382 }, { "epoch": 0.74, "learning_rate": 3.758774398488701e-05, "loss": 0.0, "step": 38384 }, { "epoch": 0.75, "learning_rate": 3.758709702462978e-05, "loss": 0.0, "step": 38386 }, { "epoch": 0.75, "learning_rate": 3.7586450064372546e-05, "loss": 0.0019, "step": 38388 }, { "epoch": 0.75, "learning_rate": 3.7585803104115315e-05, "loss": 0.0, "step": 38390 }, { "epoch": 0.75, "learning_rate": 3.7585156143858084e-05, "loss": 0.0, "step": 38392 }, { "epoch": 0.75, "learning_rate": 3.758450918360085e-05, "loss": 0.0, "step": 38394 }, { "epoch": 0.75, "learning_rate": 3.758386222334362e-05, "loss": 0.017, "step": 38396 }, { "epoch": 0.75, "learning_rate": 3.758321526308639e-05, "loss": 0.0001, "step": 38398 }, { "epoch": 0.75, "learning_rate": 3.758256830282916e-05, "loss": 0.0, "step": 38400 }, { "epoch": 0.75, "learning_rate": 3.758192134257193e-05, "loss": 0.017, "step": 38402 }, { "epoch": 0.75, "learning_rate": 3.75812743823147e-05, "loss": 0.0, "step": 38404 }, { "epoch": 0.75, "learning_rate": 3.758062742205746e-05, "loss": 0.0018, "step": 38406 }, { "epoch": 0.75, "learning_rate": 3.7579980461800236e-05, "loss": 0.0002, "step": 38408 }, { "epoch": 0.75, "learning_rate": 3.7579333501543e-05, "loss": 0.0, "step": 38410 }, { "epoch": 0.75, "learning_rate": 3.757868654128577e-05, "loss": 0.0, "step": 38412 }, { "epoch": 0.75, "learning_rate": 3.757803958102854e-05, "loss": 0.0, "step": 38414 }, { "epoch": 0.75, "learning_rate": 3.7577392620771306e-05, "loss": 0.0, "step": 38416 }, { "epoch": 0.75, "learning_rate": 3.757674566051408e-05, "loss": 0.0, "step": 38418 }, { "epoch": 0.75, "learning_rate": 3.7576098700256844e-05, "loss": 0.0, "step": 38420 }, { "epoch": 0.75, "learning_rate": 3.757545173999961e-05, "loss": 0.015, "step": 38422 }, { "epoch": 0.75, "learning_rate": 3.757480477974238e-05, "loss": 0.0, "step": 38424 }, { "epoch": 0.75, "learning_rate": 3.757415781948515e-05, "loss": 0.0001, "step": 38426 }, { "epoch": 0.75, "learning_rate": 3.757351085922792e-05, "loss": 0.0026, "step": 38428 }, { "epoch": 0.75, "learning_rate": 3.757286389897069e-05, "loss": 0.0046, "step": 38430 }, { "epoch": 0.75, "learning_rate": 3.757221693871346e-05, "loss": 0.0007, "step": 38432 }, { "epoch": 0.75, "learning_rate": 3.757156997845622e-05, "loss": 0.0, "step": 38434 }, { "epoch": 0.75, "learning_rate": 3.7570923018199e-05, "loss": 0.0, "step": 38436 }, { "epoch": 0.75, "learning_rate": 3.757027605794176e-05, "loss": 0.0002, "step": 38438 }, { "epoch": 0.75, "learning_rate": 3.7569629097684535e-05, "loss": 0.0001, "step": 38440 }, { "epoch": 0.75, "learning_rate": 3.75689821374273e-05, "loss": 0.005, "step": 38442 }, { "epoch": 0.75, "learning_rate": 3.7568335177170066e-05, "loss": 0.0001, "step": 38444 }, { "epoch": 0.75, "learning_rate": 3.7567688216912835e-05, "loss": 0.0, "step": 38446 }, { "epoch": 0.75, "learning_rate": 3.7567041256655605e-05, "loss": 0.0077, "step": 38448 }, { "epoch": 0.75, "learning_rate": 3.7566394296398374e-05, "loss": 0.0012, "step": 38450 }, { "epoch": 0.75, "learning_rate": 3.756574733614114e-05, "loss": 0.0, "step": 38452 }, { "epoch": 0.75, "learning_rate": 3.756510037588391e-05, "loss": 0.0024, "step": 38454 }, { "epoch": 0.75, "learning_rate": 3.756445341562668e-05, "loss": 0.0039, "step": 38456 }, { "epoch": 0.75, "learning_rate": 3.756380645536945e-05, "loss": 0.0007, "step": 38458 }, { "epoch": 0.75, "learning_rate": 3.756315949511221e-05, "loss": 0.0, "step": 38460 }, { "epoch": 0.75, "learning_rate": 3.756251253485499e-05, "loss": 0.009, "step": 38462 }, { "epoch": 0.75, "learning_rate": 3.756186557459776e-05, "loss": 0.0003, "step": 38464 }, { "epoch": 0.75, "learning_rate": 3.756121861434052e-05, "loss": 0.0001, "step": 38466 }, { "epoch": 0.75, "learning_rate": 3.7560571654083295e-05, "loss": 0.0019, "step": 38468 }, { "epoch": 0.75, "learning_rate": 3.755992469382606e-05, "loss": 0.0, "step": 38470 }, { "epoch": 0.75, "learning_rate": 3.755927773356883e-05, "loss": 0.0008, "step": 38472 }, { "epoch": 0.75, "learning_rate": 3.7558630773311596e-05, "loss": 0.0009, "step": 38474 }, { "epoch": 0.75, "learning_rate": 3.7557983813054365e-05, "loss": 0.0006, "step": 38476 }, { "epoch": 0.75, "learning_rate": 3.7557336852797134e-05, "loss": 0.0023, "step": 38478 }, { "epoch": 0.75, "learning_rate": 3.75566898925399e-05, "loss": 0.008, "step": 38480 }, { "epoch": 0.75, "learning_rate": 3.755604293228267e-05, "loss": 0.0002, "step": 38482 }, { "epoch": 0.75, "learning_rate": 3.755539597202544e-05, "loss": 0.0, "step": 38484 }, { "epoch": 0.75, "learning_rate": 3.755474901176821e-05, "loss": 0.0007, "step": 38486 }, { "epoch": 0.75, "learning_rate": 3.755410205151097e-05, "loss": 0.0, "step": 38488 }, { "epoch": 0.75, "learning_rate": 3.755345509125375e-05, "loss": 0.0, "step": 38490 }, { "epoch": 0.75, "learning_rate": 3.755280813099651e-05, "loss": 0.0013, "step": 38492 }, { "epoch": 0.75, "learning_rate": 3.755216117073928e-05, "loss": 0.0011, "step": 38494 }, { "epoch": 0.75, "learning_rate": 3.7551514210482056e-05, "loss": 0.0001, "step": 38496 }, { "epoch": 0.75, "learning_rate": 3.755086725022482e-05, "loss": 0.0, "step": 38498 }, { "epoch": 0.75, "learning_rate": 3.7550220289967594e-05, "loss": 0.0, "step": 38500 }, { "epoch": 0.75, "learning_rate": 3.7549573329710356e-05, "loss": 0.0003, "step": 38502 }, { "epoch": 0.75, "learning_rate": 3.7548926369453125e-05, "loss": 0.0, "step": 38504 }, { "epoch": 0.75, "learning_rate": 3.7548279409195894e-05, "loss": 0.0008, "step": 38506 }, { "epoch": 0.75, "learning_rate": 3.7547632448938664e-05, "loss": 0.0, "step": 38508 }, { "epoch": 0.75, "learning_rate": 3.754698548868143e-05, "loss": 0.0001, "step": 38510 }, { "epoch": 0.75, "learning_rate": 3.75463385284242e-05, "loss": 0.0, "step": 38512 }, { "epoch": 0.75, "learning_rate": 3.754569156816697e-05, "loss": 0.0001, "step": 38514 }, { "epoch": 0.75, "learning_rate": 3.754504460790974e-05, "loss": 0.0011, "step": 38516 }, { "epoch": 0.75, "learning_rate": 3.754439764765251e-05, "loss": 0.0034, "step": 38518 }, { "epoch": 0.75, "learning_rate": 3.754375068739527e-05, "loss": 0.0019, "step": 38520 }, { "epoch": 0.75, "learning_rate": 3.754310372713805e-05, "loss": 0.0003, "step": 38522 }, { "epoch": 0.75, "learning_rate": 3.754245676688081e-05, "loss": 0.0, "step": 38524 }, { "epoch": 0.75, "learning_rate": 3.754180980662358e-05, "loss": 0.0083, "step": 38526 }, { "epoch": 0.75, "learning_rate": 3.7541162846366354e-05, "loss": 0.0001, "step": 38528 }, { "epoch": 0.75, "learning_rate": 3.754051588610912e-05, "loss": 0.0003, "step": 38530 }, { "epoch": 0.75, "learning_rate": 3.7539868925851886e-05, "loss": 0.0007, "step": 38532 }, { "epoch": 0.75, "learning_rate": 3.7539221965594655e-05, "loss": 0.0, "step": 38534 }, { "epoch": 0.75, "learning_rate": 3.7538575005337424e-05, "loss": 0.0004, "step": 38536 }, { "epoch": 0.75, "learning_rate": 3.753792804508019e-05, "loss": 0.0, "step": 38538 }, { "epoch": 0.75, "learning_rate": 3.753728108482296e-05, "loss": 0.0001, "step": 38540 }, { "epoch": 0.75, "learning_rate": 3.753663412456573e-05, "loss": 0.0, "step": 38542 }, { "epoch": 0.75, "learning_rate": 3.75359871643085e-05, "loss": 0.0009, "step": 38544 }, { "epoch": 0.75, "learning_rate": 3.753534020405127e-05, "loss": 0.0006, "step": 38546 }, { "epoch": 0.75, "learning_rate": 3.753469324379403e-05, "loss": 0.0002, "step": 38548 }, { "epoch": 0.75, "learning_rate": 3.753404628353681e-05, "loss": 0.0031, "step": 38550 }, { "epoch": 0.75, "learning_rate": 3.753339932327957e-05, "loss": 0.0, "step": 38552 }, { "epoch": 0.75, "learning_rate": 3.753275236302234e-05, "loss": 0.0002, "step": 38554 }, { "epoch": 0.75, "learning_rate": 3.753210540276511e-05, "loss": 0.0, "step": 38556 }, { "epoch": 0.75, "learning_rate": 3.753145844250788e-05, "loss": 0.0059, "step": 38558 }, { "epoch": 0.75, "learning_rate": 3.753081148225065e-05, "loss": 0.0283, "step": 38560 }, { "epoch": 0.75, "learning_rate": 3.7530164521993415e-05, "loss": 0.0001, "step": 38562 }, { "epoch": 0.75, "learning_rate": 3.7529517561736184e-05, "loss": 0.0, "step": 38564 }, { "epoch": 0.75, "learning_rate": 3.7528870601478953e-05, "loss": 0.0, "step": 38566 }, { "epoch": 0.75, "learning_rate": 3.752822364122172e-05, "loss": 0.0002, "step": 38568 }, { "epoch": 0.75, "learning_rate": 3.7527576680964485e-05, "loss": 0.0026, "step": 38570 }, { "epoch": 0.75, "learning_rate": 3.752692972070726e-05, "loss": 0.0, "step": 38572 }, { "epoch": 0.75, "learning_rate": 3.752628276045003e-05, "loss": 0.0, "step": 38574 }, { "epoch": 0.75, "learning_rate": 3.752563580019279e-05, "loss": 0.0, "step": 38576 }, { "epoch": 0.75, "learning_rate": 3.752498883993557e-05, "loss": 0.0, "step": 38578 }, { "epoch": 0.75, "learning_rate": 3.752434187967833e-05, "loss": 0.0202, "step": 38580 }, { "epoch": 0.75, "learning_rate": 3.7523694919421106e-05, "loss": 0.0, "step": 38582 }, { "epoch": 0.75, "learning_rate": 3.752304795916387e-05, "loss": 0.0002, "step": 38584 }, { "epoch": 0.75, "learning_rate": 3.752240099890664e-05, "loss": 0.0072, "step": 38586 }, { "epoch": 0.75, "learning_rate": 3.7521754038649407e-05, "loss": 0.0, "step": 38588 }, { "epoch": 0.75, "learning_rate": 3.7521107078392176e-05, "loss": 0.0003, "step": 38590 }, { "epoch": 0.75, "learning_rate": 3.7520460118134945e-05, "loss": 0.0025, "step": 38592 }, { "epoch": 0.75, "learning_rate": 3.7519813157877714e-05, "loss": 0.0, "step": 38594 }, { "epoch": 0.75, "learning_rate": 3.751916619762048e-05, "loss": 0.0001, "step": 38596 }, { "epoch": 0.75, "learning_rate": 3.751851923736325e-05, "loss": 0.0013, "step": 38598 }, { "epoch": 0.75, "learning_rate": 3.751787227710602e-05, "loss": 0.0, "step": 38600 }, { "epoch": 0.75, "learning_rate": 3.7517225316848783e-05, "loss": 0.0024, "step": 38602 }, { "epoch": 0.75, "learning_rate": 3.751657835659156e-05, "loss": 0.0281, "step": 38604 }, { "epoch": 0.75, "learning_rate": 3.751593139633433e-05, "loss": 0.0181, "step": 38606 }, { "epoch": 0.75, "learning_rate": 3.751528443607709e-05, "loss": 0.0015, "step": 38608 }, { "epoch": 0.75, "learning_rate": 3.7514637475819867e-05, "loss": 0.0038, "step": 38610 }, { "epoch": 0.75, "learning_rate": 3.751399051556263e-05, "loss": 0.0011, "step": 38612 }, { "epoch": 0.75, "learning_rate": 3.75133435553054e-05, "loss": 0.0, "step": 38614 }, { "epoch": 0.75, "learning_rate": 3.751269659504817e-05, "loss": 0.0003, "step": 38616 }, { "epoch": 0.75, "learning_rate": 3.7512049634790936e-05, "loss": 0.0003, "step": 38618 }, { "epoch": 0.75, "learning_rate": 3.7511402674533705e-05, "loss": 0.0014, "step": 38620 }, { "epoch": 0.75, "learning_rate": 3.7510755714276474e-05, "loss": 0.0041, "step": 38622 }, { "epoch": 0.75, "learning_rate": 3.751010875401924e-05, "loss": 0.0004, "step": 38624 }, { "epoch": 0.75, "learning_rate": 3.750946179376201e-05, "loss": 0.0, "step": 38626 }, { "epoch": 0.75, "learning_rate": 3.750881483350478e-05, "loss": 0.0001, "step": 38628 }, { "epoch": 0.75, "learning_rate": 3.7508167873247544e-05, "loss": 0.0001, "step": 38630 }, { "epoch": 0.75, "learning_rate": 3.750752091299032e-05, "loss": 0.0065, "step": 38632 }, { "epoch": 0.75, "learning_rate": 3.750687395273308e-05, "loss": 0.0003, "step": 38634 }, { "epoch": 0.75, "learning_rate": 3.750622699247585e-05, "loss": 0.0006, "step": 38636 }, { "epoch": 0.75, "learning_rate": 3.750558003221862e-05, "loss": 0.005, "step": 38638 }, { "epoch": 0.75, "learning_rate": 3.750493307196139e-05, "loss": 0.0001, "step": 38640 }, { "epoch": 0.75, "learning_rate": 3.7504286111704165e-05, "loss": 0.0001, "step": 38642 }, { "epoch": 0.75, "learning_rate": 3.750363915144693e-05, "loss": 0.0002, "step": 38644 }, { "epoch": 0.75, "learning_rate": 3.7502992191189696e-05, "loss": 0.0, "step": 38646 }, { "epoch": 0.75, "learning_rate": 3.7502345230932466e-05, "loss": 0.0, "step": 38648 }, { "epoch": 0.75, "learning_rate": 3.7501698270675235e-05, "loss": 0.0013, "step": 38650 }, { "epoch": 0.75, "learning_rate": 3.7501051310418004e-05, "loss": 0.0, "step": 38652 }, { "epoch": 0.75, "learning_rate": 3.750040435016077e-05, "loss": 0.0001, "step": 38654 }, { "epoch": 0.75, "learning_rate": 3.749975738990354e-05, "loss": 0.0003, "step": 38656 }, { "epoch": 0.75, "learning_rate": 3.749911042964631e-05, "loss": 0.0, "step": 38658 }, { "epoch": 0.75, "learning_rate": 3.749846346938908e-05, "loss": 0.0144, "step": 38660 }, { "epoch": 0.75, "learning_rate": 3.749781650913184e-05, "loss": 0.0002, "step": 38662 }, { "epoch": 0.75, "learning_rate": 3.749716954887462e-05, "loss": 0.0, "step": 38664 }, { "epoch": 0.75, "learning_rate": 3.749652258861738e-05, "loss": 0.0077, "step": 38666 }, { "epoch": 0.75, "learning_rate": 3.749587562836015e-05, "loss": 0.0, "step": 38668 }, { "epoch": 0.75, "learning_rate": 3.749522866810292e-05, "loss": 0.0001, "step": 38670 }, { "epoch": 0.75, "learning_rate": 3.749458170784569e-05, "loss": 0.0001, "step": 38672 }, { "epoch": 0.75, "learning_rate": 3.749393474758846e-05, "loss": 0.0001, "step": 38674 }, { "epoch": 0.75, "learning_rate": 3.7493287787331226e-05, "loss": 0.0, "step": 38676 }, { "epoch": 0.75, "learning_rate": 3.7492640827073995e-05, "loss": 0.0039, "step": 38678 }, { "epoch": 0.75, "learning_rate": 3.7491993866816764e-05, "loss": 0.0, "step": 38680 }, { "epoch": 0.75, "learning_rate": 3.749134690655953e-05, "loss": 0.0004, "step": 38682 }, { "epoch": 0.75, "learning_rate": 3.74906999463023e-05, "loss": 0.0, "step": 38684 }, { "epoch": 0.75, "learning_rate": 3.749005298604507e-05, "loss": 0.0054, "step": 38686 }, { "epoch": 0.75, "learning_rate": 3.748940602578784e-05, "loss": 0.0002, "step": 38688 }, { "epoch": 0.75, "learning_rate": 3.74887590655306e-05, "loss": 0.0, "step": 38690 }, { "epoch": 0.75, "learning_rate": 3.748811210527338e-05, "loss": 0.0026, "step": 38692 }, { "epoch": 0.75, "learning_rate": 3.748746514501614e-05, "loss": 0.0003, "step": 38694 }, { "epoch": 0.75, "learning_rate": 3.748681818475891e-05, "loss": 0.0001, "step": 38696 }, { "epoch": 0.75, "learning_rate": 3.748617122450168e-05, "loss": 0.0001, "step": 38698 }, { "epoch": 0.75, "learning_rate": 3.748552426424445e-05, "loss": 0.0009, "step": 38700 }, { "epoch": 0.75, "learning_rate": 3.748487730398722e-05, "loss": 0.0, "step": 38702 }, { "epoch": 0.75, "learning_rate": 3.7484230343729986e-05, "loss": 0.0007, "step": 38704 }, { "epoch": 0.75, "learning_rate": 3.7483583383472755e-05, "loss": 0.0, "step": 38706 }, { "epoch": 0.75, "learning_rate": 3.7482936423215525e-05, "loss": 0.0012, "step": 38708 }, { "epoch": 0.75, "learning_rate": 3.7482289462958294e-05, "loss": 0.0001, "step": 38710 }, { "epoch": 0.75, "learning_rate": 3.7481642502701056e-05, "loss": 0.0001, "step": 38712 }, { "epoch": 0.75, "learning_rate": 3.748099554244383e-05, "loss": 0.0, "step": 38714 }, { "epoch": 0.75, "learning_rate": 3.7480348582186594e-05, "loss": 0.0016, "step": 38716 }, { "epoch": 0.75, "learning_rate": 3.747970162192936e-05, "loss": 0.0001, "step": 38718 }, { "epoch": 0.75, "learning_rate": 3.747905466167214e-05, "loss": 0.0081, "step": 38720 }, { "epoch": 0.75, "learning_rate": 3.74784077014149e-05, "loss": 0.0, "step": 38722 }, { "epoch": 0.75, "learning_rate": 3.747776074115768e-05, "loss": 0.0004, "step": 38724 }, { "epoch": 0.75, "learning_rate": 3.747711378090044e-05, "loss": 0.0104, "step": 38726 }, { "epoch": 0.75, "learning_rate": 3.747646682064321e-05, "loss": 0.0129, "step": 38728 }, { "epoch": 0.75, "learning_rate": 3.747581986038598e-05, "loss": 0.0047, "step": 38730 }, { "epoch": 0.75, "learning_rate": 3.747517290012875e-05, "loss": 0.0, "step": 38732 }, { "epoch": 0.75, "learning_rate": 3.7474525939871516e-05, "loss": 0.0, "step": 38734 }, { "epoch": 0.75, "learning_rate": 3.7473878979614285e-05, "loss": 0.0055, "step": 38736 }, { "epoch": 0.75, "learning_rate": 3.7473232019357054e-05, "loss": 0.0002, "step": 38738 }, { "epoch": 0.75, "learning_rate": 3.747258505909982e-05, "loss": 0.0, "step": 38740 }, { "epoch": 0.75, "learning_rate": 3.747193809884259e-05, "loss": 0.0063, "step": 38742 }, { "epoch": 0.75, "learning_rate": 3.7471291138585355e-05, "loss": 0.0013, "step": 38744 }, { "epoch": 0.75, "learning_rate": 3.747064417832813e-05, "loss": 0.0, "step": 38746 }, { "epoch": 0.75, "learning_rate": 3.746999721807089e-05, "loss": 0.0, "step": 38748 }, { "epoch": 0.75, "learning_rate": 3.746935025781366e-05, "loss": 0.0, "step": 38750 }, { "epoch": 0.75, "learning_rate": 3.746870329755644e-05, "loss": 0.0, "step": 38752 }, { "epoch": 0.75, "learning_rate": 3.74680563372992e-05, "loss": 0.0001, "step": 38754 }, { "epoch": 0.75, "learning_rate": 3.746740937704197e-05, "loss": 0.0, "step": 38756 }, { "epoch": 0.75, "learning_rate": 3.746676241678474e-05, "loss": 0.0002, "step": 38758 }, { "epoch": 0.75, "learning_rate": 3.746611545652751e-05, "loss": 0.0, "step": 38760 }, { "epoch": 0.75, "learning_rate": 3.7465468496270276e-05, "loss": 0.0, "step": 38762 }, { "epoch": 0.75, "learning_rate": 3.7464821536013045e-05, "loss": 0.0005, "step": 38764 }, { "epoch": 0.75, "learning_rate": 3.7464174575755814e-05, "loss": 0.017, "step": 38766 }, { "epoch": 0.75, "learning_rate": 3.7463527615498584e-05, "loss": 0.0, "step": 38768 }, { "epoch": 0.75, "learning_rate": 3.746288065524135e-05, "loss": 0.0, "step": 38770 }, { "epoch": 0.75, "learning_rate": 3.7462233694984115e-05, "loss": 0.0001, "step": 38772 }, { "epoch": 0.75, "learning_rate": 3.746158673472689e-05, "loss": 0.0, "step": 38774 }, { "epoch": 0.75, "learning_rate": 3.746093977446965e-05, "loss": 0.0002, "step": 38776 }, { "epoch": 0.75, "learning_rate": 3.746029281421242e-05, "loss": 0.0, "step": 38778 }, { "epoch": 0.75, "learning_rate": 3.745964585395519e-05, "loss": 0.0, "step": 38780 }, { "epoch": 0.75, "learning_rate": 3.745899889369796e-05, "loss": 0.0009, "step": 38782 }, { "epoch": 0.75, "learning_rate": 3.7458351933440736e-05, "loss": 0.0002, "step": 38784 }, { "epoch": 0.75, "learning_rate": 3.74577049731835e-05, "loss": 0.0, "step": 38786 }, { "epoch": 0.75, "learning_rate": 3.745705801292627e-05, "loss": 0.0001, "step": 38788 }, { "epoch": 0.75, "learning_rate": 3.745641105266904e-05, "loss": 0.0009, "step": 38790 }, { "epoch": 0.75, "learning_rate": 3.7455764092411806e-05, "loss": 0.0, "step": 38792 }, { "epoch": 0.75, "learning_rate": 3.745511713215457e-05, "loss": 0.0001, "step": 38794 }, { "epoch": 0.75, "learning_rate": 3.7454470171897344e-05, "loss": 0.0, "step": 38796 }, { "epoch": 0.75, "learning_rate": 3.745382321164011e-05, "loss": 0.0001, "step": 38798 }, { "epoch": 0.75, "learning_rate": 3.745317625138288e-05, "loss": 0.0001, "step": 38800 }, { "epoch": 0.75, "learning_rate": 3.745252929112565e-05, "loss": 0.0, "step": 38802 }, { "epoch": 0.75, "learning_rate": 3.7451882330868413e-05, "loss": 0.0, "step": 38804 }, { "epoch": 0.75, "learning_rate": 3.745123537061119e-05, "loss": 0.0, "step": 38806 }, { "epoch": 0.75, "learning_rate": 3.745058841035395e-05, "loss": 0.0004, "step": 38808 }, { "epoch": 0.75, "learning_rate": 3.744994145009672e-05, "loss": 0.0, "step": 38810 }, { "epoch": 0.75, "learning_rate": 3.744929448983949e-05, "loss": 0.0, "step": 38812 }, { "epoch": 0.75, "learning_rate": 3.744864752958226e-05, "loss": 0.0059, "step": 38814 }, { "epoch": 0.75, "learning_rate": 3.744800056932503e-05, "loss": 0.0, "step": 38816 }, { "epoch": 0.75, "learning_rate": 3.74473536090678e-05, "loss": 0.0011, "step": 38818 }, { "epoch": 0.75, "learning_rate": 3.7446706648810566e-05, "loss": 0.0, "step": 38820 }, { "epoch": 0.75, "learning_rate": 3.7446059688553335e-05, "loss": 0.0, "step": 38822 }, { "epoch": 0.75, "learning_rate": 3.7445412728296104e-05, "loss": 0.0, "step": 38824 }, { "epoch": 0.75, "learning_rate": 3.744476576803887e-05, "loss": 0.0, "step": 38826 }, { "epoch": 0.75, "learning_rate": 3.744411880778164e-05, "loss": 0.0, "step": 38828 }, { "epoch": 0.75, "learning_rate": 3.744347184752441e-05, "loss": 0.0, "step": 38830 }, { "epoch": 0.75, "learning_rate": 3.7442824887267174e-05, "loss": 0.0009, "step": 38832 }, { "epoch": 0.75, "learning_rate": 3.744217792700995e-05, "loss": 0.0007, "step": 38834 }, { "epoch": 0.75, "learning_rate": 3.744153096675271e-05, "loss": 0.0, "step": 38836 }, { "epoch": 0.75, "learning_rate": 3.744088400649548e-05, "loss": 0.0, "step": 38838 }, { "epoch": 0.75, "learning_rate": 3.744023704623825e-05, "loss": 0.0, "step": 38840 }, { "epoch": 0.75, "learning_rate": 3.743959008598102e-05, "loss": 0.0004, "step": 38842 }, { "epoch": 0.75, "learning_rate": 3.743894312572379e-05, "loss": 0.0016, "step": 38844 }, { "epoch": 0.75, "learning_rate": 3.743829616546656e-05, "loss": 0.001, "step": 38846 }, { "epoch": 0.75, "learning_rate": 3.7437649205209327e-05, "loss": 0.0, "step": 38848 }, { "epoch": 0.75, "learning_rate": 3.7437002244952096e-05, "loss": 0.0131, "step": 38850 }, { "epoch": 0.75, "learning_rate": 3.7436355284694865e-05, "loss": 0.0001, "step": 38852 }, { "epoch": 0.75, "learning_rate": 3.743570832443763e-05, "loss": 0.0047, "step": 38854 }, { "epoch": 0.75, "learning_rate": 3.74350613641804e-05, "loss": 0.0112, "step": 38856 }, { "epoch": 0.75, "learning_rate": 3.7434414403923165e-05, "loss": 0.0046, "step": 38858 }, { "epoch": 0.75, "learning_rate": 3.7433767443665934e-05, "loss": 0.0, "step": 38860 }, { "epoch": 0.75, "learning_rate": 3.743312048340871e-05, "loss": 0.0001, "step": 38862 }, { "epoch": 0.75, "learning_rate": 3.743247352315147e-05, "loss": 0.0032, "step": 38864 }, { "epoch": 0.75, "learning_rate": 3.743182656289425e-05, "loss": 0.0, "step": 38866 }, { "epoch": 0.75, "learning_rate": 3.743117960263701e-05, "loss": 0.0, "step": 38868 }, { "epoch": 0.75, "learning_rate": 3.743053264237978e-05, "loss": 0.0001, "step": 38870 }, { "epoch": 0.75, "learning_rate": 3.742988568212255e-05, "loss": 0.0, "step": 38872 }, { "epoch": 0.75, "learning_rate": 3.742923872186532e-05, "loss": 0.0, "step": 38874 }, { "epoch": 0.75, "learning_rate": 3.742859176160809e-05, "loss": 0.0004, "step": 38876 }, { "epoch": 0.75, "learning_rate": 3.7427944801350856e-05, "loss": 0.0027, "step": 38878 }, { "epoch": 0.75, "learning_rate": 3.7427297841093625e-05, "loss": 0.0034, "step": 38880 }, { "epoch": 0.75, "learning_rate": 3.7426650880836394e-05, "loss": 0.0054, "step": 38882 }, { "epoch": 0.75, "learning_rate": 3.742600392057916e-05, "loss": 0.0165, "step": 38884 }, { "epoch": 0.75, "learning_rate": 3.7425356960321926e-05, "loss": 0.0917, "step": 38886 }, { "epoch": 0.75, "learning_rate": 3.74247100000647e-05, "loss": 0.0, "step": 38888 }, { "epoch": 0.75, "learning_rate": 3.7424063039807464e-05, "loss": 0.0, "step": 38890 }, { "epoch": 0.75, "learning_rate": 3.742341607955023e-05, "loss": 0.0, "step": 38892 }, { "epoch": 0.75, "learning_rate": 3.7422769119293e-05, "loss": 0.0117, "step": 38894 }, { "epoch": 0.75, "learning_rate": 3.742212215903577e-05, "loss": 0.0004, "step": 38896 }, { "epoch": 0.75, "learning_rate": 3.742147519877854e-05, "loss": 0.0, "step": 38898 }, { "epoch": 0.76, "learning_rate": 3.742082823852131e-05, "loss": 0.0, "step": 38900 }, { "epoch": 0.76, "learning_rate": 3.742018127826408e-05, "loss": 0.0002, "step": 38902 }, { "epoch": 0.76, "learning_rate": 3.741953431800685e-05, "loss": 0.0, "step": 38904 }, { "epoch": 0.76, "learning_rate": 3.7418887357749616e-05, "loss": 0.0001, "step": 38906 }, { "epoch": 0.76, "learning_rate": 3.7418240397492386e-05, "loss": 0.0028, "step": 38908 }, { "epoch": 0.76, "learning_rate": 3.7417593437235155e-05, "loss": 0.0, "step": 38910 }, { "epoch": 0.76, "learning_rate": 3.7416946476977924e-05, "loss": 0.0003, "step": 38912 }, { "epoch": 0.76, "learning_rate": 3.7416299516720686e-05, "loss": 0.0, "step": 38914 }, { "epoch": 0.76, "learning_rate": 3.741565255646346e-05, "loss": 0.0, "step": 38916 }, { "epoch": 0.76, "learning_rate": 3.7415005596206224e-05, "loss": 0.001, "step": 38918 }, { "epoch": 0.76, "learning_rate": 3.741435863594899e-05, "loss": 0.0, "step": 38920 }, { "epoch": 0.76, "learning_rate": 3.741371167569176e-05, "loss": 0.0001, "step": 38922 }, { "epoch": 0.76, "learning_rate": 3.741306471543453e-05, "loss": 0.0, "step": 38924 }, { "epoch": 0.76, "learning_rate": 3.74124177551773e-05, "loss": 0.0002, "step": 38926 }, { "epoch": 0.76, "learning_rate": 3.741177079492007e-05, "loss": 0.0, "step": 38928 }, { "epoch": 0.76, "learning_rate": 3.741112383466284e-05, "loss": 0.0006, "step": 38930 }, { "epoch": 0.76, "learning_rate": 3.741047687440561e-05, "loss": 0.0, "step": 38932 }, { "epoch": 0.76, "learning_rate": 3.740982991414838e-05, "loss": 0.0, "step": 38934 }, { "epoch": 0.76, "learning_rate": 3.740918295389114e-05, "loss": 0.0, "step": 38936 }, { "epoch": 0.76, "learning_rate": 3.7408535993633915e-05, "loss": 0.0, "step": 38938 }, { "epoch": 0.76, "learning_rate": 3.740788903337668e-05, "loss": 0.0007, "step": 38940 }, { "epoch": 0.76, "learning_rate": 3.740724207311945e-05, "loss": 0.0002, "step": 38942 }, { "epoch": 0.76, "learning_rate": 3.740659511286222e-05, "loss": 0.0, "step": 38944 }, { "epoch": 0.76, "learning_rate": 3.7405948152604985e-05, "loss": 0.0009, "step": 38946 }, { "epoch": 0.76, "learning_rate": 3.740530119234776e-05, "loss": 0.0, "step": 38948 }, { "epoch": 0.76, "learning_rate": 3.740465423209052e-05, "loss": 0.0, "step": 38950 }, { "epoch": 0.76, "learning_rate": 3.740400727183329e-05, "loss": 0.0, "step": 38952 }, { "epoch": 0.76, "learning_rate": 3.740336031157606e-05, "loss": 0.0073, "step": 38954 }, { "epoch": 0.76, "learning_rate": 3.740271335131883e-05, "loss": 0.0009, "step": 38956 }, { "epoch": 0.76, "learning_rate": 3.74020663910616e-05, "loss": 0.0001, "step": 38958 }, { "epoch": 0.76, "learning_rate": 3.740141943080437e-05, "loss": 0.006, "step": 38960 }, { "epoch": 0.76, "learning_rate": 3.740077247054714e-05, "loss": 0.0084, "step": 38962 }, { "epoch": 0.76, "learning_rate": 3.7400125510289906e-05, "loss": 0.0373, "step": 38964 }, { "epoch": 0.76, "learning_rate": 3.7399478550032675e-05, "loss": 0.0001, "step": 38966 }, { "epoch": 0.76, "learning_rate": 3.739883158977544e-05, "loss": 0.0001, "step": 38968 }, { "epoch": 0.76, "learning_rate": 3.7398184629518214e-05, "loss": 0.0, "step": 38970 }, { "epoch": 0.76, "learning_rate": 3.7397537669260976e-05, "loss": 0.0599, "step": 38972 }, { "epoch": 0.76, "learning_rate": 3.7396890709003745e-05, "loss": 0.0, "step": 38974 }, { "epoch": 0.76, "learning_rate": 3.739624374874652e-05, "loss": 0.0001, "step": 38976 }, { "epoch": 0.76, "learning_rate": 3.739559678848928e-05, "loss": 0.0001, "step": 38978 }, { "epoch": 0.76, "learning_rate": 3.739494982823205e-05, "loss": 0.0005, "step": 38980 }, { "epoch": 0.76, "learning_rate": 3.739430286797482e-05, "loss": 0.0001, "step": 38982 }, { "epoch": 0.76, "learning_rate": 3.739365590771759e-05, "loss": 0.0, "step": 38984 }, { "epoch": 0.76, "learning_rate": 3.739300894746036e-05, "loss": 0.001, "step": 38986 }, { "epoch": 0.76, "learning_rate": 3.739236198720313e-05, "loss": 0.006, "step": 38988 }, { "epoch": 0.76, "learning_rate": 3.73917150269459e-05, "loss": 0.0001, "step": 38990 }, { "epoch": 0.76, "learning_rate": 3.739106806668867e-05, "loss": 0.0045, "step": 38992 }, { "epoch": 0.76, "learning_rate": 3.7390421106431436e-05, "loss": 0.0079, "step": 38994 }, { "epoch": 0.76, "learning_rate": 3.73897741461742e-05, "loss": 0.0052, "step": 38996 }, { "epoch": 0.76, "learning_rate": 3.7389127185916974e-05, "loss": 0.0016, "step": 38998 }, { "epoch": 0.76, "learning_rate": 3.7388480225659736e-05, "loss": 0.0, "step": 39000 }, { "epoch": 0.76, "learning_rate": 3.7387833265402505e-05, "loss": 0.0, "step": 39002 }, { "epoch": 0.76, "learning_rate": 3.7387186305145274e-05, "loss": 0.0003, "step": 39004 }, { "epoch": 0.76, "learning_rate": 3.7386539344888044e-05, "loss": 0.0001, "step": 39006 }, { "epoch": 0.76, "learning_rate": 3.738589238463082e-05, "loss": 0.0003, "step": 39008 }, { "epoch": 0.76, "learning_rate": 3.738524542437358e-05, "loss": 0.0, "step": 39010 }, { "epoch": 0.76, "learning_rate": 3.738459846411635e-05, "loss": 0.0001, "step": 39012 }, { "epoch": 0.76, "learning_rate": 3.738395150385912e-05, "loss": 0.0121, "step": 39014 }, { "epoch": 0.76, "learning_rate": 3.738330454360189e-05, "loss": 0.0004, "step": 39016 }, { "epoch": 0.76, "learning_rate": 3.738265758334465e-05, "loss": 0.0, "step": 39018 }, { "epoch": 0.76, "learning_rate": 3.738201062308743e-05, "loss": 0.0044, "step": 39020 }, { "epoch": 0.76, "learning_rate": 3.7381363662830196e-05, "loss": 0.0, "step": 39022 }, { "epoch": 0.76, "learning_rate": 3.7380716702572965e-05, "loss": 0.0001, "step": 39024 }, { "epoch": 0.76, "learning_rate": 3.7380069742315734e-05, "loss": 0.0, "step": 39026 }, { "epoch": 0.76, "learning_rate": 3.73794227820585e-05, "loss": 0.0, "step": 39028 }, { "epoch": 0.76, "learning_rate": 3.737877582180127e-05, "loss": 0.0, "step": 39030 }, { "epoch": 0.76, "learning_rate": 3.7378128861544035e-05, "loss": 0.0072, "step": 39032 }, { "epoch": 0.76, "learning_rate": 3.7377481901286804e-05, "loss": 0.0, "step": 39034 }, { "epoch": 0.76, "learning_rate": 3.737683494102957e-05, "loss": 0.0, "step": 39036 }, { "epoch": 0.76, "learning_rate": 3.737618798077234e-05, "loss": 0.0003, "step": 39038 }, { "epoch": 0.76, "learning_rate": 3.737554102051511e-05, "loss": 0.0, "step": 39040 }, { "epoch": 0.76, "learning_rate": 3.737489406025788e-05, "loss": 0.0008, "step": 39042 }, { "epoch": 0.76, "learning_rate": 3.737424710000065e-05, "loss": 0.0, "step": 39044 }, { "epoch": 0.76, "learning_rate": 3.737360013974342e-05, "loss": 0.0, "step": 39046 }, { "epoch": 0.76, "learning_rate": 3.737295317948619e-05, "loss": 0.0008, "step": 39048 }, { "epoch": 0.76, "learning_rate": 3.737230621922895e-05, "loss": 0.0001, "step": 39050 }, { "epoch": 0.76, "learning_rate": 3.7371659258971726e-05, "loss": 0.0, "step": 39052 }, { "epoch": 0.76, "learning_rate": 3.7371012298714495e-05, "loss": 0.0, "step": 39054 }, { "epoch": 0.76, "learning_rate": 3.737036533845726e-05, "loss": 0.0009, "step": 39056 }, { "epoch": 0.76, "learning_rate": 3.736971837820003e-05, "loss": 0.0001, "step": 39058 }, { "epoch": 0.76, "learning_rate": 3.7369071417942795e-05, "loss": 0.0, "step": 39060 }, { "epoch": 0.76, "learning_rate": 3.7368424457685564e-05, "loss": 0.0001, "step": 39062 }, { "epoch": 0.76, "learning_rate": 3.7367777497428333e-05, "loss": 0.002, "step": 39064 }, { "epoch": 0.76, "learning_rate": 3.73671305371711e-05, "loss": 0.0072, "step": 39066 }, { "epoch": 0.76, "learning_rate": 3.736648357691387e-05, "loss": 0.0001, "step": 39068 }, { "epoch": 0.76, "learning_rate": 3.736583661665664e-05, "loss": 0.0001, "step": 39070 }, { "epoch": 0.76, "learning_rate": 3.736518965639941e-05, "loss": 0.0026, "step": 39072 }, { "epoch": 0.76, "learning_rate": 3.736454269614218e-05, "loss": 0.0, "step": 39074 }, { "epoch": 0.76, "learning_rate": 3.736389573588495e-05, "loss": 0.0006, "step": 39076 }, { "epoch": 0.76, "learning_rate": 3.736324877562771e-05, "loss": 0.0002, "step": 39078 }, { "epoch": 0.76, "learning_rate": 3.7362601815370486e-05, "loss": 0.0006, "step": 39080 }, { "epoch": 0.76, "learning_rate": 3.736195485511325e-05, "loss": 0.0084, "step": 39082 }, { "epoch": 0.76, "learning_rate": 3.736130789485602e-05, "loss": 0.0011, "step": 39084 }, { "epoch": 0.76, "learning_rate": 3.7360660934598793e-05, "loss": 0.0, "step": 39086 }, { "epoch": 0.76, "learning_rate": 3.7360013974341556e-05, "loss": 0.0, "step": 39088 }, { "epoch": 0.76, "learning_rate": 3.735936701408433e-05, "loss": 0.0054, "step": 39090 }, { "epoch": 0.76, "learning_rate": 3.7358720053827094e-05, "loss": 0.0, "step": 39092 }, { "epoch": 0.76, "learning_rate": 3.735807309356986e-05, "loss": 0.0, "step": 39094 }, { "epoch": 0.76, "learning_rate": 3.735742613331263e-05, "loss": 0.0005, "step": 39096 }, { "epoch": 0.76, "learning_rate": 3.73567791730554e-05, "loss": 0.0, "step": 39098 }, { "epoch": 0.76, "learning_rate": 3.735613221279817e-05, "loss": 0.0002, "step": 39100 }, { "epoch": 0.76, "learning_rate": 3.735548525254094e-05, "loss": 0.0019, "step": 39102 }, { "epoch": 0.76, "learning_rate": 3.735483829228371e-05, "loss": 0.0001, "step": 39104 }, { "epoch": 0.76, "learning_rate": 3.735419133202648e-05, "loss": 0.0, "step": 39106 }, { "epoch": 0.76, "learning_rate": 3.7353544371769247e-05, "loss": 0.0, "step": 39108 }, { "epoch": 0.76, "learning_rate": 3.735289741151201e-05, "loss": 0.0, "step": 39110 }, { "epoch": 0.76, "learning_rate": 3.7352250451254785e-05, "loss": 0.0011, "step": 39112 }, { "epoch": 0.76, "learning_rate": 3.735160349099755e-05, "loss": 0.0, "step": 39114 }, { "epoch": 0.76, "learning_rate": 3.7350956530740316e-05, "loss": 0.006, "step": 39116 }, { "epoch": 0.76, "learning_rate": 3.7350309570483085e-05, "loss": 0.0047, "step": 39118 }, { "epoch": 0.76, "learning_rate": 3.7349662610225854e-05, "loss": 0.0003, "step": 39120 }, { "epoch": 0.76, "learning_rate": 3.734901564996862e-05, "loss": 0.0003, "step": 39122 }, { "epoch": 0.76, "learning_rate": 3.734836868971139e-05, "loss": 0.0189, "step": 39124 }, { "epoch": 0.76, "learning_rate": 3.734772172945416e-05, "loss": 0.0001, "step": 39126 }, { "epoch": 0.76, "learning_rate": 3.734707476919693e-05, "loss": 0.0, "step": 39128 }, { "epoch": 0.76, "learning_rate": 3.73464278089397e-05, "loss": 0.0038, "step": 39130 }, { "epoch": 0.76, "learning_rate": 3.734578084868247e-05, "loss": 0.0027, "step": 39132 }, { "epoch": 0.76, "learning_rate": 3.734513388842524e-05, "loss": 0.0, "step": 39134 }, { "epoch": 0.76, "learning_rate": 3.734448692816801e-05, "loss": 0.0, "step": 39136 }, { "epoch": 0.76, "learning_rate": 3.734383996791077e-05, "loss": 0.0, "step": 39138 }, { "epoch": 0.76, "learning_rate": 3.7343193007653545e-05, "loss": 0.0001, "step": 39140 }, { "epoch": 0.76, "learning_rate": 3.734254604739631e-05, "loss": 0.0075, "step": 39142 }, { "epoch": 0.76, "learning_rate": 3.7341899087139077e-05, "loss": 0.0005, "step": 39144 }, { "epoch": 0.76, "learning_rate": 3.7341252126881846e-05, "loss": 0.0, "step": 39146 }, { "epoch": 0.76, "learning_rate": 3.7340605166624615e-05, "loss": 0.0, "step": 39148 }, { "epoch": 0.76, "learning_rate": 3.7339958206367384e-05, "loss": 0.0004, "step": 39150 }, { "epoch": 0.76, "learning_rate": 3.733931124611015e-05, "loss": 0.0, "step": 39152 }, { "epoch": 0.76, "learning_rate": 3.733866428585292e-05, "loss": 0.0001, "step": 39154 }, { "epoch": 0.76, "learning_rate": 3.733801732559569e-05, "loss": 0.0009, "step": 39156 }, { "epoch": 0.76, "learning_rate": 3.733737036533846e-05, "loss": 0.0, "step": 39158 }, { "epoch": 0.76, "learning_rate": 3.733672340508122e-05, "loss": 0.0001, "step": 39160 }, { "epoch": 0.76, "learning_rate": 3.7336076444824e-05, "loss": 0.0153, "step": 39162 }, { "epoch": 0.76, "learning_rate": 3.733542948456677e-05, "loss": 0.0, "step": 39164 }, { "epoch": 0.76, "learning_rate": 3.7334782524309536e-05, "loss": 0.0022, "step": 39166 }, { "epoch": 0.76, "learning_rate": 3.7334135564052306e-05, "loss": 0.0302, "step": 39168 }, { "epoch": 0.76, "learning_rate": 3.733348860379507e-05, "loss": 0.0, "step": 39170 }, { "epoch": 0.76, "learning_rate": 3.7332841643537844e-05, "loss": 0.0, "step": 39172 }, { "epoch": 0.76, "learning_rate": 3.7332194683280606e-05, "loss": 0.0, "step": 39174 }, { "epoch": 0.76, "learning_rate": 3.7331547723023375e-05, "loss": 0.0, "step": 39176 }, { "epoch": 0.76, "learning_rate": 3.7330900762766144e-05, "loss": 0.0061, "step": 39178 }, { "epoch": 0.76, "learning_rate": 3.733025380250891e-05, "loss": 0.0027, "step": 39180 }, { "epoch": 0.76, "learning_rate": 3.732960684225168e-05, "loss": 0.0077, "step": 39182 }, { "epoch": 0.76, "learning_rate": 3.732895988199445e-05, "loss": 0.0, "step": 39184 }, { "epoch": 0.76, "learning_rate": 3.732831292173722e-05, "loss": 0.0004, "step": 39186 }, { "epoch": 0.76, "learning_rate": 3.732766596147999e-05, "loss": 0.0498, "step": 39188 }, { "epoch": 0.76, "learning_rate": 3.732701900122276e-05, "loss": 0.0, "step": 39190 }, { "epoch": 0.76, "learning_rate": 3.732637204096552e-05, "loss": 0.0024, "step": 39192 }, { "epoch": 0.76, "learning_rate": 3.73257250807083e-05, "loss": 0.0, "step": 39194 }, { "epoch": 0.76, "learning_rate": 3.732507812045106e-05, "loss": 0.0001, "step": 39196 }, { "epoch": 0.76, "learning_rate": 3.732443116019383e-05, "loss": 0.0, "step": 39198 }, { "epoch": 0.76, "learning_rate": 3.7323784199936604e-05, "loss": 0.0, "step": 39200 }, { "epoch": 0.76, "learning_rate": 3.7323137239679366e-05, "loss": 0.0017, "step": 39202 }, { "epoch": 0.76, "learning_rate": 3.7322490279422135e-05, "loss": 0.0, "step": 39204 }, { "epoch": 0.76, "learning_rate": 3.7321843319164905e-05, "loss": 0.0, "step": 39206 }, { "epoch": 0.76, "learning_rate": 3.7321196358907674e-05, "loss": 0.0463, "step": 39208 }, { "epoch": 0.76, "learning_rate": 3.732054939865044e-05, "loss": 0.0, "step": 39210 }, { "epoch": 0.76, "learning_rate": 3.731990243839321e-05, "loss": 0.0, "step": 39212 }, { "epoch": 0.76, "learning_rate": 3.731925547813598e-05, "loss": 0.0001, "step": 39214 }, { "epoch": 0.76, "learning_rate": 3.731860851787875e-05, "loss": 0.0, "step": 39216 }, { "epoch": 0.76, "learning_rate": 3.731796155762152e-05, "loss": 0.0, "step": 39218 }, { "epoch": 0.76, "learning_rate": 3.731731459736428e-05, "loss": 0.0035, "step": 39220 }, { "epoch": 0.76, "learning_rate": 3.731666763710706e-05, "loss": 0.0, "step": 39222 }, { "epoch": 0.76, "learning_rate": 3.731602067684982e-05, "loss": 0.0001, "step": 39224 }, { "epoch": 0.76, "learning_rate": 3.731537371659259e-05, "loss": 0.0005, "step": 39226 }, { "epoch": 0.76, "learning_rate": 3.731472675633536e-05, "loss": 0.0001, "step": 39228 }, { "epoch": 0.76, "learning_rate": 3.731407979607813e-05, "loss": 0.0001, "step": 39230 }, { "epoch": 0.76, "learning_rate": 3.73134328358209e-05, "loss": 0.0001, "step": 39232 }, { "epoch": 0.76, "learning_rate": 3.7312785875563665e-05, "loss": 0.0, "step": 39234 }, { "epoch": 0.76, "learning_rate": 3.7312138915306434e-05, "loss": 0.0001, "step": 39236 }, { "epoch": 0.76, "learning_rate": 3.73114919550492e-05, "loss": 0.0, "step": 39238 }, { "epoch": 0.76, "learning_rate": 3.731084499479197e-05, "loss": 0.0008, "step": 39240 }, { "epoch": 0.76, "learning_rate": 3.731019803453474e-05, "loss": 0.0033, "step": 39242 }, { "epoch": 0.76, "learning_rate": 3.730955107427751e-05, "loss": 0.0042, "step": 39244 }, { "epoch": 0.76, "learning_rate": 3.730890411402028e-05, "loss": 0.0001, "step": 39246 }, { "epoch": 0.76, "learning_rate": 3.730825715376305e-05, "loss": 0.0, "step": 39248 }, { "epoch": 0.76, "learning_rate": 3.730761019350582e-05, "loss": 0.0018, "step": 39250 }, { "epoch": 0.76, "learning_rate": 3.730696323324858e-05, "loss": 0.0001, "step": 39252 }, { "epoch": 0.76, "learning_rate": 3.7306316272991356e-05, "loss": 0.0047, "step": 39254 }, { "epoch": 0.76, "learning_rate": 3.730566931273412e-05, "loss": 0.0129, "step": 39256 }, { "epoch": 0.76, "learning_rate": 3.730502235247689e-05, "loss": 0.0089, "step": 39258 }, { "epoch": 0.76, "learning_rate": 3.7304375392219656e-05, "loss": 0.0, "step": 39260 }, { "epoch": 0.76, "learning_rate": 3.7303728431962425e-05, "loss": 0.0, "step": 39262 }, { "epoch": 0.76, "learning_rate": 3.7303081471705194e-05, "loss": 0.0002, "step": 39264 }, { "epoch": 0.76, "learning_rate": 3.7302434511447964e-05, "loss": 0.0, "step": 39266 }, { "epoch": 0.76, "learning_rate": 3.730178755119073e-05, "loss": 0.0003, "step": 39268 }, { "epoch": 0.76, "learning_rate": 3.73011405909335e-05, "loss": 0.0001, "step": 39270 }, { "epoch": 0.76, "learning_rate": 3.730049363067627e-05, "loss": 0.0001, "step": 39272 }, { "epoch": 0.76, "learning_rate": 3.729984667041903e-05, "loss": 0.0002, "step": 39274 }, { "epoch": 0.76, "learning_rate": 3.729919971016181e-05, "loss": 0.0005, "step": 39276 }, { "epoch": 0.76, "learning_rate": 3.729855274990458e-05, "loss": 0.0, "step": 39278 }, { "epoch": 0.76, "learning_rate": 3.729790578964734e-05, "loss": 0.0, "step": 39280 }, { "epoch": 0.76, "learning_rate": 3.7297258829390116e-05, "loss": 0.002, "step": 39282 }, { "epoch": 0.76, "learning_rate": 3.729661186913288e-05, "loss": 0.0, "step": 39284 }, { "epoch": 0.76, "learning_rate": 3.729596490887565e-05, "loss": 0.0065, "step": 39286 }, { "epoch": 0.76, "learning_rate": 3.729531794861842e-05, "loss": 0.0008, "step": 39288 }, { "epoch": 0.76, "learning_rate": 3.7294670988361186e-05, "loss": 0.0001, "step": 39290 }, { "epoch": 0.76, "learning_rate": 3.7294024028103955e-05, "loss": 0.0, "step": 39292 }, { "epoch": 0.76, "learning_rate": 3.7293377067846724e-05, "loss": 0.0001, "step": 39294 }, { "epoch": 0.76, "learning_rate": 3.729273010758949e-05, "loss": 0.0, "step": 39296 }, { "epoch": 0.76, "learning_rate": 3.729208314733226e-05, "loss": 0.0102, "step": 39298 }, { "epoch": 0.76, "learning_rate": 3.729143618707503e-05, "loss": 0.0, "step": 39300 }, { "epoch": 0.76, "learning_rate": 3.7290789226817794e-05, "loss": 0.0001, "step": 39302 }, { "epoch": 0.76, "learning_rate": 3.729014226656057e-05, "loss": 0.0003, "step": 39304 }, { "epoch": 0.76, "learning_rate": 3.728949530630333e-05, "loss": 0.0009, "step": 39306 }, { "epoch": 0.76, "learning_rate": 3.728884834604611e-05, "loss": 0.0, "step": 39308 }, { "epoch": 0.76, "learning_rate": 3.728820138578888e-05, "loss": 0.0001, "step": 39310 }, { "epoch": 0.76, "learning_rate": 3.728755442553164e-05, "loss": 0.0078, "step": 39312 }, { "epoch": 0.76, "learning_rate": 3.7286907465274415e-05, "loss": 0.0, "step": 39314 }, { "epoch": 0.76, "learning_rate": 3.728626050501718e-05, "loss": 0.0019, "step": 39316 }, { "epoch": 0.76, "learning_rate": 3.7285613544759946e-05, "loss": 0.0, "step": 39318 }, { "epoch": 0.76, "learning_rate": 3.7284966584502715e-05, "loss": 0.001, "step": 39320 }, { "epoch": 0.76, "learning_rate": 3.7284319624245484e-05, "loss": 0.0058, "step": 39322 }, { "epoch": 0.76, "learning_rate": 3.7283672663988253e-05, "loss": 0.0001, "step": 39324 }, { "epoch": 0.76, "learning_rate": 3.728302570373102e-05, "loss": 0.0, "step": 39326 }, { "epoch": 0.76, "learning_rate": 3.728237874347379e-05, "loss": 0.0389, "step": 39328 }, { "epoch": 0.76, "learning_rate": 3.728173178321656e-05, "loss": 0.0, "step": 39330 }, { "epoch": 0.76, "learning_rate": 3.728108482295933e-05, "loss": 0.0, "step": 39332 }, { "epoch": 0.76, "learning_rate": 3.728043786270209e-05, "loss": 0.0, "step": 39334 }, { "epoch": 0.76, "learning_rate": 3.727979090244487e-05, "loss": 0.0001, "step": 39336 }, { "epoch": 0.76, "learning_rate": 3.727914394218763e-05, "loss": 0.0, "step": 39338 }, { "epoch": 0.76, "learning_rate": 3.72784969819304e-05, "loss": 0.0018, "step": 39340 }, { "epoch": 0.76, "learning_rate": 3.7277850021673175e-05, "loss": 0.0043, "step": 39342 }, { "epoch": 0.76, "learning_rate": 3.727720306141594e-05, "loss": 0.0001, "step": 39344 }, { "epoch": 0.76, "learning_rate": 3.7276556101158707e-05, "loss": 0.0019, "step": 39346 }, { "epoch": 0.76, "learning_rate": 3.7275909140901476e-05, "loss": 0.0, "step": 39348 }, { "epoch": 0.76, "learning_rate": 3.7275262180644245e-05, "loss": 0.0, "step": 39350 }, { "epoch": 0.76, "learning_rate": 3.7274615220387014e-05, "loss": 0.0, "step": 39352 }, { "epoch": 0.76, "learning_rate": 3.727396826012978e-05, "loss": 0.0141, "step": 39354 }, { "epoch": 0.76, "learning_rate": 3.727332129987255e-05, "loss": 0.0, "step": 39356 }, { "epoch": 0.76, "learning_rate": 3.727267433961532e-05, "loss": 0.0001, "step": 39358 }, { "epoch": 0.76, "learning_rate": 3.727202737935809e-05, "loss": 0.0, "step": 39360 }, { "epoch": 0.76, "learning_rate": 3.727138041910085e-05, "loss": 0.0, "step": 39362 }, { "epoch": 0.76, "learning_rate": 3.727073345884363e-05, "loss": 0.0, "step": 39364 }, { "epoch": 0.76, "learning_rate": 3.727008649858639e-05, "loss": 0.0001, "step": 39366 }, { "epoch": 0.76, "learning_rate": 3.726943953832916e-05, "loss": 0.0001, "step": 39368 }, { "epoch": 0.76, "learning_rate": 3.726879257807193e-05, "loss": 0.0, "step": 39370 }, { "epoch": 0.76, "learning_rate": 3.72681456178147e-05, "loss": 0.0, "step": 39372 }, { "epoch": 0.76, "learning_rate": 3.726749865755747e-05, "loss": 0.0007, "step": 39374 }, { "epoch": 0.76, "learning_rate": 3.7266851697300236e-05, "loss": 0.0, "step": 39376 }, { "epoch": 0.76, "learning_rate": 3.7266204737043005e-05, "loss": 0.0041, "step": 39378 }, { "epoch": 0.76, "learning_rate": 3.7265557776785774e-05, "loss": 0.0221, "step": 39380 }, { "epoch": 0.76, "learning_rate": 3.726491081652854e-05, "loss": 0.0004, "step": 39382 }, { "epoch": 0.76, "learning_rate": 3.7264263856271306e-05, "loss": 0.0002, "step": 39384 }, { "epoch": 0.76, "learning_rate": 3.726361689601408e-05, "loss": 0.0001, "step": 39386 }, { "epoch": 0.76, "learning_rate": 3.726296993575685e-05, "loss": 0.0, "step": 39388 }, { "epoch": 0.76, "learning_rate": 3.726232297549962e-05, "loss": 0.0001, "step": 39390 }, { "epoch": 0.76, "learning_rate": 3.726167601524239e-05, "loss": 0.0001, "step": 39392 }, { "epoch": 0.76, "learning_rate": 3.726102905498515e-05, "loss": 0.0, "step": 39394 }, { "epoch": 0.76, "learning_rate": 3.726038209472793e-05, "loss": 0.0032, "step": 39396 }, { "epoch": 0.76, "learning_rate": 3.725973513447069e-05, "loss": 0.0, "step": 39398 }, { "epoch": 0.76, "learning_rate": 3.725908817421346e-05, "loss": 0.0002, "step": 39400 }, { "epoch": 0.76, "learning_rate": 3.725844121395623e-05, "loss": 0.0, "step": 39402 }, { "epoch": 0.76, "learning_rate": 3.7257794253698996e-05, "loss": 0.0007, "step": 39404 }, { "epoch": 0.76, "learning_rate": 3.7257147293441766e-05, "loss": 0.0, "step": 39406 }, { "epoch": 0.76, "learning_rate": 3.7256500333184535e-05, "loss": 0.0001, "step": 39408 }, { "epoch": 0.76, "learning_rate": 3.7255853372927304e-05, "loss": 0.0, "step": 39410 }, { "epoch": 0.76, "learning_rate": 3.725520641267007e-05, "loss": 0.0, "step": 39412 }, { "epoch": 0.76, "learning_rate": 3.725455945241284e-05, "loss": 0.0, "step": 39414 }, { "epoch": 0.77, "learning_rate": 3.7253912492155604e-05, "loss": 0.0, "step": 39416 }, { "epoch": 0.77, "learning_rate": 3.725326553189838e-05, "loss": 0.0, "step": 39418 }, { "epoch": 0.77, "learning_rate": 3.725261857164114e-05, "loss": 0.0, "step": 39420 }, { "epoch": 0.77, "learning_rate": 3.725197161138391e-05, "loss": 0.0001, "step": 39422 }, { "epoch": 0.77, "learning_rate": 3.725132465112669e-05, "loss": 0.0, "step": 39424 }, { "epoch": 0.77, "learning_rate": 3.725067769086945e-05, "loss": 0.0001, "step": 39426 }, { "epoch": 0.77, "learning_rate": 3.725003073061222e-05, "loss": 0.0, "step": 39428 }, { "epoch": 0.77, "learning_rate": 3.724938377035499e-05, "loss": 0.0004, "step": 39430 }, { "epoch": 0.77, "learning_rate": 3.724873681009776e-05, "loss": 0.0106, "step": 39432 }, { "epoch": 0.77, "learning_rate": 3.7248089849840526e-05, "loss": 0.0001, "step": 39434 }, { "epoch": 0.77, "learning_rate": 3.7247442889583295e-05, "loss": 0.0, "step": 39436 }, { "epoch": 0.77, "learning_rate": 3.7246795929326064e-05, "loss": 0.0035, "step": 39438 }, { "epoch": 0.77, "learning_rate": 3.724614896906883e-05, "loss": 0.0001, "step": 39440 }, { "epoch": 0.77, "learning_rate": 3.72455020088116e-05, "loss": 0.0046, "step": 39442 }, { "epoch": 0.77, "learning_rate": 3.7244855048554365e-05, "loss": 0.0, "step": 39444 }, { "epoch": 0.77, "learning_rate": 3.724420808829714e-05, "loss": 0.0077, "step": 39446 }, { "epoch": 0.77, "learning_rate": 3.72435611280399e-05, "loss": 0.0, "step": 39448 }, { "epoch": 0.77, "learning_rate": 3.724291416778268e-05, "loss": 0.0, "step": 39450 }, { "epoch": 0.77, "learning_rate": 3.724226720752544e-05, "loss": 0.0, "step": 39452 }, { "epoch": 0.77, "learning_rate": 3.724162024726821e-05, "loss": 0.0001, "step": 39454 }, { "epoch": 0.77, "learning_rate": 3.7240973287010986e-05, "loss": 0.0001, "step": 39456 }, { "epoch": 0.77, "learning_rate": 3.724032632675375e-05, "loss": 0.0, "step": 39458 }, { "epoch": 0.77, "learning_rate": 3.723967936649652e-05, "loss": 0.0027, "step": 39460 }, { "epoch": 0.77, "learning_rate": 3.7239032406239286e-05, "loss": 0.0, "step": 39462 }, { "epoch": 0.77, "learning_rate": 3.7238385445982055e-05, "loss": 0.0001, "step": 39464 }, { "epoch": 0.77, "learning_rate": 3.7237738485724825e-05, "loss": 0.0002, "step": 39466 }, { "epoch": 0.77, "learning_rate": 3.7237091525467594e-05, "loss": 0.0001, "step": 39468 }, { "epoch": 0.77, "learning_rate": 3.723644456521036e-05, "loss": 0.0, "step": 39470 }, { "epoch": 0.77, "learning_rate": 3.723579760495313e-05, "loss": 0.0008, "step": 39472 }, { "epoch": 0.77, "learning_rate": 3.72351506446959e-05, "loss": 0.0, "step": 39474 }, { "epoch": 0.77, "learning_rate": 3.723450368443866e-05, "loss": 0.0, "step": 39476 }, { "epoch": 0.77, "learning_rate": 3.723385672418144e-05, "loss": 0.0, "step": 39478 }, { "epoch": 0.77, "learning_rate": 3.72332097639242e-05, "loss": 0.0001, "step": 39480 }, { "epoch": 0.77, "learning_rate": 3.723256280366697e-05, "loss": 0.0, "step": 39482 }, { "epoch": 0.77, "learning_rate": 3.723191584340974e-05, "loss": 0.0001, "step": 39484 }, { "epoch": 0.77, "learning_rate": 3.723126888315251e-05, "loss": 0.0, "step": 39486 }, { "epoch": 0.77, "learning_rate": 3.723062192289528e-05, "loss": 0.0021, "step": 39488 }, { "epoch": 0.77, "learning_rate": 3.722997496263805e-05, "loss": 0.0002, "step": 39490 }, { "epoch": 0.77, "learning_rate": 3.7229328002380816e-05, "loss": 0.0, "step": 39492 }, { "epoch": 0.77, "learning_rate": 3.7228681042123585e-05, "loss": 0.0003, "step": 39494 }, { "epoch": 0.77, "learning_rate": 3.7228034081866354e-05, "loss": 0.0, "step": 39496 }, { "epoch": 0.77, "learning_rate": 3.7227387121609116e-05, "loss": 0.0, "step": 39498 }, { "epoch": 0.77, "learning_rate": 3.722674016135189e-05, "loss": 0.0026, "step": 39500 }, { "epoch": 0.77, "learning_rate": 3.722609320109466e-05, "loss": 0.0046, "step": 39502 }, { "epoch": 0.77, "learning_rate": 3.7225446240837424e-05, "loss": 0.0, "step": 39504 }, { "epoch": 0.77, "learning_rate": 3.72247992805802e-05, "loss": 0.0, "step": 39506 }, { "epoch": 0.77, "learning_rate": 3.722415232032296e-05, "loss": 0.0, "step": 39508 }, { "epoch": 0.77, "learning_rate": 3.722350536006573e-05, "loss": 0.0, "step": 39510 }, { "epoch": 0.77, "learning_rate": 3.72228583998085e-05, "loss": 0.0001, "step": 39512 }, { "epoch": 0.77, "learning_rate": 3.722221143955127e-05, "loss": 0.0, "step": 39514 }, { "epoch": 0.77, "learning_rate": 3.722156447929404e-05, "loss": 0.0, "step": 39516 }, { "epoch": 0.77, "learning_rate": 3.722091751903681e-05, "loss": 0.0, "step": 39518 }, { "epoch": 0.77, "learning_rate": 3.7220270558779576e-05, "loss": 0.0, "step": 39520 }, { "epoch": 0.77, "learning_rate": 3.7219623598522345e-05, "loss": 0.0007, "step": 39522 }, { "epoch": 0.77, "learning_rate": 3.7218976638265114e-05, "loss": 0.0, "step": 39524 }, { "epoch": 0.77, "learning_rate": 3.721832967800788e-05, "loss": 0.0001, "step": 39526 }, { "epoch": 0.77, "learning_rate": 3.721768271775065e-05, "loss": 0.0, "step": 39528 }, { "epoch": 0.77, "learning_rate": 3.7217035757493415e-05, "loss": 0.0072, "step": 39530 }, { "epoch": 0.77, "learning_rate": 3.721638879723619e-05, "loss": 0.0, "step": 39532 }, { "epoch": 0.77, "learning_rate": 3.721574183697896e-05, "loss": 0.0001, "step": 39534 }, { "epoch": 0.77, "learning_rate": 3.721509487672172e-05, "loss": 0.0117, "step": 39536 }, { "epoch": 0.77, "learning_rate": 3.72144479164645e-05, "loss": 0.0067, "step": 39538 }, { "epoch": 0.77, "learning_rate": 3.721380095620726e-05, "loss": 0.0, "step": 39540 }, { "epoch": 0.77, "learning_rate": 3.721315399595003e-05, "loss": 0.0025, "step": 39542 }, { "epoch": 0.77, "learning_rate": 3.72125070356928e-05, "loss": 0.0002, "step": 39544 }, { "epoch": 0.77, "learning_rate": 3.721186007543557e-05, "loss": 0.0046, "step": 39546 }, { "epoch": 0.77, "learning_rate": 3.721121311517834e-05, "loss": 0.0001, "step": 39548 }, { "epoch": 0.77, "learning_rate": 3.7210566154921106e-05, "loss": 0.0, "step": 39550 }, { "epoch": 0.77, "learning_rate": 3.7209919194663875e-05, "loss": 0.0054, "step": 39552 }, { "epoch": 0.77, "learning_rate": 3.7209272234406644e-05, "loss": 0.0, "step": 39554 }, { "epoch": 0.77, "learning_rate": 3.720862527414941e-05, "loss": 0.0069, "step": 39556 }, { "epoch": 0.77, "learning_rate": 3.7207978313892175e-05, "loss": 0.1145, "step": 39558 }, { "epoch": 0.77, "learning_rate": 3.720733135363495e-05, "loss": 0.0141, "step": 39560 }, { "epoch": 0.77, "learning_rate": 3.7206684393377714e-05, "loss": 0.0, "step": 39562 }, { "epoch": 0.77, "learning_rate": 3.720603743312048e-05, "loss": 0.0002, "step": 39564 }, { "epoch": 0.77, "learning_rate": 3.720539047286326e-05, "loss": 0.0, "step": 39566 }, { "epoch": 0.77, "learning_rate": 3.720474351260602e-05, "loss": 0.0, "step": 39568 }, { "epoch": 0.77, "learning_rate": 3.720409655234879e-05, "loss": 0.0001, "step": 39570 }, { "epoch": 0.77, "learning_rate": 3.720344959209156e-05, "loss": 0.0, "step": 39572 }, { "epoch": 0.77, "learning_rate": 3.720280263183433e-05, "loss": 0.0001, "step": 39574 }, { "epoch": 0.77, "learning_rate": 3.72021556715771e-05, "loss": 0.0001, "step": 39576 }, { "epoch": 0.77, "learning_rate": 3.7201508711319866e-05, "loss": 0.0, "step": 39578 }, { "epoch": 0.77, "learning_rate": 3.7200861751062635e-05, "loss": 0.0255, "step": 39580 }, { "epoch": 0.77, "learning_rate": 3.7200214790805404e-05, "loss": 0.0017, "step": 39582 }, { "epoch": 0.77, "learning_rate": 3.7199567830548173e-05, "loss": 0.011, "step": 39584 }, { "epoch": 0.77, "learning_rate": 3.7198920870290936e-05, "loss": 0.0003, "step": 39586 }, { "epoch": 0.77, "learning_rate": 3.719827391003371e-05, "loss": 0.0, "step": 39588 }, { "epoch": 0.77, "learning_rate": 3.7197626949776474e-05, "loss": 0.0004, "step": 39590 }, { "epoch": 0.77, "learning_rate": 3.719697998951925e-05, "loss": 0.0, "step": 39592 }, { "epoch": 0.77, "learning_rate": 3.719633302926201e-05, "loss": 0.0, "step": 39594 }, { "epoch": 0.77, "learning_rate": 3.719568606900478e-05, "loss": 0.0, "step": 39596 }, { "epoch": 0.77, "learning_rate": 3.719503910874755e-05, "loss": 0.0326, "step": 39598 }, { "epoch": 0.77, "learning_rate": 3.719439214849032e-05, "loss": 0.0162, "step": 39600 }, { "epoch": 0.77, "learning_rate": 3.719374518823309e-05, "loss": 0.0009, "step": 39602 }, { "epoch": 0.77, "learning_rate": 3.719309822797586e-05, "loss": 0.0002, "step": 39604 }, { "epoch": 0.77, "learning_rate": 3.7192451267718627e-05, "loss": 0.0009, "step": 39606 }, { "epoch": 0.77, "learning_rate": 3.719180430746139e-05, "loss": 0.0, "step": 39608 }, { "epoch": 0.77, "learning_rate": 3.7191157347204165e-05, "loss": 0.0, "step": 39610 }, { "epoch": 0.77, "learning_rate": 3.7190510386946934e-05, "loss": 0.0045, "step": 39612 }, { "epoch": 0.77, "learning_rate": 3.71898634266897e-05, "loss": 0.0, "step": 39614 }, { "epoch": 0.77, "learning_rate": 3.718921646643247e-05, "loss": 0.0, "step": 39616 }, { "epoch": 0.77, "learning_rate": 3.7188569506175234e-05, "loss": 0.0, "step": 39618 }, { "epoch": 0.77, "learning_rate": 3.718792254591801e-05, "loss": 0.0, "step": 39620 }, { "epoch": 0.77, "learning_rate": 3.718727558566077e-05, "loss": 0.0012, "step": 39622 }, { "epoch": 0.77, "learning_rate": 3.718662862540354e-05, "loss": 0.0003, "step": 39624 }, { "epoch": 0.77, "learning_rate": 3.718598166514631e-05, "loss": 0.0016, "step": 39626 }, { "epoch": 0.77, "learning_rate": 3.718533470488908e-05, "loss": 0.0, "step": 39628 }, { "epoch": 0.77, "learning_rate": 3.718468774463185e-05, "loss": 0.0, "step": 39630 }, { "epoch": 0.77, "learning_rate": 3.718404078437462e-05, "loss": 0.0022, "step": 39632 }, { "epoch": 0.77, "learning_rate": 3.718339382411739e-05, "loss": 0.0007, "step": 39634 }, { "epoch": 0.77, "learning_rate": 3.7182746863860156e-05, "loss": 0.0, "step": 39636 }, { "epoch": 0.77, "learning_rate": 3.7182099903602925e-05, "loss": 0.0014, "step": 39638 }, { "epoch": 0.77, "learning_rate": 3.718145294334569e-05, "loss": 0.0, "step": 39640 }, { "epoch": 0.77, "learning_rate": 3.718080598308846e-05, "loss": 0.0, "step": 39642 }, { "epoch": 0.77, "learning_rate": 3.718015902283123e-05, "loss": 0.003, "step": 39644 }, { "epoch": 0.77, "learning_rate": 3.7179512062573995e-05, "loss": 0.0001, "step": 39646 }, { "epoch": 0.77, "learning_rate": 3.717886510231677e-05, "loss": 0.0, "step": 39648 }, { "epoch": 0.77, "learning_rate": 3.717821814205953e-05, "loss": 0.0002, "step": 39650 }, { "epoch": 0.77, "learning_rate": 3.71775711818023e-05, "loss": 0.0001, "step": 39652 }, { "epoch": 0.77, "learning_rate": 3.717692422154507e-05, "loss": 0.0001, "step": 39654 }, { "epoch": 0.77, "learning_rate": 3.717627726128784e-05, "loss": 0.0001, "step": 39656 }, { "epoch": 0.77, "learning_rate": 3.717563030103061e-05, "loss": 0.0, "step": 39658 }, { "epoch": 0.77, "learning_rate": 3.717498334077338e-05, "loss": 0.0064, "step": 39660 }, { "epoch": 0.77, "learning_rate": 3.717433638051615e-05, "loss": 0.0004, "step": 39662 }, { "epoch": 0.77, "learning_rate": 3.7173689420258916e-05, "loss": 0.0001, "step": 39664 }, { "epoch": 0.77, "learning_rate": 3.7173042460001686e-05, "loss": 0.0002, "step": 39666 }, { "epoch": 0.77, "learning_rate": 3.717239549974445e-05, "loss": 0.0, "step": 39668 }, { "epoch": 0.77, "learning_rate": 3.7171748539487224e-05, "loss": 0.0001, "step": 39670 }, { "epoch": 0.77, "learning_rate": 3.7171101579229986e-05, "loss": 0.0, "step": 39672 }, { "epoch": 0.77, "learning_rate": 3.717045461897276e-05, "loss": 0.0163, "step": 39674 }, { "epoch": 0.77, "learning_rate": 3.7169807658715524e-05, "loss": 0.0002, "step": 39676 }, { "epoch": 0.77, "learning_rate": 3.716916069845829e-05, "loss": 0.0, "step": 39678 }, { "epoch": 0.77, "learning_rate": 3.716851373820107e-05, "loss": 0.0, "step": 39680 }, { "epoch": 0.77, "learning_rate": 3.716786677794383e-05, "loss": 0.0082, "step": 39682 }, { "epoch": 0.77, "learning_rate": 3.71672198176866e-05, "loss": 0.0, "step": 39684 }, { "epoch": 0.77, "learning_rate": 3.716657285742937e-05, "loss": 0.0056, "step": 39686 }, { "epoch": 0.77, "learning_rate": 3.716592589717214e-05, "loss": 0.0001, "step": 39688 }, { "epoch": 0.77, "learning_rate": 3.716527893691491e-05, "loss": 0.0, "step": 39690 }, { "epoch": 0.77, "learning_rate": 3.716463197665768e-05, "loss": 0.0074, "step": 39692 }, { "epoch": 0.77, "learning_rate": 3.7163985016400446e-05, "loss": 0.0, "step": 39694 }, { "epoch": 0.77, "learning_rate": 3.7163338056143215e-05, "loss": 0.0, "step": 39696 }, { "epoch": 0.77, "learning_rate": 3.7162691095885984e-05, "loss": 0.0, "step": 39698 }, { "epoch": 0.77, "learning_rate": 3.7162044135628746e-05, "loss": 0.0102, "step": 39700 }, { "epoch": 0.77, "learning_rate": 3.716139717537152e-05, "loss": 0.0, "step": 39702 }, { "epoch": 0.77, "learning_rate": 3.7160750215114285e-05, "loss": 0.0001, "step": 39704 }, { "epoch": 0.77, "learning_rate": 3.7160103254857054e-05, "loss": 0.0, "step": 39706 }, { "epoch": 0.77, "learning_rate": 3.715945629459982e-05, "loss": 0.0001, "step": 39708 }, { "epoch": 0.77, "learning_rate": 3.715880933434259e-05, "loss": 0.0, "step": 39710 }, { "epoch": 0.77, "learning_rate": 3.715816237408536e-05, "loss": 0.0001, "step": 39712 }, { "epoch": 0.77, "learning_rate": 3.715751541382813e-05, "loss": 0.0, "step": 39714 }, { "epoch": 0.77, "learning_rate": 3.71568684535709e-05, "loss": 0.0026, "step": 39716 }, { "epoch": 0.77, "learning_rate": 3.715622149331367e-05, "loss": 0.0002, "step": 39718 }, { "epoch": 0.77, "learning_rate": 3.715557453305644e-05, "loss": 0.0, "step": 39720 }, { "epoch": 0.77, "learning_rate": 3.7154927572799206e-05, "loss": 0.0, "step": 39722 }, { "epoch": 0.77, "learning_rate": 3.7154280612541975e-05, "loss": 0.0, "step": 39724 }, { "epoch": 0.77, "learning_rate": 3.7153633652284745e-05, "loss": 0.0, "step": 39726 }, { "epoch": 0.77, "learning_rate": 3.715298669202751e-05, "loss": 0.0, "step": 39728 }, { "epoch": 0.77, "learning_rate": 3.715233973177028e-05, "loss": 0.0, "step": 39730 }, { "epoch": 0.77, "learning_rate": 3.7151692771513045e-05, "loss": 0.0002, "step": 39732 }, { "epoch": 0.77, "learning_rate": 3.715104581125582e-05, "loss": 0.0089, "step": 39734 }, { "epoch": 0.77, "learning_rate": 3.715039885099858e-05, "loss": 0.0015, "step": 39736 }, { "epoch": 0.77, "learning_rate": 3.714975189074135e-05, "loss": 0.0, "step": 39738 }, { "epoch": 0.77, "learning_rate": 3.714910493048412e-05, "loss": 0.0, "step": 39740 }, { "epoch": 0.77, "learning_rate": 3.714845797022689e-05, "loss": 0.0001, "step": 39742 }, { "epoch": 0.77, "learning_rate": 3.714781100996966e-05, "loss": 0.0087, "step": 39744 }, { "epoch": 0.77, "learning_rate": 3.714716404971243e-05, "loss": 0.0, "step": 39746 }, { "epoch": 0.77, "learning_rate": 3.71465170894552e-05, "loss": 0.0001, "step": 39748 }, { "epoch": 0.77, "learning_rate": 3.714587012919796e-05, "loss": 0.0, "step": 39750 }, { "epoch": 0.77, "learning_rate": 3.7145223168940736e-05, "loss": 0.0001, "step": 39752 }, { "epoch": 0.77, "learning_rate": 3.71445762086835e-05, "loss": 0.0, "step": 39754 }, { "epoch": 0.77, "learning_rate": 3.7143929248426274e-05, "loss": 0.0003, "step": 39756 }, { "epoch": 0.77, "learning_rate": 3.714328228816904e-05, "loss": 0.0, "step": 39758 }, { "epoch": 0.77, "learning_rate": 3.7142635327911805e-05, "loss": 0.0001, "step": 39760 }, { "epoch": 0.77, "learning_rate": 3.714198836765458e-05, "loss": 0.0, "step": 39762 }, { "epoch": 0.77, "learning_rate": 3.7141341407397344e-05, "loss": 0.0, "step": 39764 }, { "epoch": 0.77, "learning_rate": 3.714069444714011e-05, "loss": 0.0001, "step": 39766 }, { "epoch": 0.77, "learning_rate": 3.714004748688288e-05, "loss": 0.019, "step": 39768 }, { "epoch": 0.77, "learning_rate": 3.713940052662565e-05, "loss": 0.0, "step": 39770 }, { "epoch": 0.77, "learning_rate": 3.713875356636842e-05, "loss": 0.0, "step": 39772 }, { "epoch": 0.77, "learning_rate": 3.713810660611119e-05, "loss": 0.0002, "step": 39774 }, { "epoch": 0.77, "learning_rate": 3.713745964585396e-05, "loss": 0.0002, "step": 39776 }, { "epoch": 0.77, "learning_rate": 3.713681268559673e-05, "loss": 0.0055, "step": 39778 }, { "epoch": 0.77, "learning_rate": 3.7136165725339496e-05, "loss": 0.0, "step": 39780 }, { "epoch": 0.77, "learning_rate": 3.713551876508226e-05, "loss": 0.0, "step": 39782 }, { "epoch": 0.77, "learning_rate": 3.7134871804825034e-05, "loss": 0.0001, "step": 39784 }, { "epoch": 0.77, "learning_rate": 3.71342248445678e-05, "loss": 0.0002, "step": 39786 }, { "epoch": 0.77, "learning_rate": 3.7133577884310566e-05, "loss": 0.0, "step": 39788 }, { "epoch": 0.77, "learning_rate": 3.713293092405334e-05, "loss": 0.0, "step": 39790 }, { "epoch": 0.77, "learning_rate": 3.7132283963796104e-05, "loss": 0.0001, "step": 39792 }, { "epoch": 0.77, "learning_rate": 3.713163700353887e-05, "loss": 0.0001, "step": 39794 }, { "epoch": 0.77, "learning_rate": 3.713099004328164e-05, "loss": 0.0001, "step": 39796 }, { "epoch": 0.77, "learning_rate": 3.713034308302441e-05, "loss": 0.0003, "step": 39798 }, { "epoch": 0.77, "learning_rate": 3.712969612276718e-05, "loss": 0.0004, "step": 39800 }, { "epoch": 0.77, "learning_rate": 3.712904916250995e-05, "loss": 0.0002, "step": 39802 }, { "epoch": 0.77, "learning_rate": 3.712840220225272e-05, "loss": 0.0009, "step": 39804 }, { "epoch": 0.77, "learning_rate": 3.712775524199549e-05, "loss": 0.0063, "step": 39806 }, { "epoch": 0.77, "learning_rate": 3.712710828173826e-05, "loss": 0.0104, "step": 39808 }, { "epoch": 0.77, "learning_rate": 3.712646132148102e-05, "loss": 0.0, "step": 39810 }, { "epoch": 0.77, "learning_rate": 3.7125814361223795e-05, "loss": 0.0, "step": 39812 }, { "epoch": 0.77, "learning_rate": 3.712516740096656e-05, "loss": 0.0, "step": 39814 }, { "epoch": 0.77, "learning_rate": 3.712452044070933e-05, "loss": 0.0, "step": 39816 }, { "epoch": 0.77, "learning_rate": 3.7123873480452095e-05, "loss": 0.0, "step": 39818 }, { "epoch": 0.77, "learning_rate": 3.7123226520194864e-05, "loss": 0.0007, "step": 39820 }, { "epoch": 0.77, "learning_rate": 3.712257955993764e-05, "loss": 0.0, "step": 39822 }, { "epoch": 0.77, "learning_rate": 3.71219325996804e-05, "loss": 0.0, "step": 39824 }, { "epoch": 0.77, "learning_rate": 3.712128563942317e-05, "loss": 0.0, "step": 39826 }, { "epoch": 0.77, "learning_rate": 3.712063867916594e-05, "loss": 0.0, "step": 39828 }, { "epoch": 0.77, "learning_rate": 3.711999171890871e-05, "loss": 0.0, "step": 39830 }, { "epoch": 0.77, "learning_rate": 3.711934475865147e-05, "loss": 0.0, "step": 39832 }, { "epoch": 0.77, "learning_rate": 3.711869779839425e-05, "loss": 0.0, "step": 39834 }, { "epoch": 0.77, "learning_rate": 3.711805083813702e-05, "loss": 0.0002, "step": 39836 }, { "epoch": 0.77, "learning_rate": 3.7117403877879786e-05, "loss": 0.0, "step": 39838 }, { "epoch": 0.77, "learning_rate": 3.7116756917622555e-05, "loss": 0.0035, "step": 39840 }, { "epoch": 0.77, "learning_rate": 3.711610995736532e-05, "loss": 0.0046, "step": 39842 }, { "epoch": 0.77, "learning_rate": 3.7115462997108093e-05, "loss": 0.002, "step": 39844 }, { "epoch": 0.77, "learning_rate": 3.7114816036850856e-05, "loss": 0.0053, "step": 39846 }, { "epoch": 0.77, "learning_rate": 3.7114169076593625e-05, "loss": 0.0038, "step": 39848 }, { "epoch": 0.77, "learning_rate": 3.7113522116336394e-05, "loss": 0.0002, "step": 39850 }, { "epoch": 0.77, "learning_rate": 3.711287515607916e-05, "loss": 0.0542, "step": 39852 }, { "epoch": 0.77, "learning_rate": 3.711222819582193e-05, "loss": 0.0, "step": 39854 }, { "epoch": 0.77, "learning_rate": 3.71115812355647e-05, "loss": 0.0, "step": 39856 }, { "epoch": 0.77, "learning_rate": 3.711093427530747e-05, "loss": 0.0001, "step": 39858 }, { "epoch": 0.77, "learning_rate": 3.711028731505024e-05, "loss": 0.0, "step": 39860 }, { "epoch": 0.77, "learning_rate": 3.710964035479301e-05, "loss": 0.0, "step": 39862 }, { "epoch": 0.77, "learning_rate": 3.710899339453577e-05, "loss": 0.0, "step": 39864 }, { "epoch": 0.77, "learning_rate": 3.7108346434278547e-05, "loss": 0.0, "step": 39866 }, { "epoch": 0.77, "learning_rate": 3.7107699474021316e-05, "loss": 0.0004, "step": 39868 }, { "epoch": 0.77, "learning_rate": 3.710705251376408e-05, "loss": 0.0, "step": 39870 }, { "epoch": 0.77, "learning_rate": 3.7106405553506854e-05, "loss": 0.0002, "step": 39872 }, { "epoch": 0.77, "learning_rate": 3.7105758593249616e-05, "loss": 0.0001, "step": 39874 }, { "epoch": 0.77, "learning_rate": 3.710511163299239e-05, "loss": 0.0004, "step": 39876 }, { "epoch": 0.77, "learning_rate": 3.7104464672735154e-05, "loss": 0.0, "step": 39878 }, { "epoch": 0.77, "learning_rate": 3.710381771247792e-05, "loss": 0.0001, "step": 39880 }, { "epoch": 0.77, "learning_rate": 3.710317075222069e-05, "loss": 0.0018, "step": 39882 }, { "epoch": 0.77, "learning_rate": 3.710252379196346e-05, "loss": 0.0044, "step": 39884 }, { "epoch": 0.77, "learning_rate": 3.710187683170623e-05, "loss": 0.0, "step": 39886 }, { "epoch": 0.77, "learning_rate": 3.7101229871449e-05, "loss": 0.0001, "step": 39888 }, { "epoch": 0.77, "learning_rate": 3.710058291119177e-05, "loss": 0.0, "step": 39890 }, { "epoch": 0.77, "learning_rate": 3.709993595093453e-05, "loss": 0.0001, "step": 39892 }, { "epoch": 0.77, "learning_rate": 3.709928899067731e-05, "loss": 0.0, "step": 39894 }, { "epoch": 0.77, "learning_rate": 3.709864203042007e-05, "loss": 0.0036, "step": 39896 }, { "epoch": 0.77, "learning_rate": 3.7097995070162845e-05, "loss": 0.0006, "step": 39898 }, { "epoch": 0.77, "learning_rate": 3.709734810990561e-05, "loss": 0.0139, "step": 39900 }, { "epoch": 0.77, "learning_rate": 3.7096701149648377e-05, "loss": 0.0046, "step": 39902 }, { "epoch": 0.77, "learning_rate": 3.709605418939115e-05, "loss": 0.0002, "step": 39904 }, { "epoch": 0.77, "learning_rate": 3.7095407229133915e-05, "loss": 0.0006, "step": 39906 }, { "epoch": 0.77, "learning_rate": 3.7094760268876684e-05, "loss": 0.0, "step": 39908 }, { "epoch": 0.77, "learning_rate": 3.709411330861945e-05, "loss": 0.0, "step": 39910 }, { "epoch": 0.77, "learning_rate": 3.709346634836222e-05, "loss": 0.0028, "step": 39912 }, { "epoch": 0.77, "learning_rate": 3.709281938810499e-05, "loss": 0.0031, "step": 39914 }, { "epoch": 0.77, "learning_rate": 3.709217242784776e-05, "loss": 0.0, "step": 39916 }, { "epoch": 0.77, "learning_rate": 3.709152546759053e-05, "loss": 0.0, "step": 39918 }, { "epoch": 0.77, "learning_rate": 3.70908785073333e-05, "loss": 0.0, "step": 39920 }, { "epoch": 0.77, "learning_rate": 3.709023154707607e-05, "loss": 0.0, "step": 39922 }, { "epoch": 0.77, "learning_rate": 3.708958458681883e-05, "loss": 0.0119, "step": 39924 }, { "epoch": 0.77, "learning_rate": 3.7088937626561606e-05, "loss": 0.0, "step": 39926 }, { "epoch": 0.77, "learning_rate": 3.708829066630437e-05, "loss": 0.0047, "step": 39928 }, { "epoch": 0.77, "learning_rate": 3.708764370604714e-05, "loss": 0.0001, "step": 39930 }, { "epoch": 0.78, "learning_rate": 3.7086996745789906e-05, "loss": 0.0002, "step": 39932 }, { "epoch": 0.78, "learning_rate": 3.7086349785532675e-05, "loss": 0.0, "step": 39934 }, { "epoch": 0.78, "learning_rate": 3.7085702825275444e-05, "loss": 0.009, "step": 39936 }, { "epoch": 0.78, "learning_rate": 3.708505586501821e-05, "loss": 0.0, "step": 39938 }, { "epoch": 0.78, "learning_rate": 3.708440890476098e-05, "loss": 0.0079, "step": 39940 }, { "epoch": 0.78, "learning_rate": 3.708376194450375e-05, "loss": 0.0002, "step": 39942 }, { "epoch": 0.78, "learning_rate": 3.708311498424652e-05, "loss": 0.0004, "step": 39944 }, { "epoch": 0.78, "learning_rate": 3.708246802398929e-05, "loss": 0.0388, "step": 39946 }, { "epoch": 0.78, "learning_rate": 3.708182106373206e-05, "loss": 0.0, "step": 39948 }, { "epoch": 0.78, "learning_rate": 3.708117410347483e-05, "loss": 0.0, "step": 39950 }, { "epoch": 0.78, "learning_rate": 3.708052714321759e-05, "loss": 0.0001, "step": 39952 }, { "epoch": 0.78, "learning_rate": 3.7079880182960366e-05, "loss": 0.0001, "step": 39954 }, { "epoch": 0.78, "learning_rate": 3.707923322270313e-05, "loss": 0.0001, "step": 39956 }, { "epoch": 0.78, "learning_rate": 3.7078586262445904e-05, "loss": 0.0001, "step": 39958 }, { "epoch": 0.78, "learning_rate": 3.7077939302188666e-05, "loss": 0.0, "step": 39960 }, { "epoch": 0.78, "learning_rate": 3.7077292341931436e-05, "loss": 0.0, "step": 39962 }, { "epoch": 0.78, "learning_rate": 3.7076645381674205e-05, "loss": 0.0001, "step": 39964 }, { "epoch": 0.78, "learning_rate": 3.7075998421416974e-05, "loss": 0.0, "step": 39966 }, { "epoch": 0.78, "learning_rate": 3.707535146115974e-05, "loss": 0.0, "step": 39968 }, { "epoch": 0.78, "learning_rate": 3.707470450090251e-05, "loss": 0.004, "step": 39970 }, { "epoch": 0.78, "learning_rate": 3.707405754064528e-05, "loss": 0.0002, "step": 39972 }, { "epoch": 0.78, "learning_rate": 3.707341058038804e-05, "loss": 0.0001, "step": 39974 }, { "epoch": 0.78, "learning_rate": 3.707276362013082e-05, "loss": 0.0061, "step": 39976 }, { "epoch": 0.78, "learning_rate": 3.707211665987358e-05, "loss": 0.0, "step": 39978 }, { "epoch": 0.78, "learning_rate": 3.707146969961636e-05, "loss": 0.0, "step": 39980 }, { "epoch": 0.78, "learning_rate": 3.7070822739359126e-05, "loss": 0.0, "step": 39982 }, { "epoch": 0.78, "learning_rate": 3.707017577910189e-05, "loss": 0.0, "step": 39984 }, { "epoch": 0.78, "learning_rate": 3.7069528818844665e-05, "loss": 0.0001, "step": 39986 }, { "epoch": 0.78, "learning_rate": 3.706888185858743e-05, "loss": 0.0, "step": 39988 }, { "epoch": 0.78, "learning_rate": 3.7068234898330196e-05, "loss": 0.0, "step": 39990 }, { "epoch": 0.78, "learning_rate": 3.7067587938072965e-05, "loss": 0.0, "step": 39992 }, { "epoch": 0.78, "learning_rate": 3.7066940977815734e-05, "loss": 0.0001, "step": 39994 }, { "epoch": 0.78, "learning_rate": 3.70662940175585e-05, "loss": 0.0, "step": 39996 }, { "epoch": 0.78, "learning_rate": 3.706564705730127e-05, "loss": 0.0, "step": 39998 }, { "epoch": 0.78, "learning_rate": 3.706500009704404e-05, "loss": 0.0125, "step": 40000 }, { "epoch": 0.78, "learning_rate": 3.706435313678681e-05, "loss": 0.0129, "step": 40002 }, { "epoch": 0.78, "learning_rate": 3.706370617652958e-05, "loss": 0.0063, "step": 40004 }, { "epoch": 0.78, "learning_rate": 3.706305921627234e-05, "loss": 0.0, "step": 40006 }, { "epoch": 0.78, "learning_rate": 3.706241225601512e-05, "loss": 0.0003, "step": 40008 }, { "epoch": 0.78, "learning_rate": 3.706176529575788e-05, "loss": 0.0, "step": 40010 }, { "epoch": 0.78, "learning_rate": 3.706111833550065e-05, "loss": 0.0001, "step": 40012 }, { "epoch": 0.78, "learning_rate": 3.7060471375243425e-05, "loss": 0.0013, "step": 40014 }, { "epoch": 0.78, "learning_rate": 3.705982441498619e-05, "loss": 0.026, "step": 40016 }, { "epoch": 0.78, "learning_rate": 3.705917745472896e-05, "loss": 0.0004, "step": 40018 }, { "epoch": 0.78, "learning_rate": 3.7058530494471725e-05, "loss": 0.0, "step": 40020 }, { "epoch": 0.78, "learning_rate": 3.7057883534214494e-05, "loss": 0.0, "step": 40022 }, { "epoch": 0.78, "learning_rate": 3.7057236573957264e-05, "loss": 0.0001, "step": 40024 }, { "epoch": 0.78, "learning_rate": 3.705658961370003e-05, "loss": 0.0002, "step": 40026 }, { "epoch": 0.78, "learning_rate": 3.70559426534428e-05, "loss": 0.0036, "step": 40028 }, { "epoch": 0.78, "learning_rate": 3.705529569318557e-05, "loss": 0.0, "step": 40030 }, { "epoch": 0.78, "learning_rate": 3.705464873292834e-05, "loss": 0.0005, "step": 40032 }, { "epoch": 0.78, "learning_rate": 3.70540017726711e-05, "loss": 0.0, "step": 40034 }, { "epoch": 0.78, "learning_rate": 3.705335481241388e-05, "loss": 0.0, "step": 40036 }, { "epoch": 0.78, "learning_rate": 3.705270785215664e-05, "loss": 0.0, "step": 40038 }, { "epoch": 0.78, "learning_rate": 3.7052060891899416e-05, "loss": 0.0001, "step": 40040 }, { "epoch": 0.78, "learning_rate": 3.705141393164218e-05, "loss": 0.0, "step": 40042 }, { "epoch": 0.78, "learning_rate": 3.705076697138495e-05, "loss": 0.0001, "step": 40044 }, { "epoch": 0.78, "learning_rate": 3.7050120011127723e-05, "loss": 0.0, "step": 40046 }, { "epoch": 0.78, "learning_rate": 3.7049473050870486e-05, "loss": 0.0001, "step": 40048 }, { "epoch": 0.78, "learning_rate": 3.7048826090613255e-05, "loss": 0.0093, "step": 40050 }, { "epoch": 0.78, "learning_rate": 3.7048179130356024e-05, "loss": 0.0034, "step": 40052 }, { "epoch": 0.78, "learning_rate": 3.704753217009879e-05, "loss": 0.0001, "step": 40054 }, { "epoch": 0.78, "learning_rate": 3.7046885209841555e-05, "loss": 0.0, "step": 40056 }, { "epoch": 0.78, "learning_rate": 3.704623824958433e-05, "loss": 0.0, "step": 40058 }, { "epoch": 0.78, "learning_rate": 3.70455912893271e-05, "loss": 0.0, "step": 40060 }, { "epoch": 0.78, "learning_rate": 3.704494432906987e-05, "loss": 0.0, "step": 40062 }, { "epoch": 0.78, "learning_rate": 3.704429736881264e-05, "loss": 0.0, "step": 40064 }, { "epoch": 0.78, "learning_rate": 3.70436504085554e-05, "loss": 0.0249, "step": 40066 }, { "epoch": 0.78, "learning_rate": 3.704300344829818e-05, "loss": 0.0, "step": 40068 }, { "epoch": 0.78, "learning_rate": 3.704235648804094e-05, "loss": 0.0, "step": 40070 }, { "epoch": 0.78, "learning_rate": 3.704170952778371e-05, "loss": 0.0016, "step": 40072 }, { "epoch": 0.78, "learning_rate": 3.704106256752648e-05, "loss": 0.0, "step": 40074 }, { "epoch": 0.78, "learning_rate": 3.7040415607269246e-05, "loss": 0.0016, "step": 40076 }, { "epoch": 0.78, "learning_rate": 3.7039768647012015e-05, "loss": 0.0, "step": 40078 }, { "epoch": 0.78, "learning_rate": 3.7039121686754784e-05, "loss": 0.0001, "step": 40080 }, { "epoch": 0.78, "learning_rate": 3.7038474726497553e-05, "loss": 0.0045, "step": 40082 }, { "epoch": 0.78, "learning_rate": 3.703782776624032e-05, "loss": 0.0066, "step": 40084 }, { "epoch": 0.78, "learning_rate": 3.703718080598309e-05, "loss": 0.0, "step": 40086 }, { "epoch": 0.78, "learning_rate": 3.7036533845725854e-05, "loss": 0.0, "step": 40088 }, { "epoch": 0.78, "learning_rate": 3.703588688546863e-05, "loss": 0.006, "step": 40090 }, { "epoch": 0.78, "learning_rate": 3.70352399252114e-05, "loss": 0.0006, "step": 40092 }, { "epoch": 0.78, "learning_rate": 3.703459296495416e-05, "loss": 0.0084, "step": 40094 }, { "epoch": 0.78, "learning_rate": 3.703394600469694e-05, "loss": 0.0111, "step": 40096 }, { "epoch": 0.78, "learning_rate": 3.70332990444397e-05, "loss": 0.0039, "step": 40098 }, { "epoch": 0.78, "learning_rate": 3.7032652084182475e-05, "loss": 0.0, "step": 40100 }, { "epoch": 0.78, "learning_rate": 3.703200512392524e-05, "loss": 0.0004, "step": 40102 }, { "epoch": 0.78, "learning_rate": 3.703135816366801e-05, "loss": 0.0, "step": 40104 }, { "epoch": 0.78, "learning_rate": 3.7030711203410776e-05, "loss": 0.0089, "step": 40106 }, { "epoch": 0.78, "learning_rate": 3.7030064243153545e-05, "loss": 0.0, "step": 40108 }, { "epoch": 0.78, "learning_rate": 3.7029417282896314e-05, "loss": 0.0002, "step": 40110 }, { "epoch": 0.78, "learning_rate": 3.702877032263908e-05, "loss": 0.0001, "step": 40112 }, { "epoch": 0.78, "learning_rate": 3.702812336238185e-05, "loss": 0.009, "step": 40114 }, { "epoch": 0.78, "learning_rate": 3.7027476402124614e-05, "loss": 0.0001, "step": 40116 }, { "epoch": 0.78, "learning_rate": 3.702682944186739e-05, "loss": 0.0012, "step": 40118 }, { "epoch": 0.78, "learning_rate": 3.702618248161015e-05, "loss": 0.0002, "step": 40120 }, { "epoch": 0.78, "learning_rate": 3.702553552135293e-05, "loss": 0.0004, "step": 40122 }, { "epoch": 0.78, "learning_rate": 3.70248885610957e-05, "loss": 0.0002, "step": 40124 }, { "epoch": 0.78, "learning_rate": 3.702424160083846e-05, "loss": 0.0002, "step": 40126 }, { "epoch": 0.78, "learning_rate": 3.7023594640581236e-05, "loss": 0.0, "step": 40128 }, { "epoch": 0.78, "learning_rate": 3.7022947680324e-05, "loss": 0.0002, "step": 40130 }, { "epoch": 0.78, "learning_rate": 3.702230072006677e-05, "loss": 0.0, "step": 40132 }, { "epoch": 0.78, "learning_rate": 3.7021653759809536e-05, "loss": 0.0001, "step": 40134 }, { "epoch": 0.78, "learning_rate": 3.7021006799552305e-05, "loss": 0.0003, "step": 40136 }, { "epoch": 0.78, "learning_rate": 3.7020359839295074e-05, "loss": 0.0092, "step": 40138 }, { "epoch": 0.78, "learning_rate": 3.701971287903784e-05, "loss": 0.0007, "step": 40140 }, { "epoch": 0.78, "learning_rate": 3.701906591878061e-05, "loss": 0.0046, "step": 40142 }, { "epoch": 0.78, "learning_rate": 3.701841895852338e-05, "loss": 0.0004, "step": 40144 }, { "epoch": 0.78, "learning_rate": 3.701777199826615e-05, "loss": 0.0005, "step": 40146 }, { "epoch": 0.78, "learning_rate": 3.701712503800891e-05, "loss": 0.0, "step": 40148 }, { "epoch": 0.78, "learning_rate": 3.701647807775169e-05, "loss": 0.0001, "step": 40150 }, { "epoch": 0.78, "learning_rate": 3.701583111749445e-05, "loss": 0.001, "step": 40152 }, { "epoch": 0.78, "learning_rate": 3.701518415723722e-05, "loss": 0.0001, "step": 40154 }, { "epoch": 0.78, "learning_rate": 3.701453719697999e-05, "loss": 0.0, "step": 40156 }, { "epoch": 0.78, "learning_rate": 3.701389023672276e-05, "loss": 0.0059, "step": 40158 }, { "epoch": 0.78, "learning_rate": 3.7013243276465534e-05, "loss": 0.0, "step": 40160 }, { "epoch": 0.78, "learning_rate": 3.7012596316208297e-05, "loss": 0.0011, "step": 40162 }, { "epoch": 0.78, "learning_rate": 3.7011949355951066e-05, "loss": 0.0, "step": 40164 }, { "epoch": 0.78, "learning_rate": 3.7011302395693835e-05, "loss": 0.0004, "step": 40166 }, { "epoch": 0.78, "learning_rate": 3.7010655435436604e-05, "loss": 0.0006, "step": 40168 }, { "epoch": 0.78, "learning_rate": 3.701000847517937e-05, "loss": 0.0001, "step": 40170 }, { "epoch": 0.78, "learning_rate": 3.700936151492214e-05, "loss": 0.0, "step": 40172 }, { "epoch": 0.78, "learning_rate": 3.700871455466491e-05, "loss": 0.0, "step": 40174 }, { "epoch": 0.78, "learning_rate": 3.700806759440767e-05, "loss": 0.0007, "step": 40176 }, { "epoch": 0.78, "learning_rate": 3.700742063415045e-05, "loss": 0.0, "step": 40178 }, { "epoch": 0.78, "learning_rate": 3.700677367389321e-05, "loss": 0.0, "step": 40180 }, { "epoch": 0.78, "learning_rate": 3.700612671363599e-05, "loss": 0.0001, "step": 40182 }, { "epoch": 0.78, "learning_rate": 3.700547975337875e-05, "loss": 0.0002, "step": 40184 }, { "epoch": 0.78, "learning_rate": 3.700483279312152e-05, "loss": 0.0108, "step": 40186 }, { "epoch": 0.78, "learning_rate": 3.700418583286429e-05, "loss": 0.0121, "step": 40188 }, { "epoch": 0.78, "learning_rate": 3.700353887260706e-05, "loss": 0.0001, "step": 40190 }, { "epoch": 0.78, "learning_rate": 3.7002891912349826e-05, "loss": 0.0074, "step": 40192 }, { "epoch": 0.78, "learning_rate": 3.7002244952092595e-05, "loss": 0.001, "step": 40194 }, { "epoch": 0.78, "learning_rate": 3.7001597991835364e-05, "loss": 0.0, "step": 40196 }, { "epoch": 0.78, "learning_rate": 3.7000951031578126e-05, "loss": 0.0001, "step": 40198 }, { "epoch": 0.78, "learning_rate": 3.70003040713209e-05, "loss": 0.0002, "step": 40200 }, { "epoch": 0.78, "learning_rate": 3.699965711106367e-05, "loss": 0.0004, "step": 40202 }, { "epoch": 0.78, "learning_rate": 3.699901015080644e-05, "loss": 0.0001, "step": 40204 }, { "epoch": 0.78, "learning_rate": 3.699836319054921e-05, "loss": 0.0002, "step": 40206 }, { "epoch": 0.78, "learning_rate": 3.699771623029197e-05, "loss": 0.0005, "step": 40208 }, { "epoch": 0.78, "learning_rate": 3.699706927003475e-05, "loss": 0.0, "step": 40210 }, { "epoch": 0.78, "learning_rate": 3.699642230977751e-05, "loss": 0.0, "step": 40212 }, { "epoch": 0.78, "learning_rate": 3.699577534952028e-05, "loss": 0.0004, "step": 40214 }, { "epoch": 0.78, "learning_rate": 3.699512838926305e-05, "loss": 0.004, "step": 40216 }, { "epoch": 0.78, "learning_rate": 3.699448142900582e-05, "loss": 0.0, "step": 40218 }, { "epoch": 0.78, "learning_rate": 3.6993834468748586e-05, "loss": 0.0002, "step": 40220 }, { "epoch": 0.78, "learning_rate": 3.6993187508491355e-05, "loss": 0.0, "step": 40222 }, { "epoch": 0.78, "learning_rate": 3.6992540548234125e-05, "loss": 0.0002, "step": 40224 }, { "epoch": 0.78, "learning_rate": 3.6991893587976894e-05, "loss": 0.0, "step": 40226 }, { "epoch": 0.78, "learning_rate": 3.699124662771966e-05, "loss": 0.0, "step": 40228 }, { "epoch": 0.78, "learning_rate": 3.6990599667462425e-05, "loss": 0.0058, "step": 40230 }, { "epoch": 0.78, "learning_rate": 3.69899527072052e-05, "loss": 0.0037, "step": 40232 }, { "epoch": 0.78, "learning_rate": 3.698930574694796e-05, "loss": 0.0, "step": 40234 }, { "epoch": 0.78, "learning_rate": 3.698865878669073e-05, "loss": 0.0001, "step": 40236 }, { "epoch": 0.78, "learning_rate": 3.698801182643351e-05, "loss": 0.0052, "step": 40238 }, { "epoch": 0.78, "learning_rate": 3.698736486617627e-05, "loss": 0.0, "step": 40240 }, { "epoch": 0.78, "learning_rate": 3.6986717905919046e-05, "loss": 0.0, "step": 40242 }, { "epoch": 0.78, "learning_rate": 3.698607094566181e-05, "loss": 0.0, "step": 40244 }, { "epoch": 0.78, "learning_rate": 3.698542398540458e-05, "loss": 0.0, "step": 40246 }, { "epoch": 0.78, "learning_rate": 3.698477702514735e-05, "loss": 0.0, "step": 40248 }, { "epoch": 0.78, "learning_rate": 3.6984130064890116e-05, "loss": 0.0024, "step": 40250 }, { "epoch": 0.78, "learning_rate": 3.6983483104632885e-05, "loss": 0.0, "step": 40252 }, { "epoch": 0.78, "learning_rate": 3.6982836144375654e-05, "loss": 0.0, "step": 40254 }, { "epoch": 0.78, "learning_rate": 3.698218918411842e-05, "loss": 0.0, "step": 40256 }, { "epoch": 0.78, "learning_rate": 3.6981542223861185e-05, "loss": 0.0, "step": 40258 }, { "epoch": 0.78, "learning_rate": 3.698089526360396e-05, "loss": 0.0, "step": 40260 }, { "epoch": 0.78, "learning_rate": 3.6980248303346724e-05, "loss": 0.0019, "step": 40262 }, { "epoch": 0.78, "learning_rate": 3.69796013430895e-05, "loss": 0.0, "step": 40264 }, { "epoch": 0.78, "learning_rate": 3.697895438283226e-05, "loss": 0.0, "step": 40266 }, { "epoch": 0.78, "learning_rate": 3.697830742257503e-05, "loss": 0.0, "step": 40268 }, { "epoch": 0.78, "learning_rate": 3.697766046231781e-05, "loss": 0.0, "step": 40270 }, { "epoch": 0.78, "learning_rate": 3.697701350206057e-05, "loss": 0.0, "step": 40272 }, { "epoch": 0.78, "learning_rate": 3.697636654180334e-05, "loss": 0.0, "step": 40274 }, { "epoch": 0.78, "learning_rate": 3.697571958154611e-05, "loss": 0.0, "step": 40276 }, { "epoch": 0.78, "learning_rate": 3.6975072621288876e-05, "loss": 0.0, "step": 40278 }, { "epoch": 0.78, "learning_rate": 3.697442566103164e-05, "loss": 0.0001, "step": 40280 }, { "epoch": 0.78, "learning_rate": 3.6973778700774414e-05, "loss": 0.0015, "step": 40282 }, { "epoch": 0.78, "learning_rate": 3.6973131740517184e-05, "loss": 0.0, "step": 40284 }, { "epoch": 0.78, "learning_rate": 3.697248478025995e-05, "loss": 0.0, "step": 40286 }, { "epoch": 0.78, "learning_rate": 3.697183782000272e-05, "loss": 0.0038, "step": 40288 }, { "epoch": 0.78, "learning_rate": 3.6971190859745484e-05, "loss": 0.0009, "step": 40290 }, { "epoch": 0.78, "learning_rate": 3.697054389948826e-05, "loss": 0.0001, "step": 40292 }, { "epoch": 0.78, "learning_rate": 3.696989693923102e-05, "loss": 0.0004, "step": 40294 }, { "epoch": 0.78, "learning_rate": 3.696924997897379e-05, "loss": 0.0, "step": 40296 }, { "epoch": 0.78, "learning_rate": 3.696860301871656e-05, "loss": 0.0119, "step": 40298 }, { "epoch": 0.78, "learning_rate": 3.696795605845933e-05, "loss": 0.0, "step": 40300 }, { "epoch": 0.78, "learning_rate": 3.6967309098202105e-05, "loss": 0.0133, "step": 40302 }, { "epoch": 0.78, "learning_rate": 3.696666213794487e-05, "loss": 0.0002, "step": 40304 }, { "epoch": 0.78, "learning_rate": 3.696601517768764e-05, "loss": 0.0, "step": 40306 }, { "epoch": 0.78, "learning_rate": 3.6965368217430406e-05, "loss": 0.0001, "step": 40308 }, { "epoch": 0.78, "learning_rate": 3.6964721257173175e-05, "loss": 0.0, "step": 40310 }, { "epoch": 0.78, "learning_rate": 3.696407429691594e-05, "loss": 0.0, "step": 40312 }, { "epoch": 0.78, "learning_rate": 3.696342733665871e-05, "loss": 0.0027, "step": 40314 }, { "epoch": 0.78, "learning_rate": 3.696278037640148e-05, "loss": 0.0184, "step": 40316 }, { "epoch": 0.78, "learning_rate": 3.6962133416144244e-05, "loss": 0.0003, "step": 40318 }, { "epoch": 0.78, "learning_rate": 3.696148645588702e-05, "loss": 0.0001, "step": 40320 }, { "epoch": 0.78, "learning_rate": 3.696083949562978e-05, "loss": 0.0129, "step": 40322 }, { "epoch": 0.78, "learning_rate": 3.696019253537256e-05, "loss": 0.0055, "step": 40324 }, { "epoch": 0.78, "learning_rate": 3.695954557511532e-05, "loss": 0.0, "step": 40326 }, { "epoch": 0.78, "learning_rate": 3.695889861485809e-05, "loss": 0.0009, "step": 40328 }, { "epoch": 0.78, "learning_rate": 3.695825165460086e-05, "loss": 0.0012, "step": 40330 }, { "epoch": 0.78, "learning_rate": 3.695760469434363e-05, "loss": 0.0006, "step": 40332 }, { "epoch": 0.78, "learning_rate": 3.69569577340864e-05, "loss": 0.006, "step": 40334 }, { "epoch": 0.78, "learning_rate": 3.6956310773829166e-05, "loss": 0.0026, "step": 40336 }, { "epoch": 0.78, "learning_rate": 3.6955663813571935e-05, "loss": 0.0, "step": 40338 }, { "epoch": 0.78, "learning_rate": 3.69550168533147e-05, "loss": 0.0001, "step": 40340 }, { "epoch": 0.78, "learning_rate": 3.6954369893057473e-05, "loss": 0.0, "step": 40342 }, { "epoch": 0.78, "learning_rate": 3.6953722932800236e-05, "loss": 0.004, "step": 40344 }, { "epoch": 0.78, "learning_rate": 3.695307597254301e-05, "loss": 0.0, "step": 40346 }, { "epoch": 0.78, "learning_rate": 3.695242901228578e-05, "loss": 0.0, "step": 40348 }, { "epoch": 0.78, "learning_rate": 3.695178205202854e-05, "loss": 0.001, "step": 40350 }, { "epoch": 0.78, "learning_rate": 3.695113509177132e-05, "loss": 0.0001, "step": 40352 }, { "epoch": 0.78, "learning_rate": 3.695048813151408e-05, "loss": 0.0011, "step": 40354 }, { "epoch": 0.78, "learning_rate": 3.694984117125685e-05, "loss": 0.0, "step": 40356 }, { "epoch": 0.78, "learning_rate": 3.694919421099962e-05, "loss": 0.0001, "step": 40358 }, { "epoch": 0.78, "learning_rate": 3.694854725074239e-05, "loss": 0.0001, "step": 40360 }, { "epoch": 0.78, "learning_rate": 3.694790029048516e-05, "loss": 0.0, "step": 40362 }, { "epoch": 0.78, "learning_rate": 3.6947253330227927e-05, "loss": 0.0139, "step": 40364 }, { "epoch": 0.78, "learning_rate": 3.6946606369970696e-05, "loss": 0.0018, "step": 40366 }, { "epoch": 0.78, "learning_rate": 3.6945959409713465e-05, "loss": 0.0038, "step": 40368 }, { "epoch": 0.78, "learning_rate": 3.6945312449456234e-05, "loss": 0.0017, "step": 40370 }, { "epoch": 0.78, "learning_rate": 3.6944665489198996e-05, "loss": 0.012, "step": 40372 }, { "epoch": 0.78, "learning_rate": 3.694401852894177e-05, "loss": 0.0009, "step": 40374 }, { "epoch": 0.78, "learning_rate": 3.6943371568684534e-05, "loss": 0.003, "step": 40376 }, { "epoch": 0.78, "learning_rate": 3.6942724608427303e-05, "loss": 0.0, "step": 40378 }, { "epoch": 0.78, "learning_rate": 3.694207764817007e-05, "loss": 0.0008, "step": 40380 }, { "epoch": 0.78, "learning_rate": 3.694143068791284e-05, "loss": 0.0, "step": 40382 }, { "epoch": 0.78, "learning_rate": 3.694078372765562e-05, "loss": 0.0002, "step": 40384 }, { "epoch": 0.78, "learning_rate": 3.694013676739838e-05, "loss": 0.0, "step": 40386 }, { "epoch": 0.78, "learning_rate": 3.693948980714115e-05, "loss": 0.0223, "step": 40388 }, { "epoch": 0.78, "learning_rate": 3.693884284688392e-05, "loss": 0.0, "step": 40390 }, { "epoch": 0.78, "learning_rate": 3.693819588662669e-05, "loss": 0.0002, "step": 40392 }, { "epoch": 0.78, "learning_rate": 3.6937548926369456e-05, "loss": 0.0004, "step": 40394 }, { "epoch": 0.78, "learning_rate": 3.6936901966112225e-05, "loss": 0.0001, "step": 40396 }, { "epoch": 0.78, "learning_rate": 3.6936255005854994e-05, "loss": 0.0, "step": 40398 }, { "epoch": 0.78, "learning_rate": 3.6935608045597757e-05, "loss": 0.0, "step": 40400 }, { "epoch": 0.78, "learning_rate": 3.693496108534053e-05, "loss": 0.0, "step": 40402 }, { "epoch": 0.78, "learning_rate": 3.6934314125083295e-05, "loss": 0.0001, "step": 40404 }, { "epoch": 0.78, "learning_rate": 3.693366716482607e-05, "loss": 0.0, "step": 40406 }, { "epoch": 0.78, "learning_rate": 3.693302020456883e-05, "loss": 0.0001, "step": 40408 }, { "epoch": 0.78, "learning_rate": 3.69323732443116e-05, "loss": 0.003, "step": 40410 }, { "epoch": 0.78, "learning_rate": 3.693172628405437e-05, "loss": 0.0048, "step": 40412 }, { "epoch": 0.78, "learning_rate": 3.693107932379714e-05, "loss": 0.0001, "step": 40414 }, { "epoch": 0.78, "learning_rate": 3.693043236353991e-05, "loss": 0.0003, "step": 40416 }, { "epoch": 0.78, "learning_rate": 3.692978540328268e-05, "loss": 0.0, "step": 40418 }, { "epoch": 0.78, "learning_rate": 3.692913844302545e-05, "loss": 0.0125, "step": 40420 }, { "epoch": 0.78, "learning_rate": 3.692849148276821e-05, "loss": 0.0, "step": 40422 }, { "epoch": 0.78, "learning_rate": 3.6927844522510986e-05, "loss": 0.0001, "step": 40424 }, { "epoch": 0.78, "learning_rate": 3.6927197562253755e-05, "loss": 0.0086, "step": 40426 }, { "epoch": 0.78, "learning_rate": 3.6926550601996524e-05, "loss": 0.021, "step": 40428 }, { "epoch": 0.78, "learning_rate": 3.692590364173929e-05, "loss": 0.0125, "step": 40430 }, { "epoch": 0.78, "learning_rate": 3.6925256681482055e-05, "loss": 0.0004, "step": 40432 }, { "epoch": 0.78, "learning_rate": 3.692460972122483e-05, "loss": 0.0034, "step": 40434 }, { "epoch": 0.78, "learning_rate": 3.692396276096759e-05, "loss": 0.0, "step": 40436 }, { "epoch": 0.78, "learning_rate": 3.692331580071036e-05, "loss": 0.0, "step": 40438 }, { "epoch": 0.78, "learning_rate": 3.692266884045313e-05, "loss": 0.0, "step": 40440 }, { "epoch": 0.78, "learning_rate": 3.69220218801959e-05, "loss": 0.0002, "step": 40442 }, { "epoch": 0.78, "learning_rate": 3.692137491993867e-05, "loss": 0.0, "step": 40444 }, { "epoch": 0.79, "learning_rate": 3.692072795968144e-05, "loss": 0.0002, "step": 40446 }, { "epoch": 0.79, "learning_rate": 3.692008099942421e-05, "loss": 0.0011, "step": 40448 }, { "epoch": 0.79, "learning_rate": 3.691943403916698e-05, "loss": 0.0065, "step": 40450 }, { "epoch": 0.79, "learning_rate": 3.6918787078909746e-05, "loss": 0.0, "step": 40452 }, { "epoch": 0.79, "learning_rate": 3.691814011865251e-05, "loss": 0.0, "step": 40454 }, { "epoch": 0.79, "learning_rate": 3.6917493158395284e-05, "loss": 0.0, "step": 40456 }, { "epoch": 0.79, "learning_rate": 3.6916846198138046e-05, "loss": 0.002, "step": 40458 }, { "epoch": 0.79, "learning_rate": 3.6916199237880816e-05, "loss": 0.0001, "step": 40460 }, { "epoch": 0.79, "learning_rate": 3.691555227762359e-05, "loss": 0.0009, "step": 40462 }, { "epoch": 0.79, "learning_rate": 3.6914905317366354e-05, "loss": 0.0, "step": 40464 }, { "epoch": 0.79, "learning_rate": 3.691425835710913e-05, "loss": 0.0141, "step": 40466 }, { "epoch": 0.79, "learning_rate": 3.691393487698051e-05, "loss": 0.0679, "step": 40468 }, { "epoch": 0.79, "learning_rate": 3.6913287916723276e-05, "loss": 0.0258, "step": 40470 }, { "epoch": 0.79, "learning_rate": 3.6912640956466046e-05, "loss": 0.0, "step": 40472 }, { "epoch": 0.79, "learning_rate": 3.6911993996208815e-05, "loss": 0.0001, "step": 40474 }, { "epoch": 0.79, "learning_rate": 3.6911347035951584e-05, "loss": 0.0, "step": 40476 }, { "epoch": 0.79, "learning_rate": 3.691070007569435e-05, "loss": 0.0003, "step": 40478 }, { "epoch": 0.79, "learning_rate": 3.691005311543712e-05, "loss": 0.0, "step": 40480 }, { "epoch": 0.79, "learning_rate": 3.6909406155179884e-05, "loss": 0.0051, "step": 40482 }, { "epoch": 0.79, "learning_rate": 3.690875919492266e-05, "loss": 0.0013, "step": 40484 }, { "epoch": 0.79, "learning_rate": 3.690811223466543e-05, "loss": 0.0, "step": 40486 }, { "epoch": 0.79, "learning_rate": 3.690746527440819e-05, "loss": 0.0005, "step": 40488 }, { "epoch": 0.79, "learning_rate": 3.690681831415097e-05, "loss": 0.0034, "step": 40490 }, { "epoch": 0.79, "learning_rate": 3.690617135389373e-05, "loss": 0.0, "step": 40492 }, { "epoch": 0.79, "learning_rate": 3.69055243936365e-05, "loss": 0.0007, "step": 40494 }, { "epoch": 0.79, "learning_rate": 3.690487743337927e-05, "loss": 0.0, "step": 40496 }, { "epoch": 0.79, "learning_rate": 3.690423047312204e-05, "loss": 0.0004, "step": 40498 }, { "epoch": 0.79, "learning_rate": 3.6903583512864806e-05, "loss": 0.0, "step": 40500 }, { "epoch": 0.79, "learning_rate": 3.6902936552607575e-05, "loss": 0.0016, "step": 40502 }, { "epoch": 0.79, "learning_rate": 3.6902289592350344e-05, "loss": 0.0002, "step": 40504 }, { "epoch": 0.79, "learning_rate": 3.690164263209311e-05, "loss": 0.0, "step": 40506 }, { "epoch": 0.79, "learning_rate": 3.690099567183588e-05, "loss": 0.0, "step": 40508 }, { "epoch": 0.79, "learning_rate": 3.6900348711578645e-05, "loss": 0.0084, "step": 40510 }, { "epoch": 0.79, "learning_rate": 3.689970175132142e-05, "loss": 0.0, "step": 40512 }, { "epoch": 0.79, "learning_rate": 3.689905479106418e-05, "loss": 0.0407, "step": 40514 }, { "epoch": 0.79, "learning_rate": 3.689840783080696e-05, "loss": 0.0, "step": 40516 }, { "epoch": 0.79, "learning_rate": 3.689776087054973e-05, "loss": 0.0, "step": 40518 }, { "epoch": 0.79, "learning_rate": 3.689711391029249e-05, "loss": 0.0, "step": 40520 }, { "epoch": 0.79, "learning_rate": 3.6896466950035266e-05, "loss": 0.0402, "step": 40522 }, { "epoch": 0.79, "learning_rate": 3.689581998977803e-05, "loss": 0.0, "step": 40524 }, { "epoch": 0.79, "learning_rate": 3.68951730295208e-05, "loss": 0.0001, "step": 40526 }, { "epoch": 0.79, "learning_rate": 3.6894526069263566e-05, "loss": 0.0001, "step": 40528 }, { "epoch": 0.79, "learning_rate": 3.6893879109006335e-05, "loss": 0.0, "step": 40530 }, { "epoch": 0.79, "learning_rate": 3.6893232148749104e-05, "loss": 0.0016, "step": 40532 }, { "epoch": 0.79, "learning_rate": 3.6892585188491874e-05, "loss": 0.0026, "step": 40534 }, { "epoch": 0.79, "learning_rate": 3.689193822823464e-05, "loss": 0.0125, "step": 40536 }, { "epoch": 0.79, "learning_rate": 3.689129126797741e-05, "loss": 0.0002, "step": 40538 }, { "epoch": 0.79, "learning_rate": 3.689064430772018e-05, "loss": 0.0, "step": 40540 }, { "epoch": 0.79, "learning_rate": 3.688999734746294e-05, "loss": 0.0041, "step": 40542 }, { "epoch": 0.79, "learning_rate": 3.688935038720572e-05, "loss": 0.0049, "step": 40544 }, { "epoch": 0.79, "learning_rate": 3.688870342694848e-05, "loss": 0.0001, "step": 40546 }, { "epoch": 0.79, "learning_rate": 3.688805646669125e-05, "loss": 0.0, "step": 40548 }, { "epoch": 0.79, "learning_rate": 3.6887409506434026e-05, "loss": 0.0, "step": 40550 }, { "epoch": 0.79, "learning_rate": 3.688676254617679e-05, "loss": 0.0, "step": 40552 }, { "epoch": 0.79, "learning_rate": 3.688611558591956e-05, "loss": 0.0, "step": 40554 }, { "epoch": 0.79, "learning_rate": 3.688546862566233e-05, "loss": 0.0001, "step": 40556 }, { "epoch": 0.79, "learning_rate": 3.6884821665405096e-05, "loss": 0.0001, "step": 40558 }, { "epoch": 0.79, "learning_rate": 3.6884174705147865e-05, "loss": 0.004, "step": 40560 }, { "epoch": 0.79, "learning_rate": 3.6883527744890634e-05, "loss": 0.0111, "step": 40562 }, { "epoch": 0.79, "learning_rate": 3.68828807846334e-05, "loss": 0.0001, "step": 40564 }, { "epoch": 0.79, "learning_rate": 3.688223382437617e-05, "loss": 0.0001, "step": 40566 }, { "epoch": 0.79, "learning_rate": 3.688158686411894e-05, "loss": 0.0001, "step": 40568 }, { "epoch": 0.79, "learning_rate": 3.6880939903861704e-05, "loss": 0.025, "step": 40570 }, { "epoch": 0.79, "learning_rate": 3.688029294360448e-05, "loss": 0.0, "step": 40572 }, { "epoch": 0.79, "learning_rate": 3.687964598334724e-05, "loss": 0.0, "step": 40574 }, { "epoch": 0.79, "learning_rate": 3.687899902309001e-05, "loss": 0.0, "step": 40576 }, { "epoch": 0.79, "learning_rate": 3.687835206283278e-05, "loss": 0.0005, "step": 40578 }, { "epoch": 0.79, "learning_rate": 3.687770510257555e-05, "loss": 0.0001, "step": 40580 }, { "epoch": 0.79, "learning_rate": 3.6877058142318325e-05, "loss": 0.0, "step": 40582 }, { "epoch": 0.79, "learning_rate": 3.687641118206109e-05, "loss": 0.0004, "step": 40584 }, { "epoch": 0.79, "learning_rate": 3.6875764221803856e-05, "loss": 0.0014, "step": 40586 }, { "epoch": 0.79, "learning_rate": 3.6875117261546625e-05, "loss": 0.0004, "step": 40588 }, { "epoch": 0.79, "learning_rate": 3.6874470301289394e-05, "loss": 0.0036, "step": 40590 }, { "epoch": 0.79, "learning_rate": 3.687382334103216e-05, "loss": 0.0, "step": 40592 }, { "epoch": 0.79, "learning_rate": 3.687317638077493e-05, "loss": 0.0001, "step": 40594 }, { "epoch": 0.79, "learning_rate": 3.68725294205177e-05, "loss": 0.0, "step": 40596 }, { "epoch": 0.79, "learning_rate": 3.687188246026047e-05, "loss": 0.0001, "step": 40598 }, { "epoch": 0.79, "learning_rate": 3.687123550000324e-05, "loss": 0.0114, "step": 40600 }, { "epoch": 0.79, "learning_rate": 3.6870588539746e-05, "loss": 0.0, "step": 40602 }, { "epoch": 0.79, "learning_rate": 3.686994157948878e-05, "loss": 0.0101, "step": 40604 }, { "epoch": 0.79, "learning_rate": 3.686929461923154e-05, "loss": 0.0, "step": 40606 }, { "epoch": 0.79, "learning_rate": 3.686864765897431e-05, "loss": 0.0058, "step": 40608 }, { "epoch": 0.79, "learning_rate": 3.686800069871708e-05, "loss": 0.0, "step": 40610 }, { "epoch": 0.79, "learning_rate": 3.686735373845985e-05, "loss": 0.0, "step": 40612 }, { "epoch": 0.79, "learning_rate": 3.686670677820262e-05, "loss": 0.0001, "step": 40614 }, { "epoch": 0.79, "learning_rate": 3.6866059817945386e-05, "loss": 0.0, "step": 40616 }, { "epoch": 0.79, "learning_rate": 3.6865412857688155e-05, "loss": 0.0001, "step": 40618 }, { "epoch": 0.79, "learning_rate": 3.6864765897430924e-05, "loss": 0.0, "step": 40620 }, { "epoch": 0.79, "learning_rate": 3.686411893717369e-05, "loss": 0.0, "step": 40622 }, { "epoch": 0.79, "learning_rate": 3.6863471976916455e-05, "loss": 0.0, "step": 40624 }, { "epoch": 0.79, "learning_rate": 3.686282501665923e-05, "loss": 0.0, "step": 40626 }, { "epoch": 0.79, "learning_rate": 3.6862178056402e-05, "loss": 0.0, "step": 40628 }, { "epoch": 0.79, "learning_rate": 3.686153109614476e-05, "loss": 0.0, "step": 40630 }, { "epoch": 0.79, "learning_rate": 3.686088413588754e-05, "loss": 0.0007, "step": 40632 }, { "epoch": 0.79, "learning_rate": 3.68602371756303e-05, "loss": 0.0, "step": 40634 }, { "epoch": 0.79, "learning_rate": 3.685959021537307e-05, "loss": 0.0001, "step": 40636 }, { "epoch": 0.79, "learning_rate": 3.685894325511584e-05, "loss": 0.0001, "step": 40638 }, { "epoch": 0.79, "learning_rate": 3.685829629485861e-05, "loss": 0.0009, "step": 40640 }, { "epoch": 0.79, "learning_rate": 3.685764933460138e-05, "loss": 0.0005, "step": 40642 }, { "epoch": 0.79, "learning_rate": 3.6857002374344146e-05, "loss": 0.0006, "step": 40644 }, { "epoch": 0.79, "learning_rate": 3.6856355414086915e-05, "loss": 0.0001, "step": 40646 }, { "epoch": 0.79, "learning_rate": 3.6855708453829684e-05, "loss": 0.0, "step": 40648 }, { "epoch": 0.79, "learning_rate": 3.685506149357245e-05, "loss": 0.0, "step": 40650 }, { "epoch": 0.79, "learning_rate": 3.6854414533315216e-05, "loss": 0.0043, "step": 40652 }, { "epoch": 0.79, "learning_rate": 3.685376757305799e-05, "loss": 0.0, "step": 40654 }, { "epoch": 0.79, "learning_rate": 3.6853120612800754e-05, "loss": 0.0, "step": 40656 }, { "epoch": 0.79, "learning_rate": 3.685247365254352e-05, "loss": 0.0, "step": 40658 }, { "epoch": 0.79, "learning_rate": 3.685182669228629e-05, "loss": 0.0001, "step": 40660 }, { "epoch": 0.79, "learning_rate": 3.685117973202906e-05, "loss": 0.0, "step": 40662 }, { "epoch": 0.79, "learning_rate": 3.685053277177184e-05, "loss": 0.0, "step": 40664 }, { "epoch": 0.79, "learning_rate": 3.68498858115146e-05, "loss": 0.0042, "step": 40666 }, { "epoch": 0.79, "learning_rate": 3.684923885125737e-05, "loss": 0.0002, "step": 40668 }, { "epoch": 0.79, "learning_rate": 3.684859189100014e-05, "loss": 0.0, "step": 40670 }, { "epoch": 0.79, "learning_rate": 3.6847944930742907e-05, "loss": 0.0, "step": 40672 }, { "epoch": 0.79, "learning_rate": 3.6847297970485676e-05, "loss": 0.0001, "step": 40674 }, { "epoch": 0.79, "learning_rate": 3.6846651010228445e-05, "loss": 0.0, "step": 40676 }, { "epoch": 0.79, "learning_rate": 3.6846004049971214e-05, "loss": 0.0, "step": 40678 }, { "epoch": 0.79, "learning_rate": 3.684535708971398e-05, "loss": 0.0, "step": 40680 }, { "epoch": 0.79, "learning_rate": 3.684471012945675e-05, "loss": 0.0, "step": 40682 }, { "epoch": 0.79, "learning_rate": 3.6844063169199514e-05, "loss": 0.0001, "step": 40684 }, { "epoch": 0.79, "learning_rate": 3.684341620894229e-05, "loss": 0.0003, "step": 40686 }, { "epoch": 0.79, "learning_rate": 3.684276924868505e-05, "loss": 0.0, "step": 40688 }, { "epoch": 0.79, "learning_rate": 3.684212228842782e-05, "loss": 0.0, "step": 40690 }, { "epoch": 0.79, "learning_rate": 3.684147532817059e-05, "loss": 0.0002, "step": 40692 }, { "epoch": 0.79, "learning_rate": 3.684082836791336e-05, "loss": 0.0, "step": 40694 }, { "epoch": 0.79, "learning_rate": 3.684018140765613e-05, "loss": 0.0, "step": 40696 }, { "epoch": 0.79, "learning_rate": 3.68395344473989e-05, "loss": 0.0001, "step": 40698 }, { "epoch": 0.79, "learning_rate": 3.683888748714167e-05, "loss": 0.008, "step": 40700 }, { "epoch": 0.79, "learning_rate": 3.6838240526884436e-05, "loss": 0.0, "step": 40702 }, { "epoch": 0.79, "learning_rate": 3.6837593566627205e-05, "loss": 0.0003, "step": 40704 }, { "epoch": 0.79, "learning_rate": 3.6836946606369974e-05, "loss": 0.0002, "step": 40706 }, { "epoch": 0.79, "learning_rate": 3.683629964611274e-05, "loss": 0.0, "step": 40708 }, { "epoch": 0.79, "learning_rate": 3.683565268585551e-05, "loss": 0.0001, "step": 40710 }, { "epoch": 0.79, "learning_rate": 3.6835005725598275e-05, "loss": 0.0, "step": 40712 }, { "epoch": 0.79, "learning_rate": 3.683435876534105e-05, "loss": 0.0, "step": 40714 }, { "epoch": 0.79, "learning_rate": 3.683371180508381e-05, "loss": 0.0, "step": 40716 }, { "epoch": 0.79, "learning_rate": 3.683306484482658e-05, "loss": 0.0, "step": 40718 }, { "epoch": 0.79, "learning_rate": 3.683241788456935e-05, "loss": 0.0, "step": 40720 }, { "epoch": 0.79, "learning_rate": 3.683177092431212e-05, "loss": 0.0, "step": 40722 }, { "epoch": 0.79, "learning_rate": 3.683112396405489e-05, "loss": 0.0001, "step": 40724 }, { "epoch": 0.79, "learning_rate": 3.683047700379766e-05, "loss": 0.0, "step": 40726 }, { "epoch": 0.79, "learning_rate": 3.682983004354043e-05, "loss": 0.0, "step": 40728 }, { "epoch": 0.79, "learning_rate": 3.6829183083283196e-05, "loss": 0.0, "step": 40730 }, { "epoch": 0.79, "learning_rate": 3.6828536123025965e-05, "loss": 0.0001, "step": 40732 }, { "epoch": 0.79, "learning_rate": 3.682788916276873e-05, "loss": 0.0001, "step": 40734 }, { "epoch": 0.79, "learning_rate": 3.6827242202511504e-05, "loss": 0.006, "step": 40736 }, { "epoch": 0.79, "learning_rate": 3.6826595242254266e-05, "loss": 0.016, "step": 40738 }, { "epoch": 0.79, "learning_rate": 3.682594828199704e-05, "loss": 0.0, "step": 40740 }, { "epoch": 0.79, "learning_rate": 3.682530132173981e-05, "loss": 0.0144, "step": 40742 }, { "epoch": 0.79, "learning_rate": 3.682465436148257e-05, "loss": 0.0007, "step": 40744 }, { "epoch": 0.79, "learning_rate": 3.682400740122535e-05, "loss": 0.0001, "step": 40746 }, { "epoch": 0.79, "learning_rate": 3.682336044096811e-05, "loss": 0.0153, "step": 40748 }, { "epoch": 0.79, "learning_rate": 3.682271348071088e-05, "loss": 0.0003, "step": 40750 }, { "epoch": 0.79, "learning_rate": 3.682206652045365e-05, "loss": 0.0002, "step": 40752 }, { "epoch": 0.79, "learning_rate": 3.682141956019642e-05, "loss": 0.0, "step": 40754 }, { "epoch": 0.79, "learning_rate": 3.682077259993919e-05, "loss": 0.0095, "step": 40756 }, { "epoch": 0.79, "learning_rate": 3.682012563968196e-05, "loss": 0.0001, "step": 40758 }, { "epoch": 0.79, "learning_rate": 3.6819478679424726e-05, "loss": 0.0057, "step": 40760 }, { "epoch": 0.79, "learning_rate": 3.6818831719167495e-05, "loss": 0.0056, "step": 40762 }, { "epoch": 0.79, "learning_rate": 3.6818184758910264e-05, "loss": 0.0071, "step": 40764 }, { "epoch": 0.79, "learning_rate": 3.6817537798653026e-05, "loss": 0.0098, "step": 40766 }, { "epoch": 0.79, "learning_rate": 3.68168908383958e-05, "loss": 0.0074, "step": 40768 }, { "epoch": 0.79, "learning_rate": 3.6816243878138565e-05, "loss": 0.0001, "step": 40770 }, { "epoch": 0.79, "learning_rate": 3.6815596917881334e-05, "loss": 0.0018, "step": 40772 }, { "epoch": 0.79, "learning_rate": 3.681494995762411e-05, "loss": 0.0, "step": 40774 }, { "epoch": 0.79, "learning_rate": 3.681430299736687e-05, "loss": 0.0, "step": 40776 }, { "epoch": 0.79, "learning_rate": 3.681365603710964e-05, "loss": 0.0039, "step": 40778 }, { "epoch": 0.79, "learning_rate": 3.681300907685241e-05, "loss": 0.0, "step": 40780 }, { "epoch": 0.79, "learning_rate": 3.681236211659518e-05, "loss": 0.0055, "step": 40782 }, { "epoch": 0.79, "learning_rate": 3.681171515633795e-05, "loss": 0.0, "step": 40784 }, { "epoch": 0.79, "learning_rate": 3.681106819608072e-05, "loss": 0.0, "step": 40786 }, { "epoch": 0.79, "learning_rate": 3.6810421235823486e-05, "loss": 0.0, "step": 40788 }, { "epoch": 0.79, "learning_rate": 3.6809774275566255e-05, "loss": 0.0017, "step": 40790 }, { "epoch": 0.79, "learning_rate": 3.6809127315309024e-05, "loss": 0.0001, "step": 40792 }, { "epoch": 0.79, "learning_rate": 3.680848035505179e-05, "loss": 0.0003, "step": 40794 }, { "epoch": 0.79, "learning_rate": 3.680783339479456e-05, "loss": 0.0, "step": 40796 }, { "epoch": 0.79, "learning_rate": 3.6807186434537325e-05, "loss": 0.0, "step": 40798 }, { "epoch": 0.79, "learning_rate": 3.6806539474280094e-05, "loss": 0.0003, "step": 40800 }, { "epoch": 0.79, "learning_rate": 3.680589251402286e-05, "loss": 0.0, "step": 40802 }, { "epoch": 0.79, "learning_rate": 3.680524555376563e-05, "loss": 0.0, "step": 40804 }, { "epoch": 0.79, "learning_rate": 3.680459859350841e-05, "loss": 0.0002, "step": 40806 }, { "epoch": 0.79, "learning_rate": 3.680395163325117e-05, "loss": 0.0014, "step": 40808 }, { "epoch": 0.79, "learning_rate": 3.680330467299394e-05, "loss": 0.0, "step": 40810 }, { "epoch": 0.79, "learning_rate": 3.680265771273671e-05, "loss": 0.0031, "step": 40812 }, { "epoch": 0.79, "learning_rate": 3.680201075247948e-05, "loss": 0.0014, "step": 40814 }, { "epoch": 0.79, "learning_rate": 3.680136379222224e-05, "loss": 0.0001, "step": 40816 }, { "epoch": 0.79, "learning_rate": 3.6800716831965016e-05, "loss": 0.0, "step": 40818 }, { "epoch": 0.79, "learning_rate": 3.6800069871707785e-05, "loss": 0.0, "step": 40820 }, { "epoch": 0.79, "learning_rate": 3.6799422911450554e-05, "loss": 0.0, "step": 40822 }, { "epoch": 0.79, "learning_rate": 3.679877595119332e-05, "loss": 0.0001, "step": 40824 }, { "epoch": 0.79, "learning_rate": 3.6798128990936085e-05, "loss": 0.004, "step": 40826 }, { "epoch": 0.79, "learning_rate": 3.679748203067886e-05, "loss": 0.0, "step": 40828 }, { "epoch": 0.79, "learning_rate": 3.6796835070421624e-05, "loss": 0.0041, "step": 40830 }, { "epoch": 0.79, "learning_rate": 3.679618811016439e-05, "loss": 0.0, "step": 40832 }, { "epoch": 0.79, "learning_rate": 3.679554114990716e-05, "loss": 0.0, "step": 40834 }, { "epoch": 0.79, "learning_rate": 3.679489418964993e-05, "loss": 0.0085, "step": 40836 }, { "epoch": 0.79, "learning_rate": 3.67942472293927e-05, "loss": 0.0124, "step": 40838 }, { "epoch": 0.79, "learning_rate": 3.679360026913547e-05, "loss": 0.0001, "step": 40840 }, { "epoch": 0.79, "learning_rate": 3.679295330887824e-05, "loss": 0.0, "step": 40842 }, { "epoch": 0.79, "learning_rate": 3.679230634862101e-05, "loss": 0.0, "step": 40844 }, { "epoch": 0.79, "learning_rate": 3.6791659388363776e-05, "loss": 0.0, "step": 40846 }, { "epoch": 0.79, "learning_rate": 3.679101242810654e-05, "loss": 0.0, "step": 40848 }, { "epoch": 0.79, "learning_rate": 3.6790365467849314e-05, "loss": 0.0016, "step": 40850 }, { "epoch": 0.79, "learning_rate": 3.6789718507592083e-05, "loss": 0.0067, "step": 40852 }, { "epoch": 0.79, "learning_rate": 3.6789071547334846e-05, "loss": 0.0002, "step": 40854 }, { "epoch": 0.79, "learning_rate": 3.678842458707762e-05, "loss": 0.0001, "step": 40856 }, { "epoch": 0.79, "learning_rate": 3.6787777626820384e-05, "loss": 0.0001, "step": 40858 }, { "epoch": 0.79, "learning_rate": 3.678713066656315e-05, "loss": 0.0001, "step": 40860 }, { "epoch": 0.79, "learning_rate": 3.678648370630592e-05, "loss": 0.0001, "step": 40862 }, { "epoch": 0.79, "learning_rate": 3.678583674604869e-05, "loss": 0.0197, "step": 40864 }, { "epoch": 0.79, "learning_rate": 3.678518978579146e-05, "loss": 0.0, "step": 40866 }, { "epoch": 0.79, "learning_rate": 3.678454282553423e-05, "loss": 0.0, "step": 40868 }, { "epoch": 0.79, "learning_rate": 3.6783895865277e-05, "loss": 0.0002, "step": 40870 }, { "epoch": 0.79, "learning_rate": 3.678324890501977e-05, "loss": 0.0019, "step": 40872 }, { "epoch": 0.79, "learning_rate": 3.6782601944762537e-05, "loss": 0.0, "step": 40874 }, { "epoch": 0.79, "learning_rate": 3.67819549845053e-05, "loss": 0.0001, "step": 40876 }, { "epoch": 0.79, "learning_rate": 3.6781308024248075e-05, "loss": 0.0187, "step": 40878 }, { "epoch": 0.79, "learning_rate": 3.678066106399084e-05, "loss": 0.0008, "step": 40880 }, { "epoch": 0.79, "learning_rate": 3.678001410373361e-05, "loss": 0.0, "step": 40882 }, { "epoch": 0.79, "learning_rate": 3.677936714347638e-05, "loss": 0.0001, "step": 40884 }, { "epoch": 0.79, "learning_rate": 3.6778720183219144e-05, "loss": 0.0005, "step": 40886 }, { "epoch": 0.79, "learning_rate": 3.677807322296192e-05, "loss": 0.0063, "step": 40888 }, { "epoch": 0.79, "learning_rate": 3.677742626270468e-05, "loss": 0.0112, "step": 40890 }, { "epoch": 0.79, "learning_rate": 3.677677930244745e-05, "loss": 0.0, "step": 40892 }, { "epoch": 0.79, "learning_rate": 3.677613234219022e-05, "loss": 0.0, "step": 40894 }, { "epoch": 0.79, "learning_rate": 3.677548538193299e-05, "loss": 0.0, "step": 40896 }, { "epoch": 0.79, "learning_rate": 3.677483842167576e-05, "loss": 0.0, "step": 40898 }, { "epoch": 0.79, "learning_rate": 3.677419146141853e-05, "loss": 0.0, "step": 40900 }, { "epoch": 0.79, "learning_rate": 3.67735445011613e-05, "loss": 0.0, "step": 40902 }, { "epoch": 0.79, "learning_rate": 3.6772897540904066e-05, "loss": 0.0829, "step": 40904 }, { "epoch": 0.79, "learning_rate": 3.6772250580646835e-05, "loss": 0.0021, "step": 40906 }, { "epoch": 0.79, "learning_rate": 3.67716036203896e-05, "loss": 0.0, "step": 40908 }, { "epoch": 0.79, "learning_rate": 3.677095666013237e-05, "loss": 0.0001, "step": 40910 }, { "epoch": 0.79, "learning_rate": 3.6770309699875136e-05, "loss": 0.0, "step": 40912 }, { "epoch": 0.79, "learning_rate": 3.6769662739617905e-05, "loss": 0.0, "step": 40914 }, { "epoch": 0.79, "learning_rate": 3.6769015779360674e-05, "loss": 0.0, "step": 40916 }, { "epoch": 0.79, "learning_rate": 3.676836881910344e-05, "loss": 0.0059, "step": 40918 }, { "epoch": 0.79, "learning_rate": 3.676772185884621e-05, "loss": 0.0, "step": 40920 }, { "epoch": 0.79, "learning_rate": 3.676707489858898e-05, "loss": 0.0001, "step": 40922 }, { "epoch": 0.79, "learning_rate": 3.676642793833175e-05, "loss": 0.0, "step": 40924 }, { "epoch": 0.79, "learning_rate": 3.676578097807452e-05, "loss": 0.0, "step": 40926 }, { "epoch": 0.79, "learning_rate": 3.676513401781729e-05, "loss": 0.0, "step": 40928 }, { "epoch": 0.79, "learning_rate": 3.676448705756006e-05, "loss": 0.0001, "step": 40930 }, { "epoch": 0.79, "learning_rate": 3.6763840097302826e-05, "loss": 0.0005, "step": 40932 }, { "epoch": 0.79, "learning_rate": 3.6763193137045596e-05, "loss": 0.0032, "step": 40934 }, { "epoch": 0.79, "learning_rate": 3.676254617678836e-05, "loss": 0.0, "step": 40936 }, { "epoch": 0.79, "learning_rate": 3.6761899216531134e-05, "loss": 0.0, "step": 40938 }, { "epoch": 0.79, "learning_rate": 3.6761252256273896e-05, "loss": 0.0, "step": 40940 }, { "epoch": 0.79, "learning_rate": 3.6760605296016665e-05, "loss": 0.0001, "step": 40942 }, { "epoch": 0.79, "learning_rate": 3.6759958335759434e-05, "loss": 0.0, "step": 40944 }, { "epoch": 0.79, "learning_rate": 3.67593113755022e-05, "loss": 0.0, "step": 40946 }, { "epoch": 0.79, "learning_rate": 3.675866441524497e-05, "loss": 0.0042, "step": 40948 }, { "epoch": 0.79, "learning_rate": 3.675801745498774e-05, "loss": 0.0013, "step": 40950 }, { "epoch": 0.79, "learning_rate": 3.675737049473051e-05, "loss": 0.0, "step": 40952 }, { "epoch": 0.79, "learning_rate": 3.675672353447328e-05, "loss": 0.0004, "step": 40954 }, { "epoch": 0.79, "learning_rate": 3.675607657421605e-05, "loss": 0.0, "step": 40956 }, { "epoch": 0.79, "learning_rate": 3.675542961395881e-05, "loss": 0.0026, "step": 40958 }, { "epoch": 0.79, "learning_rate": 3.675478265370159e-05, "loss": 0.0006, "step": 40960 }, { "epoch": 0.8, "learning_rate": 3.675413569344435e-05, "loss": 0.0001, "step": 40962 }, { "epoch": 0.8, "learning_rate": 3.6753488733187125e-05, "loss": 0.0015, "step": 40964 }, { "epoch": 0.8, "learning_rate": 3.6752841772929894e-05, "loss": 0.001, "step": 40966 }, { "epoch": 0.8, "learning_rate": 3.6752194812672656e-05, "loss": 0.0086, "step": 40968 }, { "epoch": 0.8, "learning_rate": 3.675154785241543e-05, "loss": 0.0, "step": 40970 }, { "epoch": 0.8, "learning_rate": 3.6750900892158195e-05, "loss": 0.0, "step": 40972 }, { "epoch": 0.8, "learning_rate": 3.6750253931900964e-05, "loss": 0.0052, "step": 40974 }, { "epoch": 0.8, "learning_rate": 3.674960697164373e-05, "loss": 0.0, "step": 40976 }, { "epoch": 0.8, "learning_rate": 3.67489600113865e-05, "loss": 0.0063, "step": 40978 }, { "epoch": 0.8, "learning_rate": 3.674831305112927e-05, "loss": 0.0, "step": 40980 }, { "epoch": 0.8, "learning_rate": 3.674766609087204e-05, "loss": 0.0, "step": 40982 }, { "epoch": 0.8, "learning_rate": 3.674701913061481e-05, "loss": 0.0, "step": 40984 }, { "epoch": 0.8, "learning_rate": 3.674637217035758e-05, "loss": 0.0001, "step": 40986 }, { "epoch": 0.8, "learning_rate": 3.674572521010035e-05, "loss": 0.0, "step": 40988 }, { "epoch": 0.8, "learning_rate": 3.674507824984311e-05, "loss": 0.0033, "step": 40990 }, { "epoch": 0.8, "learning_rate": 3.6744431289585885e-05, "loss": 0.0075, "step": 40992 }, { "epoch": 0.8, "learning_rate": 3.674378432932865e-05, "loss": 0.0008, "step": 40994 }, { "epoch": 0.8, "learning_rate": 3.674313736907142e-05, "loss": 0.0, "step": 40996 }, { "epoch": 0.8, "learning_rate": 3.674249040881419e-05, "loss": 0.0, "step": 40998 }, { "epoch": 0.8, "learning_rate": 3.6741843448556955e-05, "loss": 0.0001, "step": 41000 }, { "epoch": 0.8, "learning_rate": 3.6741196488299724e-05, "loss": 0.0009, "step": 41002 }, { "epoch": 0.8, "learning_rate": 3.674054952804249e-05, "loss": 0.0, "step": 41004 }, { "epoch": 0.8, "learning_rate": 3.673990256778526e-05, "loss": 0.0, "step": 41006 }, { "epoch": 0.8, "learning_rate": 3.673925560752803e-05, "loss": 0.0001, "step": 41008 }, { "epoch": 0.8, "learning_rate": 3.67386086472708e-05, "loss": 0.0001, "step": 41010 }, { "epoch": 0.8, "learning_rate": 3.673796168701357e-05, "loss": 0.0, "step": 41012 }, { "epoch": 0.8, "learning_rate": 3.673731472675634e-05, "loss": 0.0, "step": 41014 }, { "epoch": 0.8, "learning_rate": 3.673666776649911e-05, "loss": 0.0077, "step": 41016 }, { "epoch": 0.8, "learning_rate": 3.673602080624187e-05, "loss": 0.0001, "step": 41018 }, { "epoch": 0.8, "learning_rate": 3.6735373845984646e-05, "loss": 0.0001, "step": 41020 }, { "epoch": 0.8, "learning_rate": 3.673472688572741e-05, "loss": 0.0, "step": 41022 }, { "epoch": 0.8, "learning_rate": 3.6734079925470184e-05, "loss": 0.0001, "step": 41024 }, { "epoch": 0.8, "learning_rate": 3.6733432965212946e-05, "loss": 0.0, "step": 41026 }, { "epoch": 0.8, "learning_rate": 3.6732786004955715e-05, "loss": 0.0111, "step": 41028 }, { "epoch": 0.8, "learning_rate": 3.673213904469849e-05, "loss": 0.0001, "step": 41030 }, { "epoch": 0.8, "learning_rate": 3.6731492084441254e-05, "loss": 0.0306, "step": 41032 }, { "epoch": 0.8, "learning_rate": 3.673084512418402e-05, "loss": 0.0005, "step": 41034 }, { "epoch": 0.8, "learning_rate": 3.673019816392679e-05, "loss": 0.0076, "step": 41036 }, { "epoch": 0.8, "learning_rate": 3.672955120366956e-05, "loss": 0.0, "step": 41038 }, { "epoch": 0.8, "learning_rate": 3.672890424341232e-05, "loss": 0.0004, "step": 41040 }, { "epoch": 0.8, "learning_rate": 3.67282572831551e-05, "loss": 0.0003, "step": 41042 }, { "epoch": 0.8, "learning_rate": 3.672761032289787e-05, "loss": 0.0, "step": 41044 }, { "epoch": 0.8, "learning_rate": 3.672696336264064e-05, "loss": 0.0294, "step": 41046 }, { "epoch": 0.8, "learning_rate": 3.6726316402383406e-05, "loss": 0.0032, "step": 41048 }, { "epoch": 0.8, "learning_rate": 3.672566944212617e-05, "loss": 0.0008, "step": 41050 }, { "epoch": 0.8, "learning_rate": 3.6725022481868944e-05, "loss": 0.0001, "step": 41052 }, { "epoch": 0.8, "learning_rate": 3.672437552161171e-05, "loss": 0.0, "step": 41054 }, { "epoch": 0.8, "learning_rate": 3.6723728561354476e-05, "loss": 0.0037, "step": 41056 }, { "epoch": 0.8, "learning_rate": 3.6723081601097245e-05, "loss": 0.0003, "step": 41058 }, { "epoch": 0.8, "learning_rate": 3.6722434640840014e-05, "loss": 0.0001, "step": 41060 }, { "epoch": 0.8, "learning_rate": 3.672178768058278e-05, "loss": 0.0, "step": 41062 }, { "epoch": 0.8, "learning_rate": 3.672114072032555e-05, "loss": 0.0, "step": 41064 }, { "epoch": 0.8, "learning_rate": 3.672049376006832e-05, "loss": 0.0018, "step": 41066 }, { "epoch": 0.8, "learning_rate": 3.671984679981109e-05, "loss": 0.0005, "step": 41068 }, { "epoch": 0.8, "learning_rate": 3.671919983955386e-05, "loss": 0.0, "step": 41070 }, { "epoch": 0.8, "learning_rate": 3.671855287929662e-05, "loss": 0.0, "step": 41072 }, { "epoch": 0.8, "learning_rate": 3.67179059190394e-05, "loss": 0.0003, "step": 41074 }, { "epoch": 0.8, "learning_rate": 3.671725895878217e-05, "loss": 0.0, "step": 41076 }, { "epoch": 0.8, "learning_rate": 3.671661199852493e-05, "loss": 0.0004, "step": 41078 }, { "epoch": 0.8, "learning_rate": 3.6715965038267705e-05, "loss": 0.0005, "step": 41080 }, { "epoch": 0.8, "learning_rate": 3.671531807801047e-05, "loss": 0.0, "step": 41082 }, { "epoch": 0.8, "learning_rate": 3.6714671117753236e-05, "loss": 0.0037, "step": 41084 }, { "epoch": 0.8, "learning_rate": 3.6714024157496005e-05, "loss": 0.0013, "step": 41086 }, { "epoch": 0.8, "learning_rate": 3.6713377197238774e-05, "loss": 0.004, "step": 41088 }, { "epoch": 0.8, "learning_rate": 3.6712730236981543e-05, "loss": 0.0, "step": 41090 }, { "epoch": 0.8, "learning_rate": 3.671208327672431e-05, "loss": 0.0, "step": 41092 }, { "epoch": 0.8, "learning_rate": 3.671143631646708e-05, "loss": 0.0001, "step": 41094 }, { "epoch": 0.8, "learning_rate": 3.671078935620985e-05, "loss": 0.0072, "step": 41096 }, { "epoch": 0.8, "learning_rate": 3.671014239595262e-05, "loss": 0.0039, "step": 41098 }, { "epoch": 0.8, "learning_rate": 3.670949543569538e-05, "loss": 0.0, "step": 41100 }, { "epoch": 0.8, "learning_rate": 3.670884847543816e-05, "loss": 0.0013, "step": 41102 }, { "epoch": 0.8, "learning_rate": 3.670820151518092e-05, "loss": 0.0059, "step": 41104 }, { "epoch": 0.8, "learning_rate": 3.6707554554923696e-05, "loss": 0.0019, "step": 41106 }, { "epoch": 0.8, "learning_rate": 3.6706907594666465e-05, "loss": 0.0, "step": 41108 }, { "epoch": 0.8, "learning_rate": 3.670626063440923e-05, "loss": 0.0, "step": 41110 }, { "epoch": 0.8, "learning_rate": 3.6705613674152003e-05, "loss": 0.0003, "step": 41112 }, { "epoch": 0.8, "learning_rate": 3.6704966713894766e-05, "loss": 0.0001, "step": 41114 }, { "epoch": 0.8, "learning_rate": 3.6704319753637535e-05, "loss": 0.0046, "step": 41116 }, { "epoch": 0.8, "learning_rate": 3.6703672793380304e-05, "loss": 0.0045, "step": 41118 }, { "epoch": 0.8, "learning_rate": 3.670302583312307e-05, "loss": 0.0001, "step": 41120 }, { "epoch": 0.8, "learning_rate": 3.670237887286584e-05, "loss": 0.0, "step": 41122 }, { "epoch": 0.8, "learning_rate": 3.670173191260861e-05, "loss": 0.0072, "step": 41124 }, { "epoch": 0.8, "learning_rate": 3.670108495235138e-05, "loss": 0.0024, "step": 41126 }, { "epoch": 0.8, "learning_rate": 3.670043799209415e-05, "loss": 0.0, "step": 41128 }, { "epoch": 0.8, "learning_rate": 3.669979103183692e-05, "loss": 0.0001, "step": 41130 }, { "epoch": 0.8, "learning_rate": 3.669914407157968e-05, "loss": 0.0, "step": 41132 }, { "epoch": 0.8, "learning_rate": 3.6698497111322457e-05, "loss": 0.0003, "step": 41134 }, { "epoch": 0.8, "learning_rate": 3.669785015106522e-05, "loss": 0.0001, "step": 41136 }, { "epoch": 0.8, "learning_rate": 3.669720319080799e-05, "loss": 0.0066, "step": 41138 }, { "epoch": 0.8, "learning_rate": 3.669655623055076e-05, "loss": 0.0011, "step": 41140 }, { "epoch": 0.8, "learning_rate": 3.6695909270293526e-05, "loss": 0.0005, "step": 41142 }, { "epoch": 0.8, "learning_rate": 3.6695262310036295e-05, "loss": 0.0, "step": 41144 }, { "epoch": 0.8, "learning_rate": 3.6694615349779064e-05, "loss": 0.004, "step": 41146 }, { "epoch": 0.8, "learning_rate": 3.6693968389521833e-05, "loss": 0.0, "step": 41148 }, { "epoch": 0.8, "learning_rate": 3.66933214292646e-05, "loss": 0.0041, "step": 41150 }, { "epoch": 0.8, "learning_rate": 3.669267446900737e-05, "loss": 0.0006, "step": 41152 }, { "epoch": 0.8, "learning_rate": 3.669202750875014e-05, "loss": 0.0005, "step": 41154 }, { "epoch": 0.8, "learning_rate": 3.669138054849291e-05, "loss": 0.0, "step": 41156 }, { "epoch": 0.8, "learning_rate": 3.669073358823568e-05, "loss": 0.0, "step": 41158 }, { "epoch": 0.8, "learning_rate": 3.669008662797844e-05, "loss": 0.0, "step": 41160 }, { "epoch": 0.8, "learning_rate": 3.668943966772122e-05, "loss": 0.0042, "step": 41162 }, { "epoch": 0.8, "learning_rate": 3.668879270746398e-05, "loss": 0.0, "step": 41164 }, { "epoch": 0.8, "learning_rate": 3.6688145747206755e-05, "loss": 0.0, "step": 41166 }, { "epoch": 0.8, "learning_rate": 3.668749878694952e-05, "loss": 0.0, "step": 41168 }, { "epoch": 0.8, "learning_rate": 3.6686851826692287e-05, "loss": 0.0001, "step": 41170 }, { "epoch": 0.8, "learning_rate": 3.6686204866435056e-05, "loss": 0.0011, "step": 41172 }, { "epoch": 0.8, "learning_rate": 3.6685557906177825e-05, "loss": 0.0, "step": 41174 }, { "epoch": 0.8, "learning_rate": 3.6684910945920594e-05, "loss": 0.0, "step": 41176 }, { "epoch": 0.8, "learning_rate": 3.668426398566336e-05, "loss": 0.0006, "step": 41178 }, { "epoch": 0.8, "learning_rate": 3.668361702540613e-05, "loss": 0.0, "step": 41180 }, { "epoch": 0.8, "learning_rate": 3.6682970065148894e-05, "loss": 0.0002, "step": 41182 }, { "epoch": 0.8, "learning_rate": 3.668232310489167e-05, "loss": 0.0, "step": 41184 }, { "epoch": 0.8, "learning_rate": 3.668167614463444e-05, "loss": 0.0, "step": 41186 }, { "epoch": 0.8, "learning_rate": 3.668102918437721e-05, "loss": 0.0, "step": 41188 }, { "epoch": 0.8, "learning_rate": 3.668038222411998e-05, "loss": 0.0017, "step": 41190 }, { "epoch": 0.8, "learning_rate": 3.667973526386274e-05, "loss": 0.0, "step": 41192 }, { "epoch": 0.8, "learning_rate": 3.6679088303605516e-05, "loss": 0.0, "step": 41194 }, { "epoch": 0.8, "learning_rate": 3.667844134334828e-05, "loss": 0.0, "step": 41196 }, { "epoch": 0.8, "learning_rate": 3.667779438309105e-05, "loss": 0.0036, "step": 41198 }, { "epoch": 0.8, "learning_rate": 3.6677147422833816e-05, "loss": 0.0066, "step": 41200 }, { "epoch": 0.8, "learning_rate": 3.6676500462576585e-05, "loss": 0.0001, "step": 41202 }, { "epoch": 0.8, "learning_rate": 3.6675853502319354e-05, "loss": 0.0, "step": 41204 }, { "epoch": 0.8, "learning_rate": 3.667520654206212e-05, "loss": 0.0, "step": 41206 }, { "epoch": 0.8, "learning_rate": 3.667455958180489e-05, "loss": 0.0198, "step": 41208 }, { "epoch": 0.8, "learning_rate": 3.667391262154766e-05, "loss": 0.0, "step": 41210 }, { "epoch": 0.8, "learning_rate": 3.667326566129043e-05, "loss": 0.0001, "step": 41212 }, { "epoch": 0.8, "learning_rate": 3.667261870103319e-05, "loss": 0.0, "step": 41214 }, { "epoch": 0.8, "learning_rate": 3.667197174077597e-05, "loss": 0.0004, "step": 41216 }, { "epoch": 0.8, "learning_rate": 3.667132478051873e-05, "loss": 0.0016, "step": 41218 }, { "epoch": 0.8, "learning_rate": 3.66706778202615e-05, "loss": 0.0001, "step": 41220 }, { "epoch": 0.8, "learning_rate": 3.6670030860004276e-05, "loss": 0.0004, "step": 41222 }, { "epoch": 0.8, "learning_rate": 3.666938389974704e-05, "loss": 0.0, "step": 41224 }, { "epoch": 0.8, "learning_rate": 3.666873693948981e-05, "loss": 0.0015, "step": 41226 }, { "epoch": 0.8, "learning_rate": 3.6668089979232576e-05, "loss": 0.0006, "step": 41228 }, { "epoch": 0.8, "learning_rate": 3.6667443018975346e-05, "loss": 0.0012, "step": 41230 }, { "epoch": 0.8, "learning_rate": 3.6666796058718115e-05, "loss": 0.0, "step": 41232 }, { "epoch": 0.8, "learning_rate": 3.6666149098460884e-05, "loss": 0.0004, "step": 41234 }, { "epoch": 0.8, "learning_rate": 3.666550213820365e-05, "loss": 0.005, "step": 41236 }, { "epoch": 0.8, "learning_rate": 3.666485517794642e-05, "loss": 0.0008, "step": 41238 }, { "epoch": 0.8, "learning_rate": 3.666420821768919e-05, "loss": 0.0, "step": 41240 }, { "epoch": 0.8, "learning_rate": 3.666356125743195e-05, "loss": 0.0, "step": 41242 }, { "epoch": 0.8, "learning_rate": 3.666291429717473e-05, "loss": 0.0001, "step": 41244 }, { "epoch": 0.8, "learning_rate": 3.666226733691749e-05, "loss": 0.0, "step": 41246 }, { "epoch": 0.8, "learning_rate": 3.666162037666027e-05, "loss": 0.0004, "step": 41248 }, { "epoch": 0.8, "learning_rate": 3.666097341640303e-05, "loss": 0.0, "step": 41250 }, { "epoch": 0.8, "learning_rate": 3.66603264561458e-05, "loss": 0.0, "step": 41252 }, { "epoch": 0.8, "learning_rate": 3.6659679495888575e-05, "loss": 0.0, "step": 41254 }, { "epoch": 0.8, "learning_rate": 3.665903253563134e-05, "loss": 0.0004, "step": 41256 }, { "epoch": 0.8, "learning_rate": 3.6658385575374106e-05, "loss": 0.0035, "step": 41258 }, { "epoch": 0.8, "learning_rate": 3.6657738615116875e-05, "loss": 0.0, "step": 41260 }, { "epoch": 0.8, "learning_rate": 3.6657091654859644e-05, "loss": 0.002, "step": 41262 }, { "epoch": 0.8, "learning_rate": 3.665644469460241e-05, "loss": 0.0, "step": 41264 }, { "epoch": 0.8, "learning_rate": 3.665579773434518e-05, "loss": 0.0, "step": 41266 }, { "epoch": 0.8, "learning_rate": 3.665515077408795e-05, "loss": 0.0003, "step": 41268 }, { "epoch": 0.8, "learning_rate": 3.665450381383072e-05, "loss": 0.0, "step": 41270 }, { "epoch": 0.8, "learning_rate": 3.665385685357349e-05, "loss": 0.0, "step": 41272 }, { "epoch": 0.8, "learning_rate": 3.665320989331625e-05, "loss": 0.0, "step": 41274 }, { "epoch": 0.8, "learning_rate": 3.665256293305903e-05, "loss": 0.0002, "step": 41276 }, { "epoch": 0.8, "learning_rate": 3.665191597280179e-05, "loss": 0.0001, "step": 41278 }, { "epoch": 0.8, "learning_rate": 3.665126901254456e-05, "loss": 0.0039, "step": 41280 }, { "epoch": 0.8, "learning_rate": 3.665062205228733e-05, "loss": 0.0004, "step": 41282 }, { "epoch": 0.8, "learning_rate": 3.66499750920301e-05, "loss": 0.0011, "step": 41284 }, { "epoch": 0.8, "learning_rate": 3.6649328131772866e-05, "loss": 0.0057, "step": 41286 }, { "epoch": 0.8, "learning_rate": 3.6648681171515635e-05, "loss": 0.0, "step": 41288 }, { "epoch": 0.8, "learning_rate": 3.6648034211258405e-05, "loss": 0.006, "step": 41290 }, { "epoch": 0.8, "learning_rate": 3.6647387251001174e-05, "loss": 0.0002, "step": 41292 }, { "epoch": 0.8, "learning_rate": 3.664674029074394e-05, "loss": 0.0, "step": 41294 }, { "epoch": 0.8, "learning_rate": 3.6646093330486705e-05, "loss": 0.0, "step": 41296 }, { "epoch": 0.8, "learning_rate": 3.664544637022948e-05, "loss": 0.0003, "step": 41298 }, { "epoch": 0.8, "learning_rate": 3.664479940997225e-05, "loss": 0.004, "step": 41300 }, { "epoch": 0.8, "learning_rate": 3.664415244971501e-05, "loss": 0.0, "step": 41302 }, { "epoch": 0.8, "learning_rate": 3.664350548945779e-05, "loss": 0.0, "step": 41304 }, { "epoch": 0.8, "learning_rate": 3.664285852920055e-05, "loss": 0.0002, "step": 41306 }, { "epoch": 0.8, "learning_rate": 3.6642211568943326e-05, "loss": 0.0012, "step": 41308 }, { "epoch": 0.8, "learning_rate": 3.664156460868609e-05, "loss": 0.0, "step": 41310 }, { "epoch": 0.8, "learning_rate": 3.664091764842886e-05, "loss": 0.0, "step": 41312 }, { "epoch": 0.8, "learning_rate": 3.664027068817163e-05, "loss": 0.0016, "step": 41314 }, { "epoch": 0.8, "learning_rate": 3.6639623727914396e-05, "loss": 0.0001, "step": 41316 }, { "epoch": 0.8, "learning_rate": 3.6638976767657165e-05, "loss": 0.0, "step": 41318 }, { "epoch": 0.8, "learning_rate": 3.6638329807399934e-05, "loss": 0.0001, "step": 41320 }, { "epoch": 0.8, "learning_rate": 3.66376828471427e-05, "loss": 0.0093, "step": 41322 }, { "epoch": 0.8, "learning_rate": 3.6637035886885465e-05, "loss": 0.0095, "step": 41324 }, { "epoch": 0.8, "learning_rate": 3.663638892662824e-05, "loss": 0.0, "step": 41326 }, { "epoch": 0.8, "learning_rate": 3.6635741966371004e-05, "loss": 0.0016, "step": 41328 }, { "epoch": 0.8, "learning_rate": 3.663509500611378e-05, "loss": 0.0175, "step": 41330 }, { "epoch": 0.8, "learning_rate": 3.663444804585655e-05, "loss": 0.0107, "step": 41332 }, { "epoch": 0.8, "learning_rate": 3.663380108559931e-05, "loss": 0.0, "step": 41334 }, { "epoch": 0.8, "learning_rate": 3.663315412534209e-05, "loss": 0.0, "step": 41336 }, { "epoch": 0.8, "learning_rate": 3.663250716508485e-05, "loss": 0.0002, "step": 41338 }, { "epoch": 0.8, "learning_rate": 3.663186020482762e-05, "loss": 0.0, "step": 41340 }, { "epoch": 0.8, "learning_rate": 3.663121324457039e-05, "loss": 0.0, "step": 41342 }, { "epoch": 0.8, "learning_rate": 3.6630566284313156e-05, "loss": 0.0, "step": 41344 }, { "epoch": 0.8, "learning_rate": 3.6629919324055925e-05, "loss": 0.0, "step": 41346 }, { "epoch": 0.8, "learning_rate": 3.6629272363798694e-05, "loss": 0.0018, "step": 41348 }, { "epoch": 0.8, "learning_rate": 3.6628625403541463e-05, "loss": 0.0, "step": 41350 }, { "epoch": 0.8, "learning_rate": 3.662797844328423e-05, "loss": 0.0012, "step": 41352 }, { "epoch": 0.8, "learning_rate": 3.6627331483027e-05, "loss": 0.0, "step": 41354 }, { "epoch": 0.8, "learning_rate": 3.6626684522769764e-05, "loss": 0.0117, "step": 41356 }, { "epoch": 0.8, "learning_rate": 3.662603756251254e-05, "loss": 0.0003, "step": 41358 }, { "epoch": 0.8, "learning_rate": 3.66253906022553e-05, "loss": 0.0, "step": 41360 }, { "epoch": 0.8, "learning_rate": 3.662474364199807e-05, "loss": 0.0055, "step": 41362 }, { "epoch": 0.8, "learning_rate": 3.662409668174085e-05, "loss": 0.0, "step": 41364 }, { "epoch": 0.8, "learning_rate": 3.662344972148361e-05, "loss": 0.0483, "step": 41366 }, { "epoch": 0.8, "learning_rate": 3.662280276122638e-05, "loss": 0.0006, "step": 41368 }, { "epoch": 0.8, "learning_rate": 3.662215580096915e-05, "loss": 0.0014, "step": 41370 }, { "epoch": 0.8, "learning_rate": 3.662150884071192e-05, "loss": 0.0, "step": 41372 }, { "epoch": 0.8, "learning_rate": 3.6620861880454686e-05, "loss": 0.0053, "step": 41374 }, { "epoch": 0.8, "learning_rate": 3.6620214920197455e-05, "loss": 0.0, "step": 41376 }, { "epoch": 0.8, "learning_rate": 3.6619567959940224e-05, "loss": 0.0, "step": 41378 }, { "epoch": 0.8, "learning_rate": 3.661892099968299e-05, "loss": 0.0001, "step": 41380 }, { "epoch": 0.8, "learning_rate": 3.661827403942576e-05, "loss": 0.0, "step": 41382 }, { "epoch": 0.8, "learning_rate": 3.6617627079168524e-05, "loss": 0.0144, "step": 41384 }, { "epoch": 0.8, "learning_rate": 3.66169801189113e-05, "loss": 0.0, "step": 41386 }, { "epoch": 0.8, "learning_rate": 3.661633315865406e-05, "loss": 0.0001, "step": 41388 }, { "epoch": 0.8, "learning_rate": 3.661568619839684e-05, "loss": 0.0, "step": 41390 }, { "epoch": 0.8, "learning_rate": 3.66150392381396e-05, "loss": 0.0002, "step": 41392 }, { "epoch": 0.8, "learning_rate": 3.661439227788237e-05, "loss": 0.0067, "step": 41394 }, { "epoch": 0.8, "learning_rate": 3.661374531762514e-05, "loss": 0.0001, "step": 41396 }, { "epoch": 0.8, "learning_rate": 3.661309835736791e-05, "loss": 0.0, "step": 41398 }, { "epoch": 0.8, "learning_rate": 3.661245139711068e-05, "loss": 0.0148, "step": 41400 }, { "epoch": 0.8, "learning_rate": 3.6611804436853446e-05, "loss": 0.0002, "step": 41402 }, { "epoch": 0.8, "learning_rate": 3.6611157476596215e-05, "loss": 0.0001, "step": 41404 }, { "epoch": 0.8, "learning_rate": 3.661051051633898e-05, "loss": 0.0, "step": 41406 }, { "epoch": 0.8, "learning_rate": 3.660986355608175e-05, "loss": 0.0, "step": 41408 }, { "epoch": 0.8, "learning_rate": 3.660921659582452e-05, "loss": 0.0047, "step": 41410 }, { "epoch": 0.8, "learning_rate": 3.660856963556729e-05, "loss": 0.0, "step": 41412 }, { "epoch": 0.8, "learning_rate": 3.660792267531006e-05, "loss": 0.0028, "step": 41414 }, { "epoch": 0.8, "learning_rate": 3.660727571505282e-05, "loss": 0.0, "step": 41416 }, { "epoch": 0.8, "learning_rate": 3.66066287547956e-05, "loss": 0.0, "step": 41418 }, { "epoch": 0.8, "learning_rate": 3.660598179453836e-05, "loss": 0.0, "step": 41420 }, { "epoch": 0.8, "learning_rate": 3.660533483428113e-05, "loss": 0.0, "step": 41422 }, { "epoch": 0.8, "learning_rate": 3.66046878740239e-05, "loss": 0.0, "step": 41424 }, { "epoch": 0.8, "learning_rate": 3.660404091376667e-05, "loss": 0.0009, "step": 41426 }, { "epoch": 0.8, "learning_rate": 3.660339395350944e-05, "loss": 0.0, "step": 41428 }, { "epoch": 0.8, "learning_rate": 3.6602746993252207e-05, "loss": 0.0085, "step": 41430 }, { "epoch": 0.8, "learning_rate": 3.6602100032994976e-05, "loss": 0.0004, "step": 41432 }, { "epoch": 0.8, "learning_rate": 3.6601453072737745e-05, "loss": 0.0, "step": 41434 }, { "epoch": 0.8, "learning_rate": 3.6600806112480514e-05, "loss": 0.0001, "step": 41436 }, { "epoch": 0.8, "learning_rate": 3.6600159152223276e-05, "loss": 0.0038, "step": 41438 }, { "epoch": 0.8, "learning_rate": 3.659951219196605e-05, "loss": 0.0064, "step": 41440 }, { "epoch": 0.8, "learning_rate": 3.659886523170882e-05, "loss": 0.0, "step": 41442 }, { "epoch": 0.8, "learning_rate": 3.659821827145158e-05, "loss": 0.0, "step": 41444 }, { "epoch": 0.8, "learning_rate": 3.659757131119436e-05, "loss": 0.0, "step": 41446 }, { "epoch": 0.8, "learning_rate": 3.659692435093712e-05, "loss": 0.0067, "step": 41448 }, { "epoch": 0.8, "learning_rate": 3.65962773906799e-05, "loss": 0.0118, "step": 41450 }, { "epoch": 0.8, "learning_rate": 3.659563043042266e-05, "loss": 0.0001, "step": 41452 }, { "epoch": 0.8, "learning_rate": 3.659498347016543e-05, "loss": 0.0, "step": 41454 }, { "epoch": 0.8, "learning_rate": 3.65943365099082e-05, "loss": 0.0, "step": 41456 }, { "epoch": 0.8, "learning_rate": 3.659368954965097e-05, "loss": 0.0084, "step": 41458 }, { "epoch": 0.8, "learning_rate": 3.6593042589393736e-05, "loss": 0.0, "step": 41460 }, { "epoch": 0.8, "learning_rate": 3.6592395629136505e-05, "loss": 0.0001, "step": 41462 }, { "epoch": 0.8, "learning_rate": 3.6591748668879274e-05, "loss": 0.0, "step": 41464 }, { "epoch": 0.8, "learning_rate": 3.6591101708622036e-05, "loss": 0.0, "step": 41466 }, { "epoch": 0.8, "learning_rate": 3.659045474836481e-05, "loss": 0.0, "step": 41468 }, { "epoch": 0.8, "learning_rate": 3.6589807788107575e-05, "loss": 0.0002, "step": 41470 }, { "epoch": 0.8, "learning_rate": 3.658916082785035e-05, "loss": 0.0058, "step": 41472 }, { "epoch": 0.8, "learning_rate": 3.658851386759311e-05, "loss": 0.0, "step": 41474 }, { "epoch": 0.8, "learning_rate": 3.658786690733588e-05, "loss": 0.0, "step": 41476 }, { "epoch": 0.81, "learning_rate": 3.658721994707866e-05, "loss": 0.0, "step": 41478 }, { "epoch": 0.81, "learning_rate": 3.658657298682142e-05, "loss": 0.0, "step": 41480 }, { "epoch": 0.81, "learning_rate": 3.658592602656419e-05, "loss": 0.005, "step": 41482 }, { "epoch": 0.81, "learning_rate": 3.658527906630696e-05, "loss": 0.0003, "step": 41484 }, { "epoch": 0.81, "learning_rate": 3.658463210604973e-05, "loss": 0.0001, "step": 41486 }, { "epoch": 0.81, "learning_rate": 3.6583985145792496e-05, "loss": 0.0, "step": 41488 }, { "epoch": 0.81, "learning_rate": 3.6583338185535266e-05, "loss": 0.0023, "step": 41490 }, { "epoch": 0.81, "learning_rate": 3.6582691225278035e-05, "loss": 0.0003, "step": 41492 }, { "epoch": 0.81, "learning_rate": 3.6582044265020804e-05, "loss": 0.0326, "step": 41494 }, { "epoch": 0.81, "learning_rate": 3.658139730476357e-05, "loss": 0.0003, "step": 41496 }, { "epoch": 0.81, "learning_rate": 3.6580750344506335e-05, "loss": 0.0, "step": 41498 }, { "epoch": 0.81, "learning_rate": 3.658010338424911e-05, "loss": 0.0001, "step": 41500 }, { "epoch": 0.81, "learning_rate": 3.657945642399187e-05, "loss": 0.0093, "step": 41502 }, { "epoch": 0.81, "learning_rate": 3.657880946373464e-05, "loss": 0.0051, "step": 41504 }, { "epoch": 0.81, "learning_rate": 3.657816250347741e-05, "loss": 0.0, "step": 41506 }, { "epoch": 0.81, "learning_rate": 3.657751554322018e-05, "loss": 0.0127, "step": 41508 }, { "epoch": 0.81, "learning_rate": 3.657686858296295e-05, "loss": 0.0089, "step": 41510 }, { "epoch": 0.81, "learning_rate": 3.657622162270572e-05, "loss": 0.0025, "step": 41512 }, { "epoch": 0.81, "learning_rate": 3.657557466244849e-05, "loss": 0.0037, "step": 41514 }, { "epoch": 0.81, "learning_rate": 3.657492770219126e-05, "loss": 0.0021, "step": 41516 }, { "epoch": 0.81, "learning_rate": 3.6574280741934026e-05, "loss": 0.0, "step": 41518 }, { "epoch": 0.81, "learning_rate": 3.657363378167679e-05, "loss": 0.0, "step": 41520 }, { "epoch": 0.81, "learning_rate": 3.6572986821419564e-05, "loss": 0.0, "step": 41522 }, { "epoch": 0.81, "learning_rate": 3.657233986116233e-05, "loss": 0.0, "step": 41524 }, { "epoch": 0.81, "learning_rate": 3.6571692900905095e-05, "loss": 0.0, "step": 41526 }, { "epoch": 0.81, "learning_rate": 3.657104594064787e-05, "loss": 0.0004, "step": 41528 }, { "epoch": 0.81, "learning_rate": 3.6570398980390634e-05, "loss": 0.0001, "step": 41530 }, { "epoch": 0.81, "learning_rate": 3.656975202013341e-05, "loss": 0.0003, "step": 41532 }, { "epoch": 0.81, "learning_rate": 3.656910505987617e-05, "loss": 0.0075, "step": 41534 }, { "epoch": 0.81, "learning_rate": 3.656845809961894e-05, "loss": 0.0001, "step": 41536 }, { "epoch": 0.81, "learning_rate": 3.656781113936171e-05, "loss": 0.0009, "step": 41538 }, { "epoch": 0.81, "learning_rate": 3.656716417910448e-05, "loss": 0.0, "step": 41540 }, { "epoch": 0.81, "learning_rate": 3.656651721884725e-05, "loss": 0.0001, "step": 41542 }, { "epoch": 0.81, "learning_rate": 3.656587025859002e-05, "loss": 0.0002, "step": 41544 }, { "epoch": 0.81, "learning_rate": 3.6565223298332786e-05, "loss": 0.0, "step": 41546 }, { "epoch": 0.81, "learning_rate": 3.656457633807555e-05, "loss": 0.0, "step": 41548 }, { "epoch": 0.81, "learning_rate": 3.6563929377818324e-05, "loss": 0.0011, "step": 41550 }, { "epoch": 0.81, "learning_rate": 3.656328241756109e-05, "loss": 0.0, "step": 41552 }, { "epoch": 0.81, "learning_rate": 3.656263545730386e-05, "loss": 0.0034, "step": 41554 }, { "epoch": 0.81, "learning_rate": 3.656198849704663e-05, "loss": 0.0002, "step": 41556 }, { "epoch": 0.81, "learning_rate": 3.6561341536789394e-05, "loss": 0.0, "step": 41558 }, { "epoch": 0.81, "learning_rate": 3.656069457653217e-05, "loss": 0.0001, "step": 41560 }, { "epoch": 0.81, "learning_rate": 3.656004761627493e-05, "loss": 0.0, "step": 41562 }, { "epoch": 0.81, "learning_rate": 3.65594006560177e-05, "loss": 0.0003, "step": 41564 }, { "epoch": 0.81, "learning_rate": 3.655875369576047e-05, "loss": 0.0002, "step": 41566 }, { "epoch": 0.81, "learning_rate": 3.655810673550324e-05, "loss": 0.0211, "step": 41568 }, { "epoch": 0.81, "learning_rate": 3.655745977524601e-05, "loss": 0.0002, "step": 41570 }, { "epoch": 0.81, "learning_rate": 3.655681281498878e-05, "loss": 0.0002, "step": 41572 }, { "epoch": 0.81, "learning_rate": 3.655616585473155e-05, "loss": 0.0, "step": 41574 }, { "epoch": 0.81, "learning_rate": 3.6555518894474316e-05, "loss": 0.0, "step": 41576 }, { "epoch": 0.81, "learning_rate": 3.6554871934217085e-05, "loss": 0.0, "step": 41578 }, { "epoch": 0.81, "learning_rate": 3.655422497395985e-05, "loss": 0.0, "step": 41580 }, { "epoch": 0.81, "learning_rate": 3.655357801370262e-05, "loss": 0.0, "step": 41582 }, { "epoch": 0.81, "learning_rate": 3.6552931053445385e-05, "loss": 0.0001, "step": 41584 }, { "epoch": 0.81, "learning_rate": 3.6552284093188154e-05, "loss": 0.0122, "step": 41586 }, { "epoch": 0.81, "learning_rate": 3.655163713293093e-05, "loss": 0.0001, "step": 41588 }, { "epoch": 0.81, "learning_rate": 3.655099017267369e-05, "loss": 0.0003, "step": 41590 }, { "epoch": 0.81, "learning_rate": 3.655034321241647e-05, "loss": 0.0, "step": 41592 }, { "epoch": 0.81, "learning_rate": 3.654969625215923e-05, "loss": 0.0, "step": 41594 }, { "epoch": 0.81, "learning_rate": 3.6549049291902e-05, "loss": 0.0, "step": 41596 }, { "epoch": 0.81, "learning_rate": 3.654840233164477e-05, "loss": 0.0001, "step": 41598 }, { "epoch": 0.81, "learning_rate": 3.654775537138754e-05, "loss": 0.0016, "step": 41600 }, { "epoch": 0.81, "learning_rate": 3.654710841113031e-05, "loss": 0.0, "step": 41602 }, { "epoch": 0.81, "learning_rate": 3.6546461450873076e-05, "loss": 0.0, "step": 41604 }, { "epoch": 0.81, "learning_rate": 3.6545814490615845e-05, "loss": 0.0, "step": 41606 }, { "epoch": 0.81, "learning_rate": 3.654516753035861e-05, "loss": 0.0002, "step": 41608 }, { "epoch": 0.81, "learning_rate": 3.6544520570101383e-05, "loss": 0.0, "step": 41610 }, { "epoch": 0.81, "learning_rate": 3.6543873609844146e-05, "loss": 0.0014, "step": 41612 }, { "epoch": 0.81, "learning_rate": 3.654322664958692e-05, "loss": 0.0, "step": 41614 }, { "epoch": 0.81, "learning_rate": 3.6542579689329684e-05, "loss": 0.0001, "step": 41616 }, { "epoch": 0.81, "learning_rate": 3.654193272907245e-05, "loss": 0.0, "step": 41618 }, { "epoch": 0.81, "learning_rate": 3.654128576881522e-05, "loss": 0.013, "step": 41620 }, { "epoch": 0.81, "learning_rate": 3.654063880855799e-05, "loss": 0.0024, "step": 41622 }, { "epoch": 0.81, "learning_rate": 3.653999184830076e-05, "loss": 0.0016, "step": 41624 }, { "epoch": 0.81, "learning_rate": 3.653934488804353e-05, "loss": 0.0, "step": 41626 }, { "epoch": 0.81, "learning_rate": 3.65386979277863e-05, "loss": 0.0, "step": 41628 }, { "epoch": 0.81, "learning_rate": 3.653805096752906e-05, "loss": 0.0062, "step": 41630 }, { "epoch": 0.81, "learning_rate": 3.653740400727184e-05, "loss": 0.0, "step": 41632 }, { "epoch": 0.81, "learning_rate": 3.6536757047014606e-05, "loss": 0.0, "step": 41634 }, { "epoch": 0.81, "learning_rate": 3.6536110086757375e-05, "loss": 0.0, "step": 41636 }, { "epoch": 0.81, "learning_rate": 3.6535463126500144e-05, "loss": 0.0, "step": 41638 }, { "epoch": 0.81, "learning_rate": 3.6534816166242906e-05, "loss": 0.0, "step": 41640 }, { "epoch": 0.81, "learning_rate": 3.653416920598568e-05, "loss": 0.0014, "step": 41642 }, { "epoch": 0.81, "learning_rate": 3.6533522245728444e-05, "loss": 0.0002, "step": 41644 }, { "epoch": 0.81, "learning_rate": 3.6532875285471213e-05, "loss": 0.0005, "step": 41646 }, { "epoch": 0.81, "learning_rate": 3.653222832521398e-05, "loss": 0.0003, "step": 41648 }, { "epoch": 0.81, "learning_rate": 3.653158136495675e-05, "loss": 0.0001, "step": 41650 }, { "epoch": 0.81, "learning_rate": 3.653093440469952e-05, "loss": 0.0, "step": 41652 }, { "epoch": 0.81, "learning_rate": 3.653028744444229e-05, "loss": 0.0, "step": 41654 }, { "epoch": 0.81, "learning_rate": 3.652964048418506e-05, "loss": 0.0, "step": 41656 }, { "epoch": 0.81, "learning_rate": 3.652899352392783e-05, "loss": 0.0, "step": 41658 }, { "epoch": 0.81, "learning_rate": 3.65283465636706e-05, "loss": 0.0002, "step": 41660 }, { "epoch": 0.81, "learning_rate": 3.652769960341336e-05, "loss": 0.0075, "step": 41662 }, { "epoch": 0.81, "learning_rate": 3.6527052643156135e-05, "loss": 0.0, "step": 41664 }, { "epoch": 0.81, "learning_rate": 3.6526405682898904e-05, "loss": 0.0051, "step": 41666 }, { "epoch": 0.81, "learning_rate": 3.6525758722641667e-05, "loss": 0.0, "step": 41668 }, { "epoch": 0.81, "learning_rate": 3.652511176238444e-05, "loss": 0.0, "step": 41670 }, { "epoch": 0.81, "learning_rate": 3.6524464802127205e-05, "loss": 0.0, "step": 41672 }, { "epoch": 0.81, "learning_rate": 3.652381784186998e-05, "loss": 0.0, "step": 41674 }, { "epoch": 0.81, "learning_rate": 3.652317088161274e-05, "loss": 0.0001, "step": 41676 }, { "epoch": 0.81, "learning_rate": 3.652252392135551e-05, "loss": 0.0001, "step": 41678 }, { "epoch": 0.81, "learning_rate": 3.652187696109828e-05, "loss": 0.0001, "step": 41680 }, { "epoch": 0.81, "learning_rate": 3.652123000084105e-05, "loss": 0.0003, "step": 41682 }, { "epoch": 0.81, "learning_rate": 3.652058304058382e-05, "loss": 0.0, "step": 41684 }, { "epoch": 0.81, "learning_rate": 3.651993608032659e-05, "loss": 0.0, "step": 41686 }, { "epoch": 0.81, "learning_rate": 3.651928912006936e-05, "loss": 0.0001, "step": 41688 }, { "epoch": 0.81, "learning_rate": 3.651864215981212e-05, "loss": 0.0, "step": 41690 }, { "epoch": 0.81, "learning_rate": 3.6517995199554896e-05, "loss": 0.0015, "step": 41692 }, { "epoch": 0.81, "learning_rate": 3.651734823929766e-05, "loss": 0.0, "step": 41694 }, { "epoch": 0.81, "learning_rate": 3.6516701279040434e-05, "loss": 0.0005, "step": 41696 }, { "epoch": 0.81, "learning_rate": 3.6516054318783196e-05, "loss": 0.0, "step": 41698 }, { "epoch": 0.81, "learning_rate": 3.6515407358525965e-05, "loss": 0.0, "step": 41700 }, { "epoch": 0.81, "learning_rate": 3.651476039826874e-05, "loss": 0.0, "step": 41702 }, { "epoch": 0.81, "learning_rate": 3.65141134380115e-05, "loss": 0.0002, "step": 41704 }, { "epoch": 0.81, "learning_rate": 3.651346647775427e-05, "loss": 0.0059, "step": 41706 }, { "epoch": 0.81, "learning_rate": 3.651281951749704e-05, "loss": 0.0, "step": 41708 }, { "epoch": 0.81, "learning_rate": 3.651217255723981e-05, "loss": 0.0004, "step": 41710 }, { "epoch": 0.81, "learning_rate": 3.651152559698258e-05, "loss": 0.0, "step": 41712 }, { "epoch": 0.81, "learning_rate": 3.651087863672535e-05, "loss": 0.0042, "step": 41714 }, { "epoch": 0.81, "learning_rate": 3.651023167646812e-05, "loss": 0.0026, "step": 41716 }, { "epoch": 0.81, "learning_rate": 3.650958471621089e-05, "loss": 0.0002, "step": 41718 }, { "epoch": 0.81, "learning_rate": 3.6508937755953656e-05, "loss": 0.0014, "step": 41720 }, { "epoch": 0.81, "learning_rate": 3.650829079569642e-05, "loss": 0.0002, "step": 41722 }, { "epoch": 0.81, "learning_rate": 3.6507643835439194e-05, "loss": 0.0, "step": 41724 }, { "epoch": 0.81, "learning_rate": 3.6506996875181956e-05, "loss": 0.0027, "step": 41726 }, { "epoch": 0.81, "learning_rate": 3.6506349914924726e-05, "loss": 0.0, "step": 41728 }, { "epoch": 0.81, "learning_rate": 3.6505702954667495e-05, "loss": 0.0, "step": 41730 }, { "epoch": 0.81, "learning_rate": 3.6505055994410264e-05, "loss": 0.0008, "step": 41732 }, { "epoch": 0.81, "learning_rate": 3.650440903415304e-05, "loss": 0.0002, "step": 41734 }, { "epoch": 0.81, "learning_rate": 3.65037620738958e-05, "loss": 0.0003, "step": 41736 }, { "epoch": 0.81, "learning_rate": 3.650311511363857e-05, "loss": 0.0004, "step": 41738 }, { "epoch": 0.81, "learning_rate": 3.650246815338134e-05, "loss": 0.0001, "step": 41740 }, { "epoch": 0.81, "learning_rate": 3.650182119312411e-05, "loss": 0.0, "step": 41742 }, { "epoch": 0.81, "learning_rate": 3.650117423286688e-05, "loss": 0.0001, "step": 41744 }, { "epoch": 0.81, "learning_rate": 3.650052727260965e-05, "loss": 0.0, "step": 41746 }, { "epoch": 0.81, "learning_rate": 3.6499880312352416e-05, "loss": 0.0, "step": 41748 }, { "epoch": 0.81, "learning_rate": 3.649923335209518e-05, "loss": 0.0, "step": 41750 }, { "epoch": 0.81, "learning_rate": 3.6498586391837955e-05, "loss": 0.0001, "step": 41752 }, { "epoch": 0.81, "learning_rate": 3.649793943158072e-05, "loss": 0.0001, "step": 41754 }, { "epoch": 0.81, "learning_rate": 3.649729247132349e-05, "loss": 0.0, "step": 41756 }, { "epoch": 0.81, "learning_rate": 3.6496645511066255e-05, "loss": 0.0012, "step": 41758 }, { "epoch": 0.81, "learning_rate": 3.6495998550809024e-05, "loss": 0.0021, "step": 41760 }, { "epoch": 0.81, "learning_rate": 3.649535159055179e-05, "loss": 0.0001, "step": 41762 }, { "epoch": 0.81, "learning_rate": 3.649470463029456e-05, "loss": 0.0001, "step": 41764 }, { "epoch": 0.81, "learning_rate": 3.649405767003733e-05, "loss": 0.0018, "step": 41766 }, { "epoch": 0.81, "learning_rate": 3.64934107097801e-05, "loss": 0.0, "step": 41768 }, { "epoch": 0.81, "learning_rate": 3.649276374952287e-05, "loss": 0.0001, "step": 41770 }, { "epoch": 0.81, "learning_rate": 3.649211678926563e-05, "loss": 0.0, "step": 41772 }, { "epoch": 0.81, "learning_rate": 3.649146982900841e-05, "loss": 0.0031, "step": 41774 }, { "epoch": 0.81, "learning_rate": 3.649082286875117e-05, "loss": 0.0001, "step": 41776 }, { "epoch": 0.81, "learning_rate": 3.6490175908493946e-05, "loss": 0.0001, "step": 41778 }, { "epoch": 0.81, "learning_rate": 3.6489528948236715e-05, "loss": 0.001, "step": 41780 }, { "epoch": 0.81, "learning_rate": 3.648888198797948e-05, "loss": 0.0065, "step": 41782 }, { "epoch": 0.81, "learning_rate": 3.648823502772225e-05, "loss": 0.0, "step": 41784 }, { "epoch": 0.81, "learning_rate": 3.6487588067465015e-05, "loss": 0.0001, "step": 41786 }, { "epoch": 0.81, "learning_rate": 3.6486941107207785e-05, "loss": 0.0, "step": 41788 }, { "epoch": 0.81, "learning_rate": 3.6486294146950554e-05, "loss": 0.0, "step": 41790 }, { "epoch": 0.81, "learning_rate": 3.648564718669332e-05, "loss": 0.0068, "step": 41792 }, { "epoch": 0.81, "learning_rate": 3.648500022643609e-05, "loss": 0.0001, "step": 41794 }, { "epoch": 0.81, "learning_rate": 3.648435326617886e-05, "loss": 0.0, "step": 41796 }, { "epoch": 0.81, "learning_rate": 3.648370630592163e-05, "loss": 0.0, "step": 41798 }, { "epoch": 0.81, "learning_rate": 3.64830593456644e-05, "loss": 0.0057, "step": 41800 }, { "epoch": 0.81, "learning_rate": 3.648241238540717e-05, "loss": 0.0, "step": 41802 }, { "epoch": 0.81, "learning_rate": 3.648176542514993e-05, "loss": 0.0173, "step": 41804 }, { "epoch": 0.81, "learning_rate": 3.6481118464892706e-05, "loss": 0.0001, "step": 41806 }, { "epoch": 0.81, "learning_rate": 3.648047150463547e-05, "loss": 0.0111, "step": 41808 }, { "epoch": 0.81, "learning_rate": 3.647982454437824e-05, "loss": 0.0001, "step": 41810 }, { "epoch": 0.81, "learning_rate": 3.6479177584121014e-05, "loss": 0.0, "step": 41812 }, { "epoch": 0.81, "learning_rate": 3.6478530623863776e-05, "loss": 0.0, "step": 41814 }, { "epoch": 0.81, "learning_rate": 3.647788366360655e-05, "loss": 0.0046, "step": 41816 }, { "epoch": 0.81, "learning_rate": 3.6477236703349314e-05, "loss": 0.0, "step": 41818 }, { "epoch": 0.81, "learning_rate": 3.647658974309208e-05, "loss": 0.0, "step": 41820 }, { "epoch": 0.81, "learning_rate": 3.647594278283485e-05, "loss": 0.0024, "step": 41822 }, { "epoch": 0.81, "learning_rate": 3.647529582257762e-05, "loss": 0.0, "step": 41824 }, { "epoch": 0.81, "learning_rate": 3.647464886232039e-05, "loss": 0.0037, "step": 41826 }, { "epoch": 0.81, "learning_rate": 3.647400190206316e-05, "loss": 0.0035, "step": 41828 }, { "epoch": 0.81, "learning_rate": 3.647335494180593e-05, "loss": 0.0001, "step": 41830 }, { "epoch": 0.81, "learning_rate": 3.647270798154869e-05, "loss": 0.0, "step": 41832 }, { "epoch": 0.81, "learning_rate": 3.647206102129147e-05, "loss": 0.0, "step": 41834 }, { "epoch": 0.81, "learning_rate": 3.647141406103423e-05, "loss": 0.0044, "step": 41836 }, { "epoch": 0.81, "learning_rate": 3.6470767100777005e-05, "loss": 0.0, "step": 41838 }, { "epoch": 0.81, "learning_rate": 3.647012014051977e-05, "loss": 0.0001, "step": 41840 }, { "epoch": 0.81, "learning_rate": 3.6469473180262536e-05, "loss": 0.0001, "step": 41842 }, { "epoch": 0.81, "learning_rate": 3.646882622000531e-05, "loss": 0.0002, "step": 41844 }, { "epoch": 0.81, "learning_rate": 3.6468179259748074e-05, "loss": 0.0007, "step": 41846 }, { "epoch": 0.81, "learning_rate": 3.6467532299490844e-05, "loss": 0.0, "step": 41848 }, { "epoch": 0.81, "learning_rate": 3.646688533923361e-05, "loss": 0.0104, "step": 41850 }, { "epoch": 0.81, "learning_rate": 3.646623837897638e-05, "loss": 0.0, "step": 41852 }, { "epoch": 0.81, "learning_rate": 3.6465591418719144e-05, "loss": 0.0001, "step": 41854 }, { "epoch": 0.81, "learning_rate": 3.646494445846192e-05, "loss": 0.0001, "step": 41856 }, { "epoch": 0.81, "learning_rate": 3.646429749820469e-05, "loss": 0.0, "step": 41858 }, { "epoch": 0.81, "learning_rate": 3.646365053794746e-05, "loss": 0.0001, "step": 41860 }, { "epoch": 0.81, "learning_rate": 3.646300357769023e-05, "loss": 0.0037, "step": 41862 }, { "epoch": 0.81, "learning_rate": 3.646235661743299e-05, "loss": 0.004, "step": 41864 }, { "epoch": 0.81, "learning_rate": 3.6461709657175765e-05, "loss": 0.0002, "step": 41866 }, { "epoch": 0.81, "learning_rate": 3.646106269691853e-05, "loss": 0.0, "step": 41868 }, { "epoch": 0.81, "learning_rate": 3.64604157366613e-05, "loss": 0.0215, "step": 41870 }, { "epoch": 0.81, "learning_rate": 3.6459768776404066e-05, "loss": 0.0, "step": 41872 }, { "epoch": 0.81, "learning_rate": 3.6459121816146835e-05, "loss": 0.0028, "step": 41874 }, { "epoch": 0.81, "learning_rate": 3.6458474855889604e-05, "loss": 0.0074, "step": 41876 }, { "epoch": 0.81, "learning_rate": 3.645782789563237e-05, "loss": 0.0, "step": 41878 }, { "epoch": 0.81, "learning_rate": 3.645718093537514e-05, "loss": 0.009, "step": 41880 }, { "epoch": 0.81, "learning_rate": 3.645653397511791e-05, "loss": 0.0001, "step": 41882 }, { "epoch": 0.81, "learning_rate": 3.645588701486068e-05, "loss": 0.0, "step": 41884 }, { "epoch": 0.81, "learning_rate": 3.645524005460344e-05, "loss": 0.0019, "step": 41886 }, { "epoch": 0.81, "learning_rate": 3.645459309434622e-05, "loss": 0.0, "step": 41888 }, { "epoch": 0.81, "learning_rate": 3.645394613408899e-05, "loss": 0.0003, "step": 41890 }, { "epoch": 0.81, "learning_rate": 3.645329917383175e-05, "loss": 0.0002, "step": 41892 }, { "epoch": 0.81, "learning_rate": 3.6452652213574526e-05, "loss": 0.0, "step": 41894 }, { "epoch": 0.81, "learning_rate": 3.645200525331729e-05, "loss": 0.0, "step": 41896 }, { "epoch": 0.81, "learning_rate": 3.6451358293060064e-05, "loss": 0.0, "step": 41898 }, { "epoch": 0.81, "learning_rate": 3.6450711332802826e-05, "loss": 0.0, "step": 41900 }, { "epoch": 0.81, "learning_rate": 3.6450064372545595e-05, "loss": 0.0, "step": 41902 }, { "epoch": 0.81, "learning_rate": 3.6449417412288364e-05, "loss": 0.0011, "step": 41904 }, { "epoch": 0.81, "learning_rate": 3.6448770452031133e-05, "loss": 0.0, "step": 41906 }, { "epoch": 0.81, "learning_rate": 3.64481234917739e-05, "loss": 0.0, "step": 41908 }, { "epoch": 0.81, "learning_rate": 3.644747653151667e-05, "loss": 0.0, "step": 41910 }, { "epoch": 0.81, "learning_rate": 3.644682957125944e-05, "loss": 0.0, "step": 41912 }, { "epoch": 0.81, "learning_rate": 3.64461826110022e-05, "loss": 0.0, "step": 41914 }, { "epoch": 0.81, "learning_rate": 3.644553565074498e-05, "loss": 0.0, "step": 41916 }, { "epoch": 0.81, "learning_rate": 3.644488869048774e-05, "loss": 0.0, "step": 41918 }, { "epoch": 0.81, "learning_rate": 3.644424173023052e-05, "loss": 0.0, "step": 41920 }, { "epoch": 0.81, "learning_rate": 3.6443594769973286e-05, "loss": 0.0002, "step": 41922 }, { "epoch": 0.81, "learning_rate": 3.644294780971605e-05, "loss": 0.0001, "step": 41924 }, { "epoch": 0.81, "learning_rate": 3.6442300849458824e-05, "loss": 0.0006, "step": 41926 }, { "epoch": 0.81, "learning_rate": 3.6441653889201587e-05, "loss": 0.0, "step": 41928 }, { "epoch": 0.81, "learning_rate": 3.6441006928944356e-05, "loss": 0.0003, "step": 41930 }, { "epoch": 0.81, "learning_rate": 3.6440359968687125e-05, "loss": 0.0, "step": 41932 }, { "epoch": 0.81, "learning_rate": 3.6439713008429894e-05, "loss": 0.0246, "step": 41934 }, { "epoch": 0.81, "learning_rate": 3.643906604817266e-05, "loss": 0.0, "step": 41936 }, { "epoch": 0.81, "learning_rate": 3.643841908791543e-05, "loss": 0.0003, "step": 41938 }, { "epoch": 0.81, "learning_rate": 3.64377721276582e-05, "loss": 0.0, "step": 41940 }, { "epoch": 0.81, "learning_rate": 3.643712516740097e-05, "loss": 0.0005, "step": 41942 }, { "epoch": 0.81, "learning_rate": 3.643647820714374e-05, "loss": 0.0, "step": 41944 }, { "epoch": 0.81, "learning_rate": 3.64358312468865e-05, "loss": 0.0063, "step": 41946 }, { "epoch": 0.81, "learning_rate": 3.643518428662928e-05, "loss": 0.0, "step": 41948 }, { "epoch": 0.81, "learning_rate": 3.643453732637204e-05, "loss": 0.0003, "step": 41950 }, { "epoch": 0.81, "learning_rate": 3.643389036611481e-05, "loss": 0.0073, "step": 41952 }, { "epoch": 0.81, "learning_rate": 3.643324340585758e-05, "loss": 0.0, "step": 41954 }, { "epoch": 0.81, "learning_rate": 3.643259644560035e-05, "loss": 0.002, "step": 41956 }, { "epoch": 0.81, "learning_rate": 3.643194948534312e-05, "loss": 0.0, "step": 41958 }, { "epoch": 0.81, "learning_rate": 3.6431302525085885e-05, "loss": 0.0036, "step": 41960 }, { "epoch": 0.81, "learning_rate": 3.6430655564828654e-05, "loss": 0.0003, "step": 41962 }, { "epoch": 0.81, "learning_rate": 3.643000860457142e-05, "loss": 0.0, "step": 41964 }, { "epoch": 0.81, "learning_rate": 3.642936164431419e-05, "loss": 0.0007, "step": 41966 }, { "epoch": 0.81, "learning_rate": 3.642871468405696e-05, "loss": 0.0074, "step": 41968 }, { "epoch": 0.81, "learning_rate": 3.642806772379973e-05, "loss": 0.0, "step": 41970 }, { "epoch": 0.81, "learning_rate": 3.64274207635425e-05, "loss": 0.0022, "step": 41972 }, { "epoch": 0.81, "learning_rate": 3.642677380328526e-05, "loss": 0.0124, "step": 41974 }, { "epoch": 0.81, "learning_rate": 3.642612684302804e-05, "loss": 0.0001, "step": 41976 }, { "epoch": 0.81, "learning_rate": 3.64254798827708e-05, "loss": 0.0, "step": 41978 }, { "epoch": 0.81, "learning_rate": 3.6424832922513576e-05, "loss": 0.0, "step": 41980 }, { "epoch": 0.81, "learning_rate": 3.642418596225634e-05, "loss": 0.0, "step": 41982 }, { "epoch": 0.81, "learning_rate": 3.642353900199911e-05, "loss": 0.0107, "step": 41984 }, { "epoch": 0.81, "learning_rate": 3.6422892041741876e-05, "loss": 0.001, "step": 41986 }, { "epoch": 0.81, "learning_rate": 3.6422245081484646e-05, "loss": 0.0, "step": 41988 }, { "epoch": 0.81, "learning_rate": 3.6421598121227415e-05, "loss": 0.0003, "step": 41990 }, { "epoch": 0.82, "learning_rate": 3.6420951160970184e-05, "loss": 0.0003, "step": 41992 }, { "epoch": 0.82, "learning_rate": 3.642030420071295e-05, "loss": 0.0023, "step": 41994 }, { "epoch": 0.82, "learning_rate": 3.6419657240455715e-05, "loss": 0.0, "step": 41996 }, { "epoch": 0.82, "learning_rate": 3.641901028019849e-05, "loss": 0.003, "step": 41998 }, { "epoch": 0.82, "learning_rate": 3.641836331994125e-05, "loss": 0.0, "step": 42000 }, { "epoch": 0.82, "learning_rate": 3.641771635968403e-05, "loss": 0.0001, "step": 42002 }, { "epoch": 0.82, "learning_rate": 3.64170693994268e-05, "loss": 0.0, "step": 42004 }, { "epoch": 0.82, "learning_rate": 3.641642243916956e-05, "loss": 0.0, "step": 42006 }, { "epoch": 0.82, "learning_rate": 3.6415775478912336e-05, "loss": 0.0007, "step": 42008 }, { "epoch": 0.82, "learning_rate": 3.64151285186551e-05, "loss": 0.0, "step": 42010 }, { "epoch": 0.82, "learning_rate": 3.641448155839787e-05, "loss": 0.0001, "step": 42012 }, { "epoch": 0.82, "learning_rate": 3.641383459814064e-05, "loss": 0.0, "step": 42014 }, { "epoch": 0.82, "learning_rate": 3.6413187637883406e-05, "loss": 0.0029, "step": 42016 }, { "epoch": 0.82, "learning_rate": 3.6412540677626175e-05, "loss": 0.0001, "step": 42018 }, { "epoch": 0.82, "learning_rate": 3.6411893717368944e-05, "loss": 0.0, "step": 42020 }, { "epoch": 0.82, "learning_rate": 3.641124675711171e-05, "loss": 0.0008, "step": 42022 }, { "epoch": 0.82, "learning_rate": 3.641059979685448e-05, "loss": 0.0, "step": 42024 }, { "epoch": 0.82, "learning_rate": 3.640995283659725e-05, "loss": 0.0, "step": 42026 }, { "epoch": 0.82, "learning_rate": 3.6409305876340014e-05, "loss": 0.0145, "step": 42028 }, { "epoch": 0.82, "learning_rate": 3.640865891608279e-05, "loss": 0.0, "step": 42030 }, { "epoch": 0.82, "learning_rate": 3.640801195582555e-05, "loss": 0.0, "step": 42032 }, { "epoch": 0.82, "learning_rate": 3.640736499556832e-05, "loss": 0.0006, "step": 42034 }, { "epoch": 0.82, "learning_rate": 3.64067180353111e-05, "loss": 0.005, "step": 42036 }, { "epoch": 0.82, "learning_rate": 3.640607107505386e-05, "loss": 0.0, "step": 42038 }, { "epoch": 0.82, "learning_rate": 3.6405424114796635e-05, "loss": 0.0, "step": 42040 }, { "epoch": 0.82, "learning_rate": 3.64047771545394e-05, "loss": 0.0031, "step": 42042 }, { "epoch": 0.82, "learning_rate": 3.6404130194282166e-05, "loss": 0.0, "step": 42044 }, { "epoch": 0.82, "learning_rate": 3.6403483234024935e-05, "loss": 0.0001, "step": 42046 }, { "epoch": 0.82, "learning_rate": 3.6402836273767705e-05, "loss": 0.0, "step": 42048 }, { "epoch": 0.82, "learning_rate": 3.6402189313510474e-05, "loss": 0.0004, "step": 42050 }, { "epoch": 0.82, "learning_rate": 3.640154235325324e-05, "loss": 0.0052, "step": 42052 }, { "epoch": 0.82, "learning_rate": 3.640089539299601e-05, "loss": 0.0001, "step": 42054 }, { "epoch": 0.82, "learning_rate": 3.6400248432738774e-05, "loss": 0.0008, "step": 42056 }, { "epoch": 0.82, "learning_rate": 3.639960147248155e-05, "loss": 0.0, "step": 42058 }, { "epoch": 0.82, "learning_rate": 3.639895451222431e-05, "loss": 0.0053, "step": 42060 }, { "epoch": 0.82, "learning_rate": 3.639830755196709e-05, "loss": 0.0004, "step": 42062 }, { "epoch": 0.82, "learning_rate": 3.639766059170985e-05, "loss": 0.0014, "step": 42064 }, { "epoch": 0.82, "learning_rate": 3.639701363145262e-05, "loss": 0.0002, "step": 42066 }, { "epoch": 0.82, "learning_rate": 3.6396366671195395e-05, "loss": 0.0087, "step": 42068 }, { "epoch": 0.82, "learning_rate": 3.639571971093816e-05, "loss": 0.0, "step": 42070 }, { "epoch": 0.82, "learning_rate": 3.639507275068093e-05, "loss": 0.0004, "step": 42072 }, { "epoch": 0.82, "learning_rate": 3.6394425790423696e-05, "loss": 0.0, "step": 42074 }, { "epoch": 0.82, "learning_rate": 3.6393778830166465e-05, "loss": 0.0099, "step": 42076 }, { "epoch": 0.82, "learning_rate": 3.639313186990923e-05, "loss": 0.0001, "step": 42078 }, { "epoch": 0.82, "learning_rate": 3.6392484909652e-05, "loss": 0.0002, "step": 42080 }, { "epoch": 0.82, "learning_rate": 3.639183794939477e-05, "loss": 0.0, "step": 42082 }, { "epoch": 0.82, "learning_rate": 3.639119098913754e-05, "loss": 0.0, "step": 42084 }, { "epoch": 0.82, "learning_rate": 3.639054402888031e-05, "loss": 0.0, "step": 42086 }, { "epoch": 0.82, "learning_rate": 3.638989706862307e-05, "loss": 0.0001, "step": 42088 }, { "epoch": 0.82, "learning_rate": 3.638925010836585e-05, "loss": 0.0001, "step": 42090 }, { "epoch": 0.82, "learning_rate": 3.638860314810861e-05, "loss": 0.0, "step": 42092 }, { "epoch": 0.82, "learning_rate": 3.638795618785138e-05, "loss": 0.0003, "step": 42094 }, { "epoch": 0.82, "learning_rate": 3.638730922759415e-05, "loss": 0.0094, "step": 42096 }, { "epoch": 0.82, "learning_rate": 3.638666226733692e-05, "loss": 0.004, "step": 42098 }, { "epoch": 0.82, "learning_rate": 3.638601530707969e-05, "loss": 0.0001, "step": 42100 }, { "epoch": 0.82, "learning_rate": 3.6385368346822456e-05, "loss": 0.0, "step": 42102 }, { "epoch": 0.82, "learning_rate": 3.6384721386565225e-05, "loss": 0.0, "step": 42104 }, { "epoch": 0.82, "learning_rate": 3.6384074426307994e-05, "loss": 0.0, "step": 42106 }, { "epoch": 0.82, "learning_rate": 3.6383427466050763e-05, "loss": 0.0, "step": 42108 }, { "epoch": 0.82, "learning_rate": 3.6382780505793526e-05, "loss": 0.0004, "step": 42110 }, { "epoch": 0.82, "learning_rate": 3.63821335455363e-05, "loss": 0.0, "step": 42112 }, { "epoch": 0.82, "learning_rate": 3.638148658527907e-05, "loss": 0.0, "step": 42114 }, { "epoch": 0.82, "learning_rate": 3.638083962502183e-05, "loss": 0.0, "step": 42116 }, { "epoch": 0.82, "learning_rate": 3.638019266476461e-05, "loss": 0.0, "step": 42118 }, { "epoch": 0.82, "learning_rate": 3.637954570450737e-05, "loss": 0.0, "step": 42120 }, { "epoch": 0.82, "learning_rate": 3.637889874425015e-05, "loss": 0.0002, "step": 42122 }, { "epoch": 0.82, "learning_rate": 3.637825178399291e-05, "loss": 0.0, "step": 42124 }, { "epoch": 0.82, "learning_rate": 3.637760482373568e-05, "loss": 0.0003, "step": 42126 }, { "epoch": 0.82, "learning_rate": 3.637695786347845e-05, "loss": 0.0, "step": 42128 }, { "epoch": 0.82, "learning_rate": 3.637631090322122e-05, "loss": 0.0, "step": 42130 }, { "epoch": 0.82, "learning_rate": 3.6375663942963986e-05, "loss": 0.0001, "step": 42132 }, { "epoch": 0.82, "learning_rate": 3.6375016982706755e-05, "loss": 0.0019, "step": 42134 }, { "epoch": 0.82, "learning_rate": 3.6374370022449524e-05, "loss": 0.0005, "step": 42136 }, { "epoch": 0.82, "learning_rate": 3.6373723062192286e-05, "loss": 0.0002, "step": 42138 }, { "epoch": 0.82, "learning_rate": 3.637307610193506e-05, "loss": 0.0, "step": 42140 }, { "epoch": 0.82, "learning_rate": 3.6372429141677824e-05, "loss": 0.0, "step": 42142 }, { "epoch": 0.82, "learning_rate": 3.63717821814206e-05, "loss": 0.0117, "step": 42144 }, { "epoch": 0.82, "learning_rate": 3.637113522116337e-05, "loss": 0.0027, "step": 42146 }, { "epoch": 0.82, "learning_rate": 3.637048826090613e-05, "loss": 0.0, "step": 42148 }, { "epoch": 0.82, "learning_rate": 3.636984130064891e-05, "loss": 0.0, "step": 42150 }, { "epoch": 0.82, "learning_rate": 3.636919434039167e-05, "loss": 0.0, "step": 42152 }, { "epoch": 0.82, "learning_rate": 3.636854738013444e-05, "loss": 0.0, "step": 42154 }, { "epoch": 0.82, "learning_rate": 3.636790041987721e-05, "loss": 0.0002, "step": 42156 }, { "epoch": 0.82, "learning_rate": 3.636725345961998e-05, "loss": 0.0002, "step": 42158 }, { "epoch": 0.82, "learning_rate": 3.6366606499362746e-05, "loss": 0.0001, "step": 42160 }, { "epoch": 0.82, "learning_rate": 3.6365959539105515e-05, "loss": 0.0, "step": 42162 }, { "epoch": 0.82, "learning_rate": 3.6365312578848284e-05, "loss": 0.0099, "step": 42164 }, { "epoch": 0.82, "learning_rate": 3.6364665618591053e-05, "loss": 0.0, "step": 42166 }, { "epoch": 0.82, "learning_rate": 3.636401865833382e-05, "loss": 0.0, "step": 42168 }, { "epoch": 0.82, "learning_rate": 3.6363371698076585e-05, "loss": 0.0, "step": 42170 }, { "epoch": 0.82, "learning_rate": 3.636272473781936e-05, "loss": 0.0001, "step": 42172 }, { "epoch": 0.82, "learning_rate": 3.636207777756212e-05, "loss": 0.0001, "step": 42174 }, { "epoch": 0.82, "learning_rate": 3.636143081730489e-05, "loss": 0.006, "step": 42176 }, { "epoch": 0.82, "learning_rate": 3.636078385704766e-05, "loss": 0.0023, "step": 42178 }, { "epoch": 0.82, "learning_rate": 3.636013689679043e-05, "loss": 0.0005, "step": 42180 }, { "epoch": 0.82, "learning_rate": 3.6359489936533206e-05, "loss": 0.0, "step": 42182 }, { "epoch": 0.82, "learning_rate": 3.635884297627597e-05, "loss": 0.0, "step": 42184 }, { "epoch": 0.82, "learning_rate": 3.635819601601874e-05, "loss": 0.0, "step": 42186 }, { "epoch": 0.82, "learning_rate": 3.6357549055761507e-05, "loss": 0.0092, "step": 42188 }, { "epoch": 0.82, "learning_rate": 3.6356902095504276e-05, "loss": 0.0, "step": 42190 }, { "epoch": 0.82, "learning_rate": 3.6356255135247045e-05, "loss": 0.0, "step": 42192 }, { "epoch": 0.82, "learning_rate": 3.6355608174989814e-05, "loss": 0.0001, "step": 42194 }, { "epoch": 0.82, "learning_rate": 3.635496121473258e-05, "loss": 0.0004, "step": 42196 }, { "epoch": 0.82, "learning_rate": 3.6354314254475345e-05, "loss": 0.0001, "step": 42198 }, { "epoch": 0.82, "learning_rate": 3.635366729421812e-05, "loss": 0.0, "step": 42200 }, { "epoch": 0.82, "learning_rate": 3.635302033396088e-05, "loss": 0.0002, "step": 42202 }, { "epoch": 0.82, "learning_rate": 3.635237337370366e-05, "loss": 0.004, "step": 42204 }, { "epoch": 0.82, "learning_rate": 3.635172641344642e-05, "loss": 0.0, "step": 42206 }, { "epoch": 0.82, "learning_rate": 3.635107945318919e-05, "loss": 0.0004, "step": 42208 }, { "epoch": 0.82, "learning_rate": 3.635043249293196e-05, "loss": 0.0, "step": 42210 }, { "epoch": 0.82, "learning_rate": 3.634978553267473e-05, "loss": 0.008, "step": 42212 }, { "epoch": 0.82, "learning_rate": 3.63491385724175e-05, "loss": 0.0, "step": 42214 }, { "epoch": 0.82, "learning_rate": 3.634849161216027e-05, "loss": 0.0227, "step": 42216 }, { "epoch": 0.82, "learning_rate": 3.6347844651903036e-05, "loss": 0.0, "step": 42218 }, { "epoch": 0.82, "learning_rate": 3.63471976916458e-05, "loss": 0.0, "step": 42220 }, { "epoch": 0.82, "learning_rate": 3.6346550731388574e-05, "loss": 0.0002, "step": 42222 }, { "epoch": 0.82, "learning_rate": 3.634590377113134e-05, "loss": 0.0012, "step": 42224 }, { "epoch": 0.82, "learning_rate": 3.634525681087411e-05, "loss": 0.0005, "step": 42226 }, { "epoch": 0.82, "learning_rate": 3.634460985061688e-05, "loss": 0.0059, "step": 42228 }, { "epoch": 0.82, "learning_rate": 3.6343962890359644e-05, "loss": 0.0031, "step": 42230 }, { "epoch": 0.82, "learning_rate": 3.634331593010242e-05, "loss": 0.0002, "step": 42232 }, { "epoch": 0.82, "learning_rate": 3.634266896984518e-05, "loss": 0.0, "step": 42234 }, { "epoch": 0.82, "learning_rate": 3.634202200958795e-05, "loss": 0.002, "step": 42236 }, { "epoch": 0.82, "learning_rate": 3.634137504933072e-05, "loss": 0.0, "step": 42238 }, { "epoch": 0.82, "learning_rate": 3.634072808907349e-05, "loss": 0.0001, "step": 42240 }, { "epoch": 0.82, "learning_rate": 3.634008112881626e-05, "loss": 0.0084, "step": 42242 }, { "epoch": 0.82, "learning_rate": 3.633943416855903e-05, "loss": 0.0, "step": 42244 }, { "epoch": 0.82, "learning_rate": 3.6338787208301796e-05, "loss": 0.0, "step": 42246 }, { "epoch": 0.82, "learning_rate": 3.6338140248044566e-05, "loss": 0.0005, "step": 42248 }, { "epoch": 0.82, "learning_rate": 3.6337493287787335e-05, "loss": 0.0235, "step": 42250 }, { "epoch": 0.82, "learning_rate": 3.63368463275301e-05, "loss": 0.0021, "step": 42252 }, { "epoch": 0.82, "learning_rate": 3.633619936727287e-05, "loss": 0.0, "step": 42254 }, { "epoch": 0.82, "learning_rate": 3.6335552407015635e-05, "loss": 0.0, "step": 42256 }, { "epoch": 0.82, "learning_rate": 3.6334905446758404e-05, "loss": 0.0, "step": 42258 }, { "epoch": 0.82, "learning_rate": 3.633425848650118e-05, "loss": 0.0, "step": 42260 }, { "epoch": 0.82, "learning_rate": 3.633361152624394e-05, "loss": 0.0, "step": 42262 }, { "epoch": 0.82, "learning_rate": 3.633296456598672e-05, "loss": 0.0, "step": 42264 }, { "epoch": 0.82, "learning_rate": 3.633231760572948e-05, "loss": 0.0, "step": 42266 }, { "epoch": 0.82, "learning_rate": 3.633167064547225e-05, "loss": 0.0001, "step": 42268 }, { "epoch": 0.82, "learning_rate": 3.633102368521502e-05, "loss": 0.0058, "step": 42270 }, { "epoch": 0.82, "learning_rate": 3.633037672495779e-05, "loss": 0.0, "step": 42272 }, { "epoch": 0.82, "learning_rate": 3.632972976470056e-05, "loss": 0.0, "step": 42274 }, { "epoch": 0.82, "learning_rate": 3.6329082804443326e-05, "loss": 0.0, "step": 42276 }, { "epoch": 0.82, "learning_rate": 3.6328435844186095e-05, "loss": 0.0017, "step": 42278 }, { "epoch": 0.82, "learning_rate": 3.632778888392886e-05, "loss": 0.0001, "step": 42280 }, { "epoch": 0.82, "learning_rate": 3.632714192367163e-05, "loss": 0.0088, "step": 42282 }, { "epoch": 0.82, "learning_rate": 3.6326494963414395e-05, "loss": 0.0, "step": 42284 }, { "epoch": 0.82, "learning_rate": 3.632584800315717e-05, "loss": 0.0004, "step": 42286 }, { "epoch": 0.82, "learning_rate": 3.6325201042899934e-05, "loss": 0.0003, "step": 42288 }, { "epoch": 0.82, "learning_rate": 3.63245540826427e-05, "loss": 0.0, "step": 42290 }, { "epoch": 0.82, "learning_rate": 3.632390712238548e-05, "loss": 0.0, "step": 42292 }, { "epoch": 0.82, "learning_rate": 3.632326016212824e-05, "loss": 0.0001, "step": 42294 }, { "epoch": 0.82, "learning_rate": 3.632261320187101e-05, "loss": 0.0002, "step": 42296 }, { "epoch": 0.82, "learning_rate": 3.632196624161378e-05, "loss": 0.0, "step": 42298 }, { "epoch": 0.82, "learning_rate": 3.632131928135655e-05, "loss": 0.0104, "step": 42300 }, { "epoch": 0.82, "learning_rate": 3.632067232109931e-05, "loss": 0.0, "step": 42302 }, { "epoch": 0.82, "learning_rate": 3.6320025360842086e-05, "loss": 0.0938, "step": 42304 }, { "epoch": 0.82, "learning_rate": 3.6319378400584855e-05, "loss": 0.0, "step": 42306 }, { "epoch": 0.82, "learning_rate": 3.6318731440327624e-05, "loss": 0.0, "step": 42308 }, { "epoch": 0.82, "learning_rate": 3.6318084480070394e-05, "loss": 0.0021, "step": 42310 }, { "epoch": 0.82, "learning_rate": 3.6317437519813156e-05, "loss": 0.0001, "step": 42312 }, { "epoch": 0.82, "learning_rate": 3.631679055955593e-05, "loss": 0.0001, "step": 42314 }, { "epoch": 0.82, "learning_rate": 3.6316143599298694e-05, "loss": 0.0, "step": 42316 }, { "epoch": 0.82, "learning_rate": 3.631549663904146e-05, "loss": 0.0026, "step": 42318 }, { "epoch": 0.82, "learning_rate": 3.631484967878423e-05, "loss": 0.0, "step": 42320 }, { "epoch": 0.82, "learning_rate": 3.6314202718527e-05, "loss": 0.0001, "step": 42322 }, { "epoch": 0.82, "learning_rate": 3.631355575826978e-05, "loss": 0.0002, "step": 42324 }, { "epoch": 0.82, "learning_rate": 3.631290879801254e-05, "loss": 0.0106, "step": 42326 }, { "epoch": 0.82, "learning_rate": 3.631226183775531e-05, "loss": 0.0, "step": 42328 }, { "epoch": 0.82, "learning_rate": 3.631161487749808e-05, "loss": 0.0, "step": 42330 }, { "epoch": 0.82, "learning_rate": 3.631096791724085e-05, "loss": 0.0, "step": 42332 }, { "epoch": 0.82, "learning_rate": 3.631032095698361e-05, "loss": 0.0001, "step": 42334 }, { "epoch": 0.82, "learning_rate": 3.6309673996726385e-05, "loss": 0.0, "step": 42336 }, { "epoch": 0.82, "learning_rate": 3.6309027036469154e-05, "loss": 0.0118, "step": 42338 }, { "epoch": 0.82, "learning_rate": 3.6308380076211916e-05, "loss": 0.0001, "step": 42340 }, { "epoch": 0.82, "learning_rate": 3.630773311595469e-05, "loss": 0.0, "step": 42342 }, { "epoch": 0.82, "learning_rate": 3.6307086155697454e-05, "loss": 0.0002, "step": 42344 }, { "epoch": 0.82, "learning_rate": 3.630643919544023e-05, "loss": 0.0043, "step": 42346 }, { "epoch": 0.82, "learning_rate": 3.630579223518299e-05, "loss": 0.0, "step": 42348 }, { "epoch": 0.82, "learning_rate": 3.630514527492576e-05, "loss": 0.0018, "step": 42350 }, { "epoch": 0.82, "learning_rate": 3.630449831466853e-05, "loss": 0.0, "step": 42352 }, { "epoch": 0.82, "learning_rate": 3.63038513544113e-05, "loss": 0.0, "step": 42354 }, { "epoch": 0.82, "learning_rate": 3.630320439415407e-05, "loss": 0.0, "step": 42356 }, { "epoch": 0.82, "learning_rate": 3.630255743389684e-05, "loss": 0.0001, "step": 42358 }, { "epoch": 0.82, "learning_rate": 3.630191047363961e-05, "loss": 0.0, "step": 42360 }, { "epoch": 0.82, "learning_rate": 3.630126351338237e-05, "loss": 0.0001, "step": 42362 }, { "epoch": 0.82, "learning_rate": 3.6300616553125145e-05, "loss": 0.0, "step": 42364 }, { "epoch": 0.82, "learning_rate": 3.629996959286791e-05, "loss": 0.0, "step": 42366 }, { "epoch": 0.82, "learning_rate": 3.6299322632610683e-05, "loss": 0.0001, "step": 42368 }, { "epoch": 0.82, "learning_rate": 3.629867567235345e-05, "loss": 0.0, "step": 42370 }, { "epoch": 0.82, "learning_rate": 3.6298028712096215e-05, "loss": 0.0, "step": 42372 }, { "epoch": 0.82, "learning_rate": 3.629738175183899e-05, "loss": 0.0002, "step": 42374 }, { "epoch": 0.82, "learning_rate": 3.629673479158175e-05, "loss": 0.0001, "step": 42376 }, { "epoch": 0.82, "learning_rate": 3.629608783132452e-05, "loss": 0.0, "step": 42378 }, { "epoch": 0.82, "learning_rate": 3.629544087106729e-05, "loss": 0.0, "step": 42380 }, { "epoch": 0.82, "learning_rate": 3.629479391081006e-05, "loss": 0.0002, "step": 42382 }, { "epoch": 0.82, "learning_rate": 3.629414695055283e-05, "loss": 0.0001, "step": 42384 }, { "epoch": 0.82, "learning_rate": 3.62934999902956e-05, "loss": 0.0004, "step": 42386 }, { "epoch": 0.82, "learning_rate": 3.629285303003837e-05, "loss": 0.0001, "step": 42388 }, { "epoch": 0.82, "learning_rate": 3.629220606978114e-05, "loss": 0.0006, "step": 42390 }, { "epoch": 0.82, "learning_rate": 3.6291559109523906e-05, "loss": 0.0, "step": 42392 }, { "epoch": 0.82, "learning_rate": 3.629091214926667e-05, "loss": 0.0, "step": 42394 }, { "epoch": 0.82, "learning_rate": 3.6290265189009444e-05, "loss": 0.0021, "step": 42396 }, { "epoch": 0.82, "learning_rate": 3.6289618228752206e-05, "loss": 0.0, "step": 42398 }, { "epoch": 0.82, "learning_rate": 3.6288971268494975e-05, "loss": 0.0, "step": 42400 }, { "epoch": 0.82, "learning_rate": 3.628832430823775e-05, "loss": 0.0005, "step": 42402 }, { "epoch": 0.82, "learning_rate": 3.6287677347980513e-05, "loss": 0.0041, "step": 42404 }, { "epoch": 0.82, "learning_rate": 3.628703038772329e-05, "loss": 0.0, "step": 42406 }, { "epoch": 0.82, "learning_rate": 3.628638342746605e-05, "loss": 0.0043, "step": 42408 }, { "epoch": 0.82, "learning_rate": 3.628573646720882e-05, "loss": 0.0007, "step": 42410 }, { "epoch": 0.82, "learning_rate": 3.628508950695159e-05, "loss": 0.0001, "step": 42412 }, { "epoch": 0.82, "learning_rate": 3.628444254669436e-05, "loss": 0.0002, "step": 42414 }, { "epoch": 0.82, "learning_rate": 3.628379558643713e-05, "loss": 0.0001, "step": 42416 }, { "epoch": 0.82, "learning_rate": 3.62831486261799e-05, "loss": 0.0005, "step": 42418 }, { "epoch": 0.82, "learning_rate": 3.6282501665922666e-05, "loss": 0.0, "step": 42420 }, { "epoch": 0.82, "learning_rate": 3.628185470566543e-05, "loss": 0.0, "step": 42422 }, { "epoch": 0.82, "learning_rate": 3.6281207745408204e-05, "loss": 0.0143, "step": 42424 }, { "epoch": 0.82, "learning_rate": 3.6280560785150967e-05, "loss": 0.0001, "step": 42426 }, { "epoch": 0.82, "learning_rate": 3.627991382489374e-05, "loss": 0.0003, "step": 42428 }, { "epoch": 0.82, "learning_rate": 3.6279266864636505e-05, "loss": 0.0, "step": 42430 }, { "epoch": 0.82, "learning_rate": 3.6278619904379274e-05, "loss": 0.0, "step": 42432 }, { "epoch": 0.82, "learning_rate": 3.627797294412204e-05, "loss": 0.0, "step": 42434 }, { "epoch": 0.82, "learning_rate": 3.627732598386481e-05, "loss": 0.0, "step": 42436 }, { "epoch": 0.82, "learning_rate": 3.627667902360758e-05, "loss": 0.0, "step": 42438 }, { "epoch": 0.82, "learning_rate": 3.627603206335035e-05, "loss": 0.0224, "step": 42440 }, { "epoch": 0.82, "learning_rate": 3.627538510309312e-05, "loss": 0.0001, "step": 42442 }, { "epoch": 0.82, "learning_rate": 3.627473814283588e-05, "loss": 0.0081, "step": 42444 }, { "epoch": 0.82, "learning_rate": 3.627409118257866e-05, "loss": 0.0044, "step": 42446 }, { "epoch": 0.82, "learning_rate": 3.6273444222321427e-05, "loss": 0.0043, "step": 42448 }, { "epoch": 0.82, "learning_rate": 3.6272797262064196e-05, "loss": 0.0019, "step": 42450 }, { "epoch": 0.82, "learning_rate": 3.6272150301806965e-05, "loss": 0.0, "step": 42452 }, { "epoch": 0.82, "learning_rate": 3.627150334154973e-05, "loss": 0.0, "step": 42454 }, { "epoch": 0.82, "learning_rate": 3.62708563812925e-05, "loss": 0.0006, "step": 42456 }, { "epoch": 0.82, "learning_rate": 3.6270209421035265e-05, "loss": 0.0078, "step": 42458 }, { "epoch": 0.82, "learning_rate": 3.6269562460778034e-05, "loss": 0.0001, "step": 42460 }, { "epoch": 0.82, "learning_rate": 3.62689155005208e-05, "loss": 0.0008, "step": 42462 }, { "epoch": 0.82, "learning_rate": 3.626826854026357e-05, "loss": 0.0003, "step": 42464 }, { "epoch": 0.82, "learning_rate": 3.626762158000634e-05, "loss": 0.0002, "step": 42466 }, { "epoch": 0.82, "learning_rate": 3.626697461974911e-05, "loss": 0.0005, "step": 42468 }, { "epoch": 0.82, "learning_rate": 3.626632765949188e-05, "loss": 0.0001, "step": 42470 }, { "epoch": 0.82, "learning_rate": 3.626568069923465e-05, "loss": 0.0053, "step": 42472 }, { "epoch": 0.82, "learning_rate": 3.626503373897742e-05, "loss": 0.0, "step": 42474 }, { "epoch": 0.82, "learning_rate": 3.626438677872018e-05, "loss": 0.0, "step": 42476 }, { "epoch": 0.82, "learning_rate": 3.6263739818462956e-05, "loss": 0.0006, "step": 42478 }, { "epoch": 0.82, "learning_rate": 3.626309285820572e-05, "loss": 0.0251, "step": 42480 }, { "epoch": 0.82, "learning_rate": 3.626244589794849e-05, "loss": 0.0, "step": 42482 }, { "epoch": 0.82, "learning_rate": 3.626179893769126e-05, "loss": 0.0004, "step": 42484 }, { "epoch": 0.82, "learning_rate": 3.6261151977434026e-05, "loss": 0.0, "step": 42486 }, { "epoch": 0.82, "learning_rate": 3.62605050171768e-05, "loss": 0.0048, "step": 42488 }, { "epoch": 0.82, "learning_rate": 3.6259858056919564e-05, "loss": 0.0005, "step": 42490 }, { "epoch": 0.82, "learning_rate": 3.625921109666233e-05, "loss": 0.0002, "step": 42492 }, { "epoch": 0.82, "learning_rate": 3.62585641364051e-05, "loss": 0.0062, "step": 42494 }, { "epoch": 0.82, "learning_rate": 3.625791717614787e-05, "loss": 0.0, "step": 42496 }, { "epoch": 0.82, "learning_rate": 3.625727021589064e-05, "loss": 0.0012, "step": 42498 }, { "epoch": 0.82, "learning_rate": 3.625662325563341e-05, "loss": 0.0007, "step": 42500 }, { "epoch": 0.82, "learning_rate": 3.625597629537618e-05, "loss": 0.0, "step": 42502 }, { "epoch": 0.82, "learning_rate": 3.625532933511894e-05, "loss": 0.0096, "step": 42504 }, { "epoch": 0.82, "learning_rate": 3.6254682374861716e-05, "loss": 0.0, "step": 42506 }, { "epoch": 0.83, "learning_rate": 3.625403541460448e-05, "loss": 0.0, "step": 42508 }, { "epoch": 0.83, "learning_rate": 3.6253388454347255e-05, "loss": 0.0, "step": 42510 }, { "epoch": 0.83, "learning_rate": 3.625274149409002e-05, "loss": 0.0, "step": 42512 }, { "epoch": 0.83, "learning_rate": 3.6252094533832786e-05, "loss": 0.0, "step": 42514 }, { "epoch": 0.83, "learning_rate": 3.625144757357556e-05, "loss": 0.0, "step": 42516 }, { "epoch": 0.83, "learning_rate": 3.6250800613318324e-05, "loss": 0.0, "step": 42518 }, { "epoch": 0.83, "learning_rate": 3.625015365306109e-05, "loss": 0.0281, "step": 42520 }, { "epoch": 0.83, "learning_rate": 3.624950669280386e-05, "loss": 0.0478, "step": 42522 }, { "epoch": 0.83, "learning_rate": 3.624885973254663e-05, "loss": 0.0034, "step": 42524 }, { "epoch": 0.83, "learning_rate": 3.62482127722894e-05, "loss": 0.0, "step": 42526 }, { "epoch": 0.83, "learning_rate": 3.624756581203217e-05, "loss": 0.0, "step": 42528 }, { "epoch": 0.83, "learning_rate": 3.624691885177494e-05, "loss": 0.0, "step": 42530 }, { "epoch": 0.83, "learning_rate": 3.624627189151771e-05, "loss": 0.0, "step": 42532 }, { "epoch": 0.83, "learning_rate": 3.624562493126048e-05, "loss": 0.0, "step": 42534 }, { "epoch": 0.83, "learning_rate": 3.624497797100324e-05, "loss": 0.0, "step": 42536 }, { "epoch": 0.83, "learning_rate": 3.6244331010746015e-05, "loss": 0.0, "step": 42538 }, { "epoch": 0.83, "learning_rate": 3.624368405048878e-05, "loss": 0.0, "step": 42540 }, { "epoch": 0.83, "learning_rate": 3.6243037090231546e-05, "loss": 0.0, "step": 42542 }, { "epoch": 0.83, "learning_rate": 3.6242390129974315e-05, "loss": 0.0, "step": 42544 }, { "epoch": 0.83, "learning_rate": 3.6241743169717085e-05, "loss": 0.0132, "step": 42546 }, { "epoch": 0.83, "learning_rate": 3.624109620945986e-05, "loss": 0.0, "step": 42548 }, { "epoch": 0.83, "learning_rate": 3.624044924920262e-05, "loss": 0.0, "step": 42550 }, { "epoch": 0.83, "learning_rate": 3.623980228894539e-05, "loss": 0.0001, "step": 42552 }, { "epoch": 0.83, "learning_rate": 3.623915532868816e-05, "loss": 0.0, "step": 42554 }, { "epoch": 0.83, "learning_rate": 3.623850836843093e-05, "loss": 0.036, "step": 42556 }, { "epoch": 0.83, "learning_rate": 3.623786140817369e-05, "loss": 0.0018, "step": 42558 }, { "epoch": 0.83, "learning_rate": 3.623721444791647e-05, "loss": 0.0001, "step": 42560 }, { "epoch": 0.83, "learning_rate": 3.623656748765924e-05, "loss": 0.0, "step": 42562 }, { "epoch": 0.83, "learning_rate": 3.6235920527402e-05, "loss": 0.0, "step": 42564 }, { "epoch": 0.83, "learning_rate": 3.6235273567144775e-05, "loss": 0.0, "step": 42566 }, { "epoch": 0.83, "learning_rate": 3.623462660688754e-05, "loss": 0.0, "step": 42568 }, { "epoch": 0.83, "learning_rate": 3.6233979646630314e-05, "loss": 0.0002, "step": 42570 }, { "epoch": 0.83, "learning_rate": 3.6233332686373076e-05, "loss": 0.0004, "step": 42572 }, { "epoch": 0.83, "learning_rate": 3.6232685726115845e-05, "loss": 0.0, "step": 42574 }, { "epoch": 0.83, "learning_rate": 3.6232038765858614e-05, "loss": 0.0, "step": 42576 }, { "epoch": 0.83, "learning_rate": 3.623139180560138e-05, "loss": 0.0001, "step": 42578 }, { "epoch": 0.83, "learning_rate": 3.623074484534415e-05, "loss": 0.0, "step": 42580 }, { "epoch": 0.83, "learning_rate": 3.623009788508692e-05, "loss": 0.0001, "step": 42582 }, { "epoch": 0.83, "learning_rate": 3.622945092482969e-05, "loss": 0.0, "step": 42584 }, { "epoch": 0.83, "learning_rate": 3.622880396457245e-05, "loss": 0.0, "step": 42586 }, { "epoch": 0.83, "learning_rate": 3.622815700431523e-05, "loss": 0.0, "step": 42588 }, { "epoch": 0.83, "learning_rate": 3.622751004405799e-05, "loss": 0.0002, "step": 42590 }, { "epoch": 0.83, "learning_rate": 3.622686308380077e-05, "loss": 0.0, "step": 42592 }, { "epoch": 0.83, "learning_rate": 3.6226216123543536e-05, "loss": 0.0079, "step": 42594 }, { "epoch": 0.83, "learning_rate": 3.62255691632863e-05, "loss": 0.0154, "step": 42596 }, { "epoch": 0.83, "learning_rate": 3.6224922203029074e-05, "loss": 0.0004, "step": 42598 }, { "epoch": 0.83, "learning_rate": 3.6224275242771836e-05, "loss": 0.0004, "step": 42600 }, { "epoch": 0.83, "learning_rate": 3.6223628282514605e-05, "loss": 0.0, "step": 42602 }, { "epoch": 0.83, "learning_rate": 3.6222981322257374e-05, "loss": 0.0, "step": 42604 }, { "epoch": 0.83, "learning_rate": 3.6222334362000144e-05, "loss": 0.0017, "step": 42606 }, { "epoch": 0.83, "learning_rate": 3.622168740174291e-05, "loss": 0.0001, "step": 42608 }, { "epoch": 0.83, "learning_rate": 3.622104044148568e-05, "loss": 0.0, "step": 42610 }, { "epoch": 0.83, "learning_rate": 3.622039348122845e-05, "loss": 0.0, "step": 42612 }, { "epoch": 0.83, "learning_rate": 3.621974652097122e-05, "loss": 0.0001, "step": 42614 }, { "epoch": 0.83, "learning_rate": 3.621909956071399e-05, "loss": 0.0, "step": 42616 }, { "epoch": 0.83, "learning_rate": 3.621845260045675e-05, "loss": 0.0, "step": 42618 }, { "epoch": 0.83, "learning_rate": 3.621780564019953e-05, "loss": 0.0, "step": 42620 }, { "epoch": 0.83, "learning_rate": 3.621715867994229e-05, "loss": 0.0, "step": 42622 }, { "epoch": 0.83, "learning_rate": 3.621651171968506e-05, "loss": 0.0192, "step": 42624 }, { "epoch": 0.83, "learning_rate": 3.6215864759427834e-05, "loss": 0.0, "step": 42626 }, { "epoch": 0.83, "learning_rate": 3.62152177991706e-05, "loss": 0.0, "step": 42628 }, { "epoch": 0.83, "learning_rate": 3.621457083891337e-05, "loss": 0.0001, "step": 42630 }, { "epoch": 0.83, "learning_rate": 3.6213923878656135e-05, "loss": 0.0, "step": 42632 }, { "epoch": 0.83, "learning_rate": 3.6213276918398904e-05, "loss": 0.0041, "step": 42634 }, { "epoch": 0.83, "learning_rate": 3.621262995814167e-05, "loss": 0.0, "step": 42636 }, { "epoch": 0.83, "learning_rate": 3.621198299788444e-05, "loss": 0.0001, "step": 42638 }, { "epoch": 0.83, "learning_rate": 3.621133603762721e-05, "loss": 0.0074, "step": 42640 }, { "epoch": 0.83, "learning_rate": 3.621068907736998e-05, "loss": 0.0, "step": 42642 }, { "epoch": 0.83, "learning_rate": 3.621004211711275e-05, "loss": 0.0, "step": 42644 }, { "epoch": 0.83, "learning_rate": 3.620939515685551e-05, "loss": 0.0, "step": 42646 }, { "epoch": 0.83, "learning_rate": 3.620874819659829e-05, "loss": 0.0, "step": 42648 }, { "epoch": 0.83, "learning_rate": 3.620810123634105e-05, "loss": 0.0001, "step": 42650 }, { "epoch": 0.83, "learning_rate": 3.6207454276083826e-05, "loss": 0.0002, "step": 42652 }, { "epoch": 0.83, "learning_rate": 3.620680731582659e-05, "loss": 0.0, "step": 42654 }, { "epoch": 0.83, "learning_rate": 3.620616035556936e-05, "loss": 0.0001, "step": 42656 }, { "epoch": 0.83, "learning_rate": 3.6205513395312126e-05, "loss": 0.0, "step": 42658 }, { "epoch": 0.83, "learning_rate": 3.6204866435054895e-05, "loss": 0.0, "step": 42660 }, { "epoch": 0.83, "learning_rate": 3.6204219474797664e-05, "loss": 0.0, "step": 42662 }, { "epoch": 0.83, "learning_rate": 3.6203572514540433e-05, "loss": 0.0, "step": 42664 }, { "epoch": 0.83, "learning_rate": 3.62029255542832e-05, "loss": 0.0308, "step": 42666 }, { "epoch": 0.83, "learning_rate": 3.620227859402597e-05, "loss": 0.0008, "step": 42668 }, { "epoch": 0.83, "learning_rate": 3.620163163376874e-05, "loss": 0.0, "step": 42670 }, { "epoch": 0.83, "learning_rate": 3.620098467351151e-05, "loss": 0.0059, "step": 42672 }, { "epoch": 0.83, "learning_rate": 3.620033771325428e-05, "loss": 0.0003, "step": 42674 }, { "epoch": 0.83, "learning_rate": 3.619969075299705e-05, "loss": 0.0, "step": 42676 }, { "epoch": 0.83, "learning_rate": 3.619904379273981e-05, "loss": 0.0, "step": 42678 }, { "epoch": 0.83, "learning_rate": 3.6198396832482586e-05, "loss": 0.0, "step": 42680 }, { "epoch": 0.83, "learning_rate": 3.619774987222535e-05, "loss": 0.0008, "step": 42682 }, { "epoch": 0.83, "learning_rate": 3.619710291196812e-05, "loss": 0.0, "step": 42684 }, { "epoch": 0.83, "learning_rate": 3.6196455951710887e-05, "loss": 0.0077, "step": 42686 }, { "epoch": 0.83, "learning_rate": 3.6195808991453656e-05, "loss": 0.0, "step": 42688 }, { "epoch": 0.83, "learning_rate": 3.6195162031196425e-05, "loss": 0.0, "step": 42690 }, { "epoch": 0.83, "learning_rate": 3.6194515070939194e-05, "loss": 0.0, "step": 42692 }, { "epoch": 0.83, "learning_rate": 3.619386811068196e-05, "loss": 0.0004, "step": 42694 }, { "epoch": 0.83, "learning_rate": 3.619322115042473e-05, "loss": 0.0081, "step": 42696 }, { "epoch": 0.83, "learning_rate": 3.61925741901675e-05, "loss": 0.0002, "step": 42698 }, { "epoch": 0.83, "learning_rate": 3.6191927229910263e-05, "loss": 0.0001, "step": 42700 }, { "epoch": 0.83, "learning_rate": 3.619128026965304e-05, "loss": 0.0, "step": 42702 }, { "epoch": 0.83, "learning_rate": 3.619063330939581e-05, "loss": 0.0075, "step": 42704 }, { "epoch": 0.83, "learning_rate": 3.618998634913857e-05, "loss": 0.0018, "step": 42706 }, { "epoch": 0.83, "learning_rate": 3.6189339388881347e-05, "loss": 0.0, "step": 42708 }, { "epoch": 0.83, "learning_rate": 3.618869242862411e-05, "loss": 0.0001, "step": 42710 }, { "epoch": 0.83, "learning_rate": 3.6188045468366885e-05, "loss": 0.0001, "step": 42712 }, { "epoch": 0.83, "learning_rate": 3.618739850810965e-05, "loss": 0.0, "step": 42714 }, { "epoch": 0.83, "learning_rate": 3.6186751547852416e-05, "loss": 0.0, "step": 42716 }, { "epoch": 0.83, "learning_rate": 3.6186104587595185e-05, "loss": 0.0, "step": 42718 }, { "epoch": 0.83, "learning_rate": 3.6185457627337954e-05, "loss": 0.0, "step": 42720 }, { "epoch": 0.83, "learning_rate": 3.618481066708072e-05, "loss": 0.0, "step": 42722 }, { "epoch": 0.83, "learning_rate": 3.618416370682349e-05, "loss": 0.0001, "step": 42724 }, { "epoch": 0.83, "learning_rate": 3.618351674656626e-05, "loss": 0.0, "step": 42726 }, { "epoch": 0.83, "learning_rate": 3.6182869786309024e-05, "loss": 0.0, "step": 42728 }, { "epoch": 0.83, "learning_rate": 3.61822228260518e-05, "loss": 0.0001, "step": 42730 }, { "epoch": 0.83, "learning_rate": 3.618157586579456e-05, "loss": 0.0122, "step": 42732 }, { "epoch": 0.83, "learning_rate": 3.618092890553734e-05, "loss": 0.0, "step": 42734 }, { "epoch": 0.83, "learning_rate": 3.61802819452801e-05, "loss": 0.0001, "step": 42736 }, { "epoch": 0.83, "learning_rate": 3.617963498502287e-05, "loss": 0.0, "step": 42738 }, { "epoch": 0.83, "learning_rate": 3.6178988024765645e-05, "loss": 0.008, "step": 42740 }, { "epoch": 0.83, "learning_rate": 3.617834106450841e-05, "loss": 0.0002, "step": 42742 }, { "epoch": 0.83, "learning_rate": 3.6177694104251176e-05, "loss": 0.0, "step": 42744 }, { "epoch": 0.83, "learning_rate": 3.6177047143993946e-05, "loss": 0.0001, "step": 42746 }, { "epoch": 0.83, "learning_rate": 3.6176400183736715e-05, "loss": 0.0005, "step": 42748 }, { "epoch": 0.83, "learning_rate": 3.6175753223479484e-05, "loss": 0.0032, "step": 42750 }, { "epoch": 0.83, "learning_rate": 3.617510626322225e-05, "loss": 0.0, "step": 42752 }, { "epoch": 0.83, "learning_rate": 3.617445930296502e-05, "loss": 0.0, "step": 42754 }, { "epoch": 0.83, "learning_rate": 3.617381234270779e-05, "loss": 0.0005, "step": 42756 }, { "epoch": 0.83, "learning_rate": 3.617316538245056e-05, "loss": 0.0, "step": 42758 }, { "epoch": 0.83, "learning_rate": 3.617251842219332e-05, "loss": 0.0001, "step": 42760 }, { "epoch": 0.83, "learning_rate": 3.61718714619361e-05, "loss": 0.0001, "step": 42762 }, { "epoch": 0.83, "learning_rate": 3.617122450167886e-05, "loss": 0.0, "step": 42764 }, { "epoch": 0.83, "learning_rate": 3.617057754142163e-05, "loss": 0.0, "step": 42766 }, { "epoch": 0.83, "learning_rate": 3.61699305811644e-05, "loss": 0.0049, "step": 42768 }, { "epoch": 0.83, "learning_rate": 3.616928362090717e-05, "loss": 0.0, "step": 42770 }, { "epoch": 0.83, "learning_rate": 3.6168636660649944e-05, "loss": 0.0037, "step": 42772 }, { "epoch": 0.83, "learning_rate": 3.6167989700392706e-05, "loss": 0.0, "step": 42774 }, { "epoch": 0.83, "learning_rate": 3.6167342740135475e-05, "loss": 0.0, "step": 42776 }, { "epoch": 0.83, "learning_rate": 3.6166695779878244e-05, "loss": 0.0, "step": 42778 }, { "epoch": 0.83, "learning_rate": 3.616604881962101e-05, "loss": 0.0, "step": 42780 }, { "epoch": 0.83, "learning_rate": 3.6165401859363776e-05, "loss": 0.0, "step": 42782 }, { "epoch": 0.83, "learning_rate": 3.616475489910655e-05, "loss": 0.0004, "step": 42784 }, { "epoch": 0.83, "learning_rate": 3.616410793884932e-05, "loss": 0.0005, "step": 42786 }, { "epoch": 0.83, "learning_rate": 3.616346097859208e-05, "loss": 0.0007, "step": 42788 }, { "epoch": 0.83, "learning_rate": 3.616281401833486e-05, "loss": 0.0407, "step": 42790 }, { "epoch": 0.83, "learning_rate": 3.616216705807762e-05, "loss": 0.0, "step": 42792 }, { "epoch": 0.83, "learning_rate": 3.61615200978204e-05, "loss": 0.0, "step": 42794 }, { "epoch": 0.83, "learning_rate": 3.616087313756316e-05, "loss": 0.0, "step": 42796 }, { "epoch": 0.83, "learning_rate": 3.616022617730593e-05, "loss": 0.0002, "step": 42798 }, { "epoch": 0.83, "learning_rate": 3.61595792170487e-05, "loss": 0.0124, "step": 42800 }, { "epoch": 0.83, "learning_rate": 3.6158932256791466e-05, "loss": 0.0, "step": 42802 }, { "epoch": 0.83, "learning_rate": 3.6158285296534235e-05, "loss": 0.0002, "step": 42804 }, { "epoch": 0.83, "learning_rate": 3.6157638336277005e-05, "loss": 0.0, "step": 42806 }, { "epoch": 0.83, "learning_rate": 3.6156991376019774e-05, "loss": 0.0, "step": 42808 }, { "epoch": 0.83, "learning_rate": 3.615634441576254e-05, "loss": 0.0, "step": 42810 }, { "epoch": 0.83, "learning_rate": 3.615569745550531e-05, "loss": 0.0006, "step": 42812 }, { "epoch": 0.83, "learning_rate": 3.6155050495248074e-05, "loss": 0.0002, "step": 42814 }, { "epoch": 0.83, "learning_rate": 3.615440353499085e-05, "loss": 0.0001, "step": 42816 }, { "epoch": 0.83, "learning_rate": 3.615375657473362e-05, "loss": 0.0, "step": 42818 }, { "epoch": 0.83, "learning_rate": 3.615310961447638e-05, "loss": 0.0, "step": 42820 }, { "epoch": 0.83, "learning_rate": 3.615246265421916e-05, "loss": 0.0, "step": 42822 }, { "epoch": 0.83, "learning_rate": 3.615181569396192e-05, "loss": 0.0037, "step": 42824 }, { "epoch": 0.83, "learning_rate": 3.615116873370469e-05, "loss": 0.0001, "step": 42826 }, { "epoch": 0.83, "learning_rate": 3.615052177344746e-05, "loss": 0.0, "step": 42828 }, { "epoch": 0.83, "learning_rate": 3.614987481319023e-05, "loss": 0.0, "step": 42830 }, { "epoch": 0.83, "learning_rate": 3.6149227852932996e-05, "loss": 0.0, "step": 42832 }, { "epoch": 0.83, "learning_rate": 3.6148580892675765e-05, "loss": 0.0, "step": 42834 }, { "epoch": 0.83, "learning_rate": 3.6147933932418534e-05, "loss": 0.0254, "step": 42836 }, { "epoch": 0.83, "learning_rate": 3.61472869721613e-05, "loss": 0.0001, "step": 42838 }, { "epoch": 0.83, "learning_rate": 3.614664001190407e-05, "loss": 0.0001, "step": 42840 }, { "epoch": 0.83, "learning_rate": 3.6145993051646835e-05, "loss": 0.0, "step": 42842 }, { "epoch": 0.83, "learning_rate": 3.614534609138961e-05, "loss": 0.0, "step": 42844 }, { "epoch": 0.83, "learning_rate": 3.614469913113237e-05, "loss": 0.0001, "step": 42846 }, { "epoch": 0.83, "learning_rate": 3.614405217087514e-05, "loss": 0.0, "step": 42848 }, { "epoch": 0.83, "learning_rate": 3.614340521061792e-05, "loss": 0.0, "step": 42850 }, { "epoch": 0.83, "learning_rate": 3.614275825036068e-05, "loss": 0.0084, "step": 42852 }, { "epoch": 0.83, "learning_rate": 3.6142111290103456e-05, "loss": 0.0, "step": 42854 }, { "epoch": 0.83, "learning_rate": 3.614146432984622e-05, "loss": 0.0002, "step": 42856 }, { "epoch": 0.83, "learning_rate": 3.614081736958899e-05, "loss": 0.0168, "step": 42858 }, { "epoch": 0.83, "learning_rate": 3.6140170409331756e-05, "loss": 0.0, "step": 42860 }, { "epoch": 0.83, "learning_rate": 3.6139523449074525e-05, "loss": 0.0, "step": 42862 }, { "epoch": 0.83, "learning_rate": 3.6138876488817294e-05, "loss": 0.0, "step": 42864 }, { "epoch": 0.83, "learning_rate": 3.6138229528560064e-05, "loss": 0.0, "step": 42866 }, { "epoch": 0.83, "learning_rate": 3.613758256830283e-05, "loss": 0.0001, "step": 42868 }, { "epoch": 0.83, "learning_rate": 3.6136935608045595e-05, "loss": 0.0, "step": 42870 }, { "epoch": 0.83, "learning_rate": 3.613628864778837e-05, "loss": 0.0248, "step": 42872 }, { "epoch": 0.83, "learning_rate": 3.613564168753113e-05, "loss": 0.0, "step": 42874 }, { "epoch": 0.83, "learning_rate": 3.613499472727391e-05, "loss": 0.0, "step": 42876 }, { "epoch": 0.83, "learning_rate": 3.613434776701667e-05, "loss": 0.0002, "step": 42878 }, { "epoch": 0.83, "learning_rate": 3.613370080675944e-05, "loss": 0.0001, "step": 42880 }, { "epoch": 0.83, "learning_rate": 3.6133053846502216e-05, "loss": 0.0, "step": 42882 }, { "epoch": 0.83, "learning_rate": 3.613240688624498e-05, "loss": 0.0007, "step": 42884 }, { "epoch": 0.83, "learning_rate": 3.613175992598775e-05, "loss": 0.0107, "step": 42886 }, { "epoch": 0.83, "learning_rate": 3.613111296573052e-05, "loss": 0.0, "step": 42888 }, { "epoch": 0.83, "learning_rate": 3.6130466005473286e-05, "loss": 0.0037, "step": 42890 }, { "epoch": 0.83, "learning_rate": 3.6129819045216055e-05, "loss": 0.0, "step": 42892 }, { "epoch": 0.83, "learning_rate": 3.6129172084958824e-05, "loss": 0.0, "step": 42894 }, { "epoch": 0.83, "learning_rate": 3.612852512470159e-05, "loss": 0.0015, "step": 42896 }, { "epoch": 0.83, "learning_rate": 3.612787816444436e-05, "loss": 0.0123, "step": 42898 }, { "epoch": 0.83, "learning_rate": 3.612723120418713e-05, "loss": 0.0, "step": 42900 }, { "epoch": 0.83, "learning_rate": 3.6126584243929893e-05, "loss": 0.0, "step": 42902 }, { "epoch": 0.83, "learning_rate": 3.612593728367267e-05, "loss": 0.0, "step": 42904 }, { "epoch": 0.83, "learning_rate": 3.612529032341543e-05, "loss": 0.0001, "step": 42906 }, { "epoch": 0.83, "learning_rate": 3.61246433631582e-05, "loss": 0.0, "step": 42908 }, { "epoch": 0.83, "learning_rate": 3.612399640290097e-05, "loss": 0.0152, "step": 42910 }, { "epoch": 0.83, "learning_rate": 3.612334944264374e-05, "loss": 0.0005, "step": 42912 }, { "epoch": 0.83, "learning_rate": 3.612270248238651e-05, "loss": 0.0, "step": 42914 }, { "epoch": 0.83, "learning_rate": 3.612205552212928e-05, "loss": 0.0, "step": 42916 }, { "epoch": 0.83, "learning_rate": 3.6121408561872046e-05, "loss": 0.0001, "step": 42918 }, { "epoch": 0.83, "learning_rate": 3.6120761601614815e-05, "loss": 0.0002, "step": 42920 }, { "epoch": 0.83, "learning_rate": 3.6120114641357584e-05, "loss": 0.0001, "step": 42922 }, { "epoch": 0.83, "learning_rate": 3.611946768110035e-05, "loss": 0.0002, "step": 42924 }, { "epoch": 0.83, "learning_rate": 3.611882072084312e-05, "loss": 0.0, "step": 42926 }, { "epoch": 0.83, "learning_rate": 3.611817376058589e-05, "loss": 0.0, "step": 42928 }, { "epoch": 0.83, "learning_rate": 3.6117526800328654e-05, "loss": 0.0004, "step": 42930 }, { "epoch": 0.83, "learning_rate": 3.611687984007143e-05, "loss": 0.0044, "step": 42932 }, { "epoch": 0.83, "learning_rate": 3.611623287981419e-05, "loss": 0.0062, "step": 42934 }, { "epoch": 0.83, "learning_rate": 3.611558591955697e-05, "loss": 0.0, "step": 42936 }, { "epoch": 0.83, "learning_rate": 3.611493895929973e-05, "loss": 0.0102, "step": 42938 }, { "epoch": 0.83, "learning_rate": 3.61142919990425e-05, "loss": 0.0039, "step": 42940 }, { "epoch": 0.83, "learning_rate": 3.611364503878527e-05, "loss": 0.0001, "step": 42942 }, { "epoch": 0.83, "learning_rate": 3.611299807852804e-05, "loss": 0.0041, "step": 42944 }, { "epoch": 0.83, "learning_rate": 3.6112351118270807e-05, "loss": 0.0, "step": 42946 }, { "epoch": 0.83, "learning_rate": 3.6111704158013576e-05, "loss": 0.0, "step": 42948 }, { "epoch": 0.83, "learning_rate": 3.6111057197756345e-05, "loss": 0.0001, "step": 42950 }, { "epoch": 0.83, "learning_rate": 3.6110410237499114e-05, "loss": 0.0, "step": 42952 }, { "epoch": 0.83, "learning_rate": 3.610976327724188e-05, "loss": 0.0, "step": 42954 }, { "epoch": 0.83, "learning_rate": 3.6109116316984645e-05, "loss": 0.0001, "step": 42956 }, { "epoch": 0.83, "learning_rate": 3.610846935672742e-05, "loss": 0.0, "step": 42958 }, { "epoch": 0.83, "learning_rate": 3.610782239647018e-05, "loss": 0.0001, "step": 42960 }, { "epoch": 0.83, "learning_rate": 3.610717543621295e-05, "loss": 0.0, "step": 42962 }, { "epoch": 0.83, "learning_rate": 3.610652847595573e-05, "loss": 0.0, "step": 42964 }, { "epoch": 0.83, "learning_rate": 3.610588151569849e-05, "loss": 0.0, "step": 42966 }, { "epoch": 0.83, "learning_rate": 3.610523455544126e-05, "loss": 0.0, "step": 42968 }, { "epoch": 0.83, "learning_rate": 3.610458759518403e-05, "loss": 0.0014, "step": 42970 }, { "epoch": 0.83, "learning_rate": 3.61039406349268e-05, "loss": 0.0102, "step": 42972 }, { "epoch": 0.83, "learning_rate": 3.610329367466957e-05, "loss": 0.0, "step": 42974 }, { "epoch": 0.83, "learning_rate": 3.6102646714412336e-05, "loss": 0.0001, "step": 42976 }, { "epoch": 0.83, "learning_rate": 3.6101999754155105e-05, "loss": 0.0, "step": 42978 }, { "epoch": 0.83, "learning_rate": 3.6101352793897874e-05, "loss": 0.0, "step": 42980 }, { "epoch": 0.83, "learning_rate": 3.610070583364064e-05, "loss": 0.0, "step": 42982 }, { "epoch": 0.83, "learning_rate": 3.6100058873383406e-05, "loss": 0.0003, "step": 42984 }, { "epoch": 0.83, "learning_rate": 3.609941191312618e-05, "loss": 0.0034, "step": 42986 }, { "epoch": 0.83, "learning_rate": 3.6098764952868944e-05, "loss": 0.004, "step": 42988 }, { "epoch": 0.83, "learning_rate": 3.609811799261171e-05, "loss": 0.0002, "step": 42990 }, { "epoch": 0.83, "learning_rate": 3.609747103235448e-05, "loss": 0.0, "step": 42992 }, { "epoch": 0.83, "learning_rate": 3.609682407209725e-05, "loss": 0.0, "step": 42994 }, { "epoch": 0.83, "learning_rate": 3.609617711184003e-05, "loss": 0.0003, "step": 42996 }, { "epoch": 0.83, "learning_rate": 3.609553015158279e-05, "loss": 0.01, "step": 42998 }, { "epoch": 0.83, "learning_rate": 3.609488319132556e-05, "loss": 0.0044, "step": 43000 }, { "epoch": 0.83, "learning_rate": 3.609423623106833e-05, "loss": 0.0004, "step": 43002 }, { "epoch": 0.83, "learning_rate": 3.6093589270811096e-05, "loss": 0.0066, "step": 43004 }, { "epoch": 0.83, "learning_rate": 3.6092942310553866e-05, "loss": 0.0, "step": 43006 }, { "epoch": 0.83, "learning_rate": 3.6092295350296635e-05, "loss": 0.0134, "step": 43008 }, { "epoch": 0.83, "learning_rate": 3.6091648390039404e-05, "loss": 0.0015, "step": 43010 }, { "epoch": 0.83, "learning_rate": 3.6091001429782166e-05, "loss": 0.0, "step": 43012 }, { "epoch": 0.83, "learning_rate": 3.609035446952494e-05, "loss": 0.0002, "step": 43014 }, { "epoch": 0.83, "learning_rate": 3.6089707509267704e-05, "loss": 0.0001, "step": 43016 }, { "epoch": 0.83, "learning_rate": 3.608906054901048e-05, "loss": 0.0032, "step": 43018 }, { "epoch": 0.83, "learning_rate": 3.608841358875324e-05, "loss": 0.0, "step": 43020 }, { "epoch": 0.84, "learning_rate": 3.608776662849601e-05, "loss": 0.0006, "step": 43022 }, { "epoch": 0.84, "learning_rate": 3.608711966823878e-05, "loss": 0.0, "step": 43024 }, { "epoch": 0.84, "learning_rate": 3.608647270798155e-05, "loss": 0.0, "step": 43026 }, { "epoch": 0.84, "learning_rate": 3.608582574772432e-05, "loss": 0.0, "step": 43028 }, { "epoch": 0.84, "learning_rate": 3.608517878746709e-05, "loss": 0.0, "step": 43030 }, { "epoch": 0.84, "learning_rate": 3.608453182720986e-05, "loss": 0.0, "step": 43032 }, { "epoch": 0.84, "learning_rate": 3.6083884866952626e-05, "loss": 0.0, "step": 43034 }, { "epoch": 0.84, "learning_rate": 3.6083237906695395e-05, "loss": 0.0009, "step": 43036 }, { "epoch": 0.84, "learning_rate": 3.608259094643816e-05, "loss": 0.0019, "step": 43038 }, { "epoch": 0.84, "learning_rate": 3.608194398618093e-05, "loss": 0.0, "step": 43040 }, { "epoch": 0.84, "learning_rate": 3.60812970259237e-05, "loss": 0.0, "step": 43042 }, { "epoch": 0.84, "learning_rate": 3.6080650065666465e-05, "loss": 0.0001, "step": 43044 }, { "epoch": 0.84, "learning_rate": 3.608000310540924e-05, "loss": 0.0001, "step": 43046 }, { "epoch": 0.84, "learning_rate": 3.6079356145152e-05, "loss": 0.0004, "step": 43048 }, { "epoch": 0.84, "learning_rate": 3.607870918489477e-05, "loss": 0.0, "step": 43050 }, { "epoch": 0.84, "learning_rate": 3.607806222463754e-05, "loss": 0.0008, "step": 43052 }, { "epoch": 0.84, "learning_rate": 3.607741526438031e-05, "loss": 0.0, "step": 43054 }, { "epoch": 0.84, "learning_rate": 3.607676830412308e-05, "loss": 0.0, "step": 43056 }, { "epoch": 0.84, "learning_rate": 3.607612134386585e-05, "loss": 0.0, "step": 43058 }, { "epoch": 0.84, "learning_rate": 3.607547438360862e-05, "loss": 0.0161, "step": 43060 }, { "epoch": 0.84, "learning_rate": 3.6074827423351386e-05, "loss": 0.0, "step": 43062 }, { "epoch": 0.84, "learning_rate": 3.6074180463094155e-05, "loss": 0.0, "step": 43064 }, { "epoch": 0.84, "learning_rate": 3.607353350283692e-05, "loss": 0.0, "step": 43066 }, { "epoch": 0.84, "learning_rate": 3.6072886542579694e-05, "loss": 0.0, "step": 43068 }, { "epoch": 0.84, "learning_rate": 3.6072239582322456e-05, "loss": 0.0001, "step": 43070 }, { "epoch": 0.84, "learning_rate": 3.6071592622065225e-05, "loss": 0.0002, "step": 43072 }, { "epoch": 0.84, "learning_rate": 3.6070945661808e-05, "loss": 0.0054, "step": 43074 }, { "epoch": 0.84, "learning_rate": 3.607029870155076e-05, "loss": 0.0002, "step": 43076 }, { "epoch": 0.84, "learning_rate": 3.606965174129354e-05, "loss": 0.0, "step": 43078 }, { "epoch": 0.84, "learning_rate": 3.60690047810363e-05, "loss": 0.0068, "step": 43080 }, { "epoch": 0.84, "learning_rate": 3.606835782077907e-05, "loss": 0.0, "step": 43082 }, { "epoch": 0.84, "learning_rate": 3.606771086052184e-05, "loss": 0.0049, "step": 43084 }, { "epoch": 0.84, "learning_rate": 3.606706390026461e-05, "loss": 0.006, "step": 43086 }, { "epoch": 0.84, "learning_rate": 3.606641694000738e-05, "loss": 0.0, "step": 43088 }, { "epoch": 0.84, "learning_rate": 3.606576997975015e-05, "loss": 0.0001, "step": 43090 }, { "epoch": 0.84, "learning_rate": 3.6065123019492916e-05, "loss": 0.0, "step": 43092 }, { "epoch": 0.84, "learning_rate": 3.6064476059235685e-05, "loss": 0.0012, "step": 43094 }, { "epoch": 0.84, "learning_rate": 3.6063829098978454e-05, "loss": 0.0, "step": 43096 }, { "epoch": 0.84, "learning_rate": 3.6063182138721216e-05, "loss": 0.0, "step": 43098 }, { "epoch": 0.84, "learning_rate": 3.606253517846399e-05, "loss": 0.0002, "step": 43100 }, { "epoch": 0.84, "learning_rate": 3.6061888218206754e-05, "loss": 0.0063, "step": 43102 }, { "epoch": 0.84, "learning_rate": 3.6061241257949524e-05, "loss": 0.004, "step": 43104 }, { "epoch": 0.84, "learning_rate": 3.60605942976923e-05, "loss": 0.0008, "step": 43106 }, { "epoch": 0.84, "learning_rate": 3.605994733743506e-05, "loss": 0.0, "step": 43108 }, { "epoch": 0.84, "learning_rate": 3.605930037717783e-05, "loss": 0.0004, "step": 43110 }, { "epoch": 0.84, "learning_rate": 3.60586534169206e-05, "loss": 0.0037, "step": 43112 }, { "epoch": 0.84, "learning_rate": 3.605800645666337e-05, "loss": 0.0, "step": 43114 }, { "epoch": 0.84, "learning_rate": 3.605735949640614e-05, "loss": 0.0177, "step": 43116 }, { "epoch": 0.84, "learning_rate": 3.605671253614891e-05, "loss": 0.0, "step": 43118 }, { "epoch": 0.84, "learning_rate": 3.6056065575891676e-05, "loss": 0.0, "step": 43120 }, { "epoch": 0.84, "learning_rate": 3.6055418615634445e-05, "loss": 0.0024, "step": 43122 }, { "epoch": 0.84, "learning_rate": 3.6054771655377214e-05, "loss": 0.0, "step": 43124 }, { "epoch": 0.84, "learning_rate": 3.605412469511998e-05, "loss": 0.0001, "step": 43126 }, { "epoch": 0.84, "learning_rate": 3.605347773486275e-05, "loss": 0.0, "step": 43128 }, { "epoch": 0.84, "learning_rate": 3.6052830774605515e-05, "loss": 0.0, "step": 43130 }, { "epoch": 0.84, "learning_rate": 3.6052183814348284e-05, "loss": 0.0, "step": 43132 }, { "epoch": 0.84, "learning_rate": 3.605153685409105e-05, "loss": 0.0002, "step": 43134 }, { "epoch": 0.84, "learning_rate": 3.605088989383382e-05, "loss": 0.0, "step": 43136 }, { "epoch": 0.84, "learning_rate": 3.605024293357659e-05, "loss": 0.0, "step": 43138 }, { "epoch": 0.84, "learning_rate": 3.604959597331936e-05, "loss": 0.0, "step": 43140 }, { "epoch": 0.84, "learning_rate": 3.604894901306213e-05, "loss": 0.0, "step": 43142 }, { "epoch": 0.84, "learning_rate": 3.60483020528049e-05, "loss": 0.0001, "step": 43144 }, { "epoch": 0.84, "learning_rate": 3.604765509254767e-05, "loss": 0.0001, "step": 43146 }, { "epoch": 0.84, "learning_rate": 3.604700813229043e-05, "loss": 0.0, "step": 43148 }, { "epoch": 0.84, "learning_rate": 3.6046361172033206e-05, "loss": 0.0002, "step": 43150 }, { "epoch": 0.84, "learning_rate": 3.6045714211775975e-05, "loss": 0.0147, "step": 43152 }, { "epoch": 0.84, "learning_rate": 3.604506725151874e-05, "loss": 0.0002, "step": 43154 }, { "epoch": 0.84, "learning_rate": 3.604442029126151e-05, "loss": 0.0, "step": 43156 }, { "epoch": 0.84, "learning_rate": 3.6043773331004275e-05, "loss": 0.0023, "step": 43158 }, { "epoch": 0.84, "learning_rate": 3.604312637074705e-05, "loss": 0.0, "step": 43160 }, { "epoch": 0.84, "learning_rate": 3.6042479410489813e-05, "loss": 0.0, "step": 43162 }, { "epoch": 0.84, "learning_rate": 3.604183245023258e-05, "loss": 0.0, "step": 43164 }, { "epoch": 0.84, "learning_rate": 3.604118548997535e-05, "loss": 0.0, "step": 43166 }, { "epoch": 0.84, "learning_rate": 3.604053852971812e-05, "loss": 0.0, "step": 43168 }, { "epoch": 0.84, "learning_rate": 3.603989156946089e-05, "loss": 0.001, "step": 43170 }, { "epoch": 0.84, "learning_rate": 3.603924460920366e-05, "loss": 0.0, "step": 43172 }, { "epoch": 0.84, "learning_rate": 3.603859764894643e-05, "loss": 0.0, "step": 43174 }, { "epoch": 0.84, "learning_rate": 3.60379506886892e-05, "loss": 0.0, "step": 43176 }, { "epoch": 0.84, "learning_rate": 3.6037303728431966e-05, "loss": 0.0, "step": 43178 }, { "epoch": 0.84, "learning_rate": 3.603665676817473e-05, "loss": 0.0003, "step": 43180 }, { "epoch": 0.84, "learning_rate": 3.6036009807917504e-05, "loss": 0.0002, "step": 43182 }, { "epoch": 0.84, "learning_rate": 3.6035362847660273e-05, "loss": 0.0007, "step": 43184 }, { "epoch": 0.84, "learning_rate": 3.6034715887403036e-05, "loss": 0.0008, "step": 43186 }, { "epoch": 0.84, "learning_rate": 3.603406892714581e-05, "loss": 0.0, "step": 43188 }, { "epoch": 0.84, "learning_rate": 3.6033421966888574e-05, "loss": 0.0001, "step": 43190 }, { "epoch": 0.84, "learning_rate": 3.603277500663134e-05, "loss": 0.0, "step": 43192 }, { "epoch": 0.84, "learning_rate": 3.603212804637411e-05, "loss": 0.0, "step": 43194 }, { "epoch": 0.84, "learning_rate": 3.603148108611688e-05, "loss": 0.0001, "step": 43196 }, { "epoch": 0.84, "learning_rate": 3.603083412585965e-05, "loss": 0.0, "step": 43198 }, { "epoch": 0.84, "learning_rate": 3.603018716560242e-05, "loss": 0.0001, "step": 43200 }, { "epoch": 0.84, "learning_rate": 3.602954020534519e-05, "loss": 0.0, "step": 43202 }, { "epoch": 0.84, "learning_rate": 3.602889324508796e-05, "loss": 0.0, "step": 43204 }, { "epoch": 0.84, "learning_rate": 3.6028246284830727e-05, "loss": 0.0, "step": 43206 }, { "epoch": 0.84, "learning_rate": 3.602759932457349e-05, "loss": 0.0001, "step": 43208 }, { "epoch": 0.84, "learning_rate": 3.6026952364316265e-05, "loss": 0.0027, "step": 43210 }, { "epoch": 0.84, "learning_rate": 3.602630540405903e-05, "loss": 0.0002, "step": 43212 }, { "epoch": 0.84, "learning_rate": 3.6025658443801796e-05, "loss": 0.0, "step": 43214 }, { "epoch": 0.84, "learning_rate": 3.6025011483544565e-05, "loss": 0.0004, "step": 43216 }, { "epoch": 0.84, "learning_rate": 3.6024364523287334e-05, "loss": 0.0072, "step": 43218 }, { "epoch": 0.84, "learning_rate": 3.602371756303011e-05, "loss": 0.0076, "step": 43220 }, { "epoch": 0.84, "learning_rate": 3.602307060277287e-05, "loss": 0.0, "step": 43222 }, { "epoch": 0.84, "learning_rate": 3.602242364251564e-05, "loss": 0.0, "step": 43224 }, { "epoch": 0.84, "learning_rate": 3.602177668225841e-05, "loss": 0.0007, "step": 43226 }, { "epoch": 0.84, "learning_rate": 3.602112972200118e-05, "loss": 0.0001, "step": 43228 }, { "epoch": 0.84, "learning_rate": 3.602048276174395e-05, "loss": 0.0, "step": 43230 }, { "epoch": 0.84, "learning_rate": 3.601983580148672e-05, "loss": 0.0001, "step": 43232 }, { "epoch": 0.84, "learning_rate": 3.601918884122949e-05, "loss": 0.0, "step": 43234 }, { "epoch": 0.84, "learning_rate": 3.6018541880972256e-05, "loss": 0.0001, "step": 43236 }, { "epoch": 0.84, "learning_rate": 3.6017894920715025e-05, "loss": 0.0, "step": 43238 }, { "epoch": 0.84, "learning_rate": 3.601724796045779e-05, "loss": 0.0001, "step": 43240 }, { "epoch": 0.84, "learning_rate": 3.601660100020056e-05, "loss": 0.0001, "step": 43242 }, { "epoch": 0.84, "learning_rate": 3.6015954039943326e-05, "loss": 0.0, "step": 43244 }, { "epoch": 0.84, "learning_rate": 3.6015307079686095e-05, "loss": 0.0011, "step": 43246 }, { "epoch": 0.84, "learning_rate": 3.6014660119428864e-05, "loss": 0.0, "step": 43248 }, { "epoch": 0.84, "learning_rate": 3.601401315917163e-05, "loss": 0.0, "step": 43250 }, { "epoch": 0.84, "learning_rate": 3.60133661989144e-05, "loss": 0.0, "step": 43252 }, { "epoch": 0.84, "learning_rate": 3.601271923865717e-05, "loss": 0.0, "step": 43254 }, { "epoch": 0.84, "learning_rate": 3.601207227839994e-05, "loss": 0.0001, "step": 43256 }, { "epoch": 0.84, "learning_rate": 3.601142531814271e-05, "loss": 0.0071, "step": 43258 }, { "epoch": 0.84, "learning_rate": 3.601077835788548e-05, "loss": 0.0022, "step": 43260 }, { "epoch": 0.84, "learning_rate": 3.601013139762824e-05, "loss": 0.0, "step": 43262 }, { "epoch": 0.84, "learning_rate": 3.6009484437371016e-05, "loss": 0.0006, "step": 43264 }, { "epoch": 0.84, "learning_rate": 3.6008837477113786e-05, "loss": 0.0, "step": 43266 }, { "epoch": 0.84, "learning_rate": 3.600819051685655e-05, "loss": 0.0001, "step": 43268 }, { "epoch": 0.84, "learning_rate": 3.6007543556599324e-05, "loss": 0.0, "step": 43270 }, { "epoch": 0.84, "learning_rate": 3.6006896596342086e-05, "loss": 0.0, "step": 43272 }, { "epoch": 0.84, "learning_rate": 3.6006249636084855e-05, "loss": 0.0002, "step": 43274 }, { "epoch": 0.84, "learning_rate": 3.6005602675827624e-05, "loss": 0.0748, "step": 43276 }, { "epoch": 0.84, "learning_rate": 3.600495571557039e-05, "loss": 0.0001, "step": 43278 }, { "epoch": 0.84, "learning_rate": 3.600430875531316e-05, "loss": 0.0, "step": 43280 }, { "epoch": 0.84, "learning_rate": 3.600366179505593e-05, "loss": 0.0001, "step": 43282 }, { "epoch": 0.84, "learning_rate": 3.60030148347987e-05, "loss": 0.0006, "step": 43284 }, { "epoch": 0.84, "learning_rate": 3.600236787454147e-05, "loss": 0.0, "step": 43286 }, { "epoch": 0.84, "learning_rate": 3.600172091428424e-05, "loss": 0.0, "step": 43288 }, { "epoch": 0.84, "learning_rate": 3.6001073954027e-05, "loss": 0.0, "step": 43290 }, { "epoch": 0.84, "learning_rate": 3.600042699376978e-05, "loss": 0.0, "step": 43292 }, { "epoch": 0.84, "learning_rate": 3.599978003351254e-05, "loss": 0.0, "step": 43294 }, { "epoch": 0.84, "learning_rate": 3.599913307325531e-05, "loss": 0.0, "step": 43296 }, { "epoch": 0.84, "learning_rate": 3.5998486112998084e-05, "loss": 0.0006, "step": 43298 }, { "epoch": 0.84, "learning_rate": 3.5997839152740846e-05, "loss": 0.0, "step": 43300 }, { "epoch": 0.84, "learning_rate": 3.599719219248362e-05, "loss": 0.0, "step": 43302 }, { "epoch": 0.84, "learning_rate": 3.5996545232226385e-05, "loss": 0.0007, "step": 43304 }, { "epoch": 0.84, "learning_rate": 3.5995898271969154e-05, "loss": 0.0, "step": 43306 }, { "epoch": 0.84, "learning_rate": 3.599525131171192e-05, "loss": 0.0016, "step": 43308 }, { "epoch": 0.84, "learning_rate": 3.599460435145469e-05, "loss": 0.0, "step": 43310 }, { "epoch": 0.84, "learning_rate": 3.599395739119746e-05, "loss": 0.0001, "step": 43312 }, { "epoch": 0.84, "learning_rate": 3.599331043094023e-05, "loss": 0.0002, "step": 43314 }, { "epoch": 0.84, "learning_rate": 3.5992663470683e-05, "loss": 0.0, "step": 43316 }, { "epoch": 0.84, "learning_rate": 3.599201651042577e-05, "loss": 0.0, "step": 43318 }, { "epoch": 0.84, "learning_rate": 3.599136955016854e-05, "loss": 0.0, "step": 43320 }, { "epoch": 0.84, "learning_rate": 3.59907225899113e-05, "loss": 0.0006, "step": 43322 }, { "epoch": 0.84, "learning_rate": 3.5990075629654075e-05, "loss": 0.0, "step": 43324 }, { "epoch": 0.84, "learning_rate": 3.598942866939684e-05, "loss": 0.0002, "step": 43326 }, { "epoch": 0.84, "learning_rate": 3.598878170913961e-05, "loss": 0.0012, "step": 43328 }, { "epoch": 0.84, "learning_rate": 3.598813474888238e-05, "loss": 0.0, "step": 43330 }, { "epoch": 0.84, "learning_rate": 3.5987487788625145e-05, "loss": 0.0051, "step": 43332 }, { "epoch": 0.84, "learning_rate": 3.5986840828367914e-05, "loss": 0.0, "step": 43334 }, { "epoch": 0.84, "learning_rate": 3.598619386811068e-05, "loss": 0.0001, "step": 43336 }, { "epoch": 0.84, "learning_rate": 3.598554690785345e-05, "loss": 0.0001, "step": 43338 }, { "epoch": 0.84, "learning_rate": 3.598489994759622e-05, "loss": 0.0, "step": 43340 }, { "epoch": 0.84, "learning_rate": 3.598425298733899e-05, "loss": 0.0, "step": 43342 }, { "epoch": 0.84, "learning_rate": 3.598360602708176e-05, "loss": 0.0001, "step": 43344 }, { "epoch": 0.84, "learning_rate": 3.598295906682453e-05, "loss": 0.0, "step": 43346 }, { "epoch": 0.84, "learning_rate": 3.59823121065673e-05, "loss": 0.0, "step": 43348 }, { "epoch": 0.84, "learning_rate": 3.598166514631006e-05, "loss": 0.0003, "step": 43350 }, { "epoch": 0.84, "learning_rate": 3.5981018186052836e-05, "loss": 0.0, "step": 43352 }, { "epoch": 0.84, "learning_rate": 3.59803712257956e-05, "loss": 0.0, "step": 43354 }, { "epoch": 0.84, "learning_rate": 3.597972426553837e-05, "loss": 0.0, "step": 43356 }, { "epoch": 0.84, "learning_rate": 3.5979077305281136e-05, "loss": 0.0015, "step": 43358 }, { "epoch": 0.84, "learning_rate": 3.5978430345023905e-05, "loss": 0.0, "step": 43360 }, { "epoch": 0.84, "learning_rate": 3.597778338476668e-05, "loss": 0.0, "step": 43362 }, { "epoch": 0.84, "learning_rate": 3.5977136424509444e-05, "loss": 0.0001, "step": 43364 }, { "epoch": 0.84, "learning_rate": 3.597648946425221e-05, "loss": 0.1263, "step": 43366 }, { "epoch": 0.84, "learning_rate": 3.597584250399498e-05, "loss": 0.0002, "step": 43368 }, { "epoch": 0.84, "learning_rate": 3.597519554373775e-05, "loss": 0.0, "step": 43370 }, { "epoch": 0.84, "learning_rate": 3.597454858348051e-05, "loss": 0.0027, "step": 43372 }, { "epoch": 0.84, "learning_rate": 3.597390162322329e-05, "loss": 0.0337, "step": 43374 }, { "epoch": 0.84, "learning_rate": 3.597325466296606e-05, "loss": 0.0, "step": 43376 }, { "epoch": 0.84, "learning_rate": 3.597260770270883e-05, "loss": 0.0, "step": 43378 }, { "epoch": 0.84, "learning_rate": 3.5971960742451596e-05, "loss": 0.0001, "step": 43380 }, { "epoch": 0.84, "learning_rate": 3.597131378219436e-05, "loss": 0.0001, "step": 43382 }, { "epoch": 0.84, "learning_rate": 3.5970666821937134e-05, "loss": 0.0, "step": 43384 }, { "epoch": 0.84, "learning_rate": 3.59700198616799e-05, "loss": 0.0, "step": 43386 }, { "epoch": 0.84, "learning_rate": 3.5969372901422666e-05, "loss": 0.0001, "step": 43388 }, { "epoch": 0.84, "learning_rate": 3.5968725941165435e-05, "loss": 0.0001, "step": 43390 }, { "epoch": 0.84, "learning_rate": 3.5968078980908204e-05, "loss": 0.0036, "step": 43392 }, { "epoch": 0.84, "learning_rate": 3.596743202065097e-05, "loss": 0.0, "step": 43394 }, { "epoch": 0.84, "learning_rate": 3.596678506039374e-05, "loss": 0.0065, "step": 43396 }, { "epoch": 0.84, "learning_rate": 3.596613810013651e-05, "loss": 0.0001, "step": 43398 }, { "epoch": 0.84, "learning_rate": 3.596549113987928e-05, "loss": 0.0039, "step": 43400 }, { "epoch": 0.84, "learning_rate": 3.596484417962205e-05, "loss": 0.0001, "step": 43402 }, { "epoch": 0.84, "learning_rate": 3.596419721936481e-05, "loss": 0.004, "step": 43404 }, { "epoch": 0.84, "learning_rate": 3.596355025910759e-05, "loss": 0.0, "step": 43406 }, { "epoch": 0.84, "learning_rate": 3.596290329885036e-05, "loss": 0.0, "step": 43408 }, { "epoch": 0.84, "learning_rate": 3.596225633859312e-05, "loss": 0.0, "step": 43410 }, { "epoch": 0.84, "learning_rate": 3.5961609378335895e-05, "loss": 0.0001, "step": 43412 }, { "epoch": 0.84, "learning_rate": 3.596096241807866e-05, "loss": 0.0002, "step": 43414 }, { "epoch": 0.84, "learning_rate": 3.5960315457821426e-05, "loss": 0.0, "step": 43416 }, { "epoch": 0.84, "learning_rate": 3.5959668497564195e-05, "loss": 0.0, "step": 43418 }, { "epoch": 0.84, "learning_rate": 3.5959021537306964e-05, "loss": 0.0, "step": 43420 }, { "epoch": 0.84, "learning_rate": 3.5958374577049733e-05, "loss": 0.0, "step": 43422 }, { "epoch": 0.84, "learning_rate": 3.59577276167925e-05, "loss": 0.006, "step": 43424 }, { "epoch": 0.84, "learning_rate": 3.595708065653527e-05, "loss": 0.0, "step": 43426 }, { "epoch": 0.84, "learning_rate": 3.595643369627804e-05, "loss": 0.0035, "step": 43428 }, { "epoch": 0.84, "learning_rate": 3.595578673602081e-05, "loss": 0.0001, "step": 43430 }, { "epoch": 0.84, "learning_rate": 3.595513977576357e-05, "loss": 0.0081, "step": 43432 }, { "epoch": 0.84, "learning_rate": 3.595449281550635e-05, "loss": 0.0, "step": 43434 }, { "epoch": 0.84, "learning_rate": 3.595384585524911e-05, "loss": 0.0, "step": 43436 }, { "epoch": 0.84, "learning_rate": 3.595319889499188e-05, "loss": 0.0029, "step": 43438 }, { "epoch": 0.84, "learning_rate": 3.595255193473465e-05, "loss": 0.0, "step": 43440 }, { "epoch": 0.84, "learning_rate": 3.595190497447742e-05, "loss": 0.0, "step": 43442 }, { "epoch": 0.84, "learning_rate": 3.595125801422019e-05, "loss": 0.0001, "step": 43444 }, { "epoch": 0.84, "learning_rate": 3.5950611053962956e-05, "loss": 0.0001, "step": 43446 }, { "epoch": 0.84, "learning_rate": 3.5949964093705725e-05, "loss": 0.0087, "step": 43448 }, { "epoch": 0.84, "learning_rate": 3.5949317133448494e-05, "loss": 0.0354, "step": 43450 }, { "epoch": 0.84, "learning_rate": 3.594867017319126e-05, "loss": 0.0074, "step": 43452 }, { "epoch": 0.84, "learning_rate": 3.594802321293403e-05, "loss": 0.0, "step": 43454 }, { "epoch": 0.84, "learning_rate": 3.59473762526768e-05, "loss": 0.0084, "step": 43456 }, { "epoch": 0.84, "learning_rate": 3.594672929241957e-05, "loss": 0.0, "step": 43458 }, { "epoch": 0.84, "learning_rate": 3.594608233216234e-05, "loss": 0.0, "step": 43460 }, { "epoch": 0.84, "learning_rate": 3.594543537190511e-05, "loss": 0.0068, "step": 43462 }, { "epoch": 0.84, "learning_rate": 3.594478841164787e-05, "loss": 0.0, "step": 43464 }, { "epoch": 0.84, "learning_rate": 3.5944141451390647e-05, "loss": 0.0, "step": 43466 }, { "epoch": 0.84, "learning_rate": 3.594349449113341e-05, "loss": 0.0036, "step": 43468 }, { "epoch": 0.84, "learning_rate": 3.594284753087618e-05, "loss": 0.0012, "step": 43470 }, { "epoch": 0.84, "learning_rate": 3.594220057061895e-05, "loss": 0.0, "step": 43472 }, { "epoch": 0.84, "learning_rate": 3.5941553610361716e-05, "loss": 0.0049, "step": 43474 }, { "epoch": 0.84, "learning_rate": 3.5940906650104485e-05, "loss": 0.0081, "step": 43476 }, { "epoch": 0.84, "learning_rate": 3.5940259689847254e-05, "loss": 0.0016, "step": 43478 }, { "epoch": 0.84, "learning_rate": 3.593961272959002e-05, "loss": 0.0137, "step": 43480 }, { "epoch": 0.84, "learning_rate": 3.593896576933279e-05, "loss": 0.0, "step": 43482 }, { "epoch": 0.84, "learning_rate": 3.593831880907556e-05, "loss": 0.0, "step": 43484 }, { "epoch": 0.84, "learning_rate": 3.593767184881833e-05, "loss": 0.0, "step": 43486 }, { "epoch": 0.84, "learning_rate": 3.59370248885611e-05, "loss": 0.0001, "step": 43488 }, { "epoch": 0.84, "learning_rate": 3.593637792830387e-05, "loss": 0.0, "step": 43490 }, { "epoch": 0.84, "learning_rate": 3.593573096804663e-05, "loss": 0.0, "step": 43492 }, { "epoch": 0.84, "learning_rate": 3.593508400778941e-05, "loss": 0.0004, "step": 43494 }, { "epoch": 0.84, "learning_rate": 3.593443704753217e-05, "loss": 0.0004, "step": 43496 }, { "epoch": 0.84, "learning_rate": 3.593379008727494e-05, "loss": 0.0093, "step": 43498 }, { "epoch": 0.84, "learning_rate": 3.593314312701771e-05, "loss": 0.0, "step": 43500 }, { "epoch": 0.84, "learning_rate": 3.5932496166760476e-05, "loss": 0.0, "step": 43502 }, { "epoch": 0.84, "learning_rate": 3.5931849206503246e-05, "loss": 0.0, "step": 43504 }, { "epoch": 0.84, "learning_rate": 3.5931202246246015e-05, "loss": 0.0, "step": 43506 }, { "epoch": 0.84, "learning_rate": 3.5930555285988784e-05, "loss": 0.0, "step": 43508 }, { "epoch": 0.84, "learning_rate": 3.592990832573155e-05, "loss": 0.0029, "step": 43510 }, { "epoch": 0.84, "learning_rate": 3.592926136547432e-05, "loss": 0.0023, "step": 43512 }, { "epoch": 0.84, "learning_rate": 3.5928614405217084e-05, "loss": 0.0002, "step": 43514 }, { "epoch": 0.84, "learning_rate": 3.592796744495986e-05, "loss": 0.0, "step": 43516 }, { "epoch": 0.84, "learning_rate": 3.592732048470262e-05, "loss": 0.0017, "step": 43518 }, { "epoch": 0.84, "learning_rate": 3.59266735244454e-05, "loss": 0.0001, "step": 43520 }, { "epoch": 0.84, "learning_rate": 3.592602656418817e-05, "loss": 0.0034, "step": 43522 }, { "epoch": 0.84, "learning_rate": 3.592537960393093e-05, "loss": 0.0, "step": 43524 }, { "epoch": 0.84, "learning_rate": 3.5924732643673706e-05, "loss": 0.0007, "step": 43526 }, { "epoch": 0.84, "learning_rate": 3.592408568341647e-05, "loss": 0.008, "step": 43528 }, { "epoch": 0.84, "learning_rate": 3.592343872315924e-05, "loss": 0.0, "step": 43530 }, { "epoch": 0.84, "learning_rate": 3.5922791762902006e-05, "loss": 0.0, "step": 43532 }, { "epoch": 0.84, "learning_rate": 3.5922144802644775e-05, "loss": 0.0014, "step": 43534 }, { "epoch": 0.84, "learning_rate": 3.5921497842387544e-05, "loss": 0.0, "step": 43536 }, { "epoch": 0.85, "learning_rate": 3.592085088213031e-05, "loss": 0.0004, "step": 43538 }, { "epoch": 0.85, "learning_rate": 3.592020392187308e-05, "loss": 0.0001, "step": 43540 }, { "epoch": 0.85, "learning_rate": 3.591955696161585e-05, "loss": 0.0005, "step": 43542 }, { "epoch": 0.85, "learning_rate": 3.591891000135862e-05, "loss": 0.004, "step": 43544 }, { "epoch": 0.85, "learning_rate": 3.591826304110138e-05, "loss": 0.0001, "step": 43546 }, { "epoch": 0.85, "learning_rate": 3.591761608084416e-05, "loss": 0.001, "step": 43548 }, { "epoch": 0.85, "learning_rate": 3.591696912058692e-05, "loss": 0.0001, "step": 43550 }, { "epoch": 0.85, "learning_rate": 3.591632216032969e-05, "loss": 0.0, "step": 43552 }, { "epoch": 0.85, "learning_rate": 3.5915675200072466e-05, "loss": 0.0, "step": 43554 }, { "epoch": 0.85, "learning_rate": 3.591502823981523e-05, "loss": 0.0, "step": 43556 }, { "epoch": 0.85, "learning_rate": 3.5914381279558e-05, "loss": 0.0002, "step": 43558 }, { "epoch": 0.85, "learning_rate": 3.5913734319300766e-05, "loss": 0.0, "step": 43560 }, { "epoch": 0.85, "learning_rate": 3.5913087359043535e-05, "loss": 0.0, "step": 43562 }, { "epoch": 0.85, "learning_rate": 3.5912440398786305e-05, "loss": 0.0, "step": 43564 }, { "epoch": 0.85, "learning_rate": 3.5911793438529074e-05, "loss": 0.0, "step": 43566 }, { "epoch": 0.85, "learning_rate": 3.591114647827184e-05, "loss": 0.0, "step": 43568 }, { "epoch": 0.85, "learning_rate": 3.591049951801461e-05, "loss": 0.0001, "step": 43570 }, { "epoch": 0.85, "learning_rate": 3.590985255775738e-05, "loss": 0.0, "step": 43572 }, { "epoch": 0.85, "learning_rate": 3.590920559750014e-05, "loss": 0.0, "step": 43574 }, { "epoch": 0.85, "learning_rate": 3.590855863724292e-05, "loss": 0.0001, "step": 43576 }, { "epoch": 0.85, "learning_rate": 3.590791167698568e-05, "loss": 0.0045, "step": 43578 }, { "epoch": 0.85, "learning_rate": 3.590726471672845e-05, "loss": 0.0001, "step": 43580 }, { "epoch": 0.85, "learning_rate": 3.590661775647122e-05, "loss": 0.0001, "step": 43582 }, { "epoch": 0.85, "learning_rate": 3.590597079621399e-05, "loss": 0.0, "step": 43584 }, { "epoch": 0.85, "learning_rate": 3.5905323835956764e-05, "loss": 0.0, "step": 43586 }, { "epoch": 0.85, "learning_rate": 3.590467687569953e-05, "loss": 0.0, "step": 43588 }, { "epoch": 0.85, "learning_rate": 3.5904029915442296e-05, "loss": 0.0, "step": 43590 }, { "epoch": 0.85, "learning_rate": 3.5903382955185065e-05, "loss": 0.0001, "step": 43592 }, { "epoch": 0.85, "learning_rate": 3.5902735994927834e-05, "loss": 0.0175, "step": 43594 }, { "epoch": 0.85, "learning_rate": 3.5902089034670596e-05, "loss": 0.0001, "step": 43596 }, { "epoch": 0.85, "learning_rate": 3.590144207441337e-05, "loss": 0.0001, "step": 43598 }, { "epoch": 0.85, "learning_rate": 3.590079511415614e-05, "loss": 0.0001, "step": 43600 }, { "epoch": 0.85, "learning_rate": 3.590014815389891e-05, "loss": 0.0, "step": 43602 }, { "epoch": 0.85, "learning_rate": 3.589950119364168e-05, "loss": 0.0226, "step": 43604 }, { "epoch": 0.85, "learning_rate": 3.589885423338444e-05, "loss": 0.0, "step": 43606 }, { "epoch": 0.85, "learning_rate": 3.5898530753255826e-05, "loss": 0.3995, "step": 43608 }, { "epoch": 0.85, "learning_rate": 3.58978837929986e-05, "loss": 0.0, "step": 43610 }, { "epoch": 0.85, "learning_rate": 3.5897236832741364e-05, "loss": 0.0086, "step": 43612 }, { "epoch": 0.85, "learning_rate": 3.589658987248414e-05, "loss": 0.0, "step": 43614 }, { "epoch": 0.85, "learning_rate": 3.58959429122269e-05, "loss": 0.0, "step": 43616 }, { "epoch": 0.85, "learning_rate": 3.589529595196967e-05, "loss": 0.0, "step": 43618 }, { "epoch": 0.85, "learning_rate": 3.589464899171244e-05, "loss": 0.0002, "step": 43620 }, { "epoch": 0.85, "learning_rate": 3.589400203145521e-05, "loss": 0.0008, "step": 43622 }, { "epoch": 0.85, "learning_rate": 3.589335507119798e-05, "loss": 0.0, "step": 43624 }, { "epoch": 0.85, "learning_rate": 3.589270811094075e-05, "loss": 0.0, "step": 43626 }, { "epoch": 0.85, "learning_rate": 3.589206115068352e-05, "loss": 0.0002, "step": 43628 }, { "epoch": 0.85, "learning_rate": 3.589141419042628e-05, "loss": 0.0, "step": 43630 }, { "epoch": 0.85, "learning_rate": 3.5890767230169055e-05, "loss": 0.0, "step": 43632 }, { "epoch": 0.85, "learning_rate": 3.589012026991182e-05, "loss": 0.0003, "step": 43634 }, { "epoch": 0.85, "learning_rate": 3.5889473309654593e-05, "loss": 0.0001, "step": 43636 }, { "epoch": 0.85, "learning_rate": 3.5888826349397356e-05, "loss": 0.0, "step": 43638 }, { "epoch": 0.85, "learning_rate": 3.5888179389140125e-05, "loss": 0.0, "step": 43640 }, { "epoch": 0.85, "learning_rate": 3.5887532428882894e-05, "loss": 0.0, "step": 43642 }, { "epoch": 0.85, "learning_rate": 3.588688546862566e-05, "loss": 0.0022, "step": 43644 }, { "epoch": 0.85, "learning_rate": 3.588623850836843e-05, "loss": 0.0, "step": 43646 }, { "epoch": 0.85, "learning_rate": 3.58855915481112e-05, "loss": 0.0044, "step": 43648 }, { "epoch": 0.85, "learning_rate": 3.588494458785397e-05, "loss": 0.0, "step": 43650 }, { "epoch": 0.85, "learning_rate": 3.588429762759673e-05, "loss": 0.0, "step": 43652 }, { "epoch": 0.85, "learning_rate": 3.588365066733951e-05, "loss": 0.0, "step": 43654 }, { "epoch": 0.85, "learning_rate": 3.588300370708228e-05, "loss": 0.0, "step": 43656 }, { "epoch": 0.85, "learning_rate": 3.588235674682505e-05, "loss": 0.0002, "step": 43658 }, { "epoch": 0.85, "learning_rate": 3.5881709786567816e-05, "loss": 0.0001, "step": 43660 }, { "epoch": 0.85, "learning_rate": 3.588106282631058e-05, "loss": 0.0, "step": 43662 }, { "epoch": 0.85, "learning_rate": 3.5880415866053354e-05, "loss": 0.0, "step": 43664 }, { "epoch": 0.85, "learning_rate": 3.5879768905796116e-05, "loss": 0.0, "step": 43666 }, { "epoch": 0.85, "learning_rate": 3.5879121945538885e-05, "loss": 0.0, "step": 43668 }, { "epoch": 0.85, "learning_rate": 3.5878474985281654e-05, "loss": 0.0013, "step": 43670 }, { "epoch": 0.85, "learning_rate": 3.5877828025024423e-05, "loss": 0.0, "step": 43672 }, { "epoch": 0.85, "learning_rate": 3.587718106476719e-05, "loss": 0.0037, "step": 43674 }, { "epoch": 0.85, "learning_rate": 3.587653410450996e-05, "loss": 0.0, "step": 43676 }, { "epoch": 0.85, "learning_rate": 3.587588714425273e-05, "loss": 0.0119, "step": 43678 }, { "epoch": 0.85, "learning_rate": 3.58752401839955e-05, "loss": 0.0021, "step": 43680 }, { "epoch": 0.85, "learning_rate": 3.587459322373827e-05, "loss": 0.0, "step": 43682 }, { "epoch": 0.85, "learning_rate": 3.587394626348103e-05, "loss": 0.0, "step": 43684 }, { "epoch": 0.85, "learning_rate": 3.587329930322381e-05, "loss": 0.0, "step": 43686 }, { "epoch": 0.85, "learning_rate": 3.5872652342966576e-05, "loss": 0.0038, "step": 43688 }, { "epoch": 0.85, "learning_rate": 3.587200538270934e-05, "loss": 0.0094, "step": 43690 }, { "epoch": 0.85, "learning_rate": 3.5871358422452114e-05, "loss": 0.0185, "step": 43692 }, { "epoch": 0.85, "learning_rate": 3.587071146219488e-05, "loss": 0.0, "step": 43694 }, { "epoch": 0.85, "learning_rate": 3.587006450193765e-05, "loss": 0.0, "step": 43696 }, { "epoch": 0.85, "learning_rate": 3.5869417541680415e-05, "loss": 0.063, "step": 43698 }, { "epoch": 0.85, "learning_rate": 3.5868770581423184e-05, "loss": 0.0, "step": 43700 }, { "epoch": 0.85, "learning_rate": 3.586812362116595e-05, "loss": 0.0001, "step": 43702 }, { "epoch": 0.85, "learning_rate": 3.586747666090872e-05, "loss": 0.0, "step": 43704 }, { "epoch": 0.85, "learning_rate": 3.586682970065149e-05, "loss": 0.0082, "step": 43706 }, { "epoch": 0.85, "learning_rate": 3.586618274039426e-05, "loss": 0.0016, "step": 43708 }, { "epoch": 0.85, "learning_rate": 3.586553578013703e-05, "loss": 0.0, "step": 43710 }, { "epoch": 0.85, "learning_rate": 3.586488881987979e-05, "loss": 0.0087, "step": 43712 }, { "epoch": 0.85, "learning_rate": 3.586424185962257e-05, "loss": 0.0, "step": 43714 }, { "epoch": 0.85, "learning_rate": 3.586359489936533e-05, "loss": 0.0001, "step": 43716 }, { "epoch": 0.85, "learning_rate": 3.5862947939108106e-05, "loss": 0.0, "step": 43718 }, { "epoch": 0.85, "learning_rate": 3.586230097885087e-05, "loss": 0.0005, "step": 43720 }, { "epoch": 0.85, "learning_rate": 3.586165401859364e-05, "loss": 0.0006, "step": 43722 }, { "epoch": 0.85, "learning_rate": 3.586100705833641e-05, "loss": 0.0, "step": 43724 }, { "epoch": 0.85, "learning_rate": 3.5860360098079175e-05, "loss": 0.0025, "step": 43726 }, { "epoch": 0.85, "learning_rate": 3.5859713137821944e-05, "loss": 0.0, "step": 43728 }, { "epoch": 0.85, "learning_rate": 3.585906617756471e-05, "loss": 0.0001, "step": 43730 }, { "epoch": 0.85, "learning_rate": 3.585841921730748e-05, "loss": 0.0, "step": 43732 }, { "epoch": 0.85, "learning_rate": 3.585777225705025e-05, "loss": 0.0007, "step": 43734 }, { "epoch": 0.85, "learning_rate": 3.585712529679302e-05, "loss": 0.0, "step": 43736 }, { "epoch": 0.85, "learning_rate": 3.585647833653579e-05, "loss": 0.0003, "step": 43738 }, { "epoch": 0.85, "learning_rate": 3.585583137627856e-05, "loss": 0.0019, "step": 43740 }, { "epoch": 0.85, "learning_rate": 3.585518441602133e-05, "loss": 0.0064, "step": 43742 }, { "epoch": 0.85, "learning_rate": 3.585453745576409e-05, "loss": 0.0, "step": 43744 }, { "epoch": 0.85, "learning_rate": 3.5853890495506866e-05, "loss": 0.0, "step": 43746 }, { "epoch": 0.85, "learning_rate": 3.585324353524963e-05, "loss": 0.0, "step": 43748 }, { "epoch": 0.85, "learning_rate": 3.58525965749924e-05, "loss": 0.0011, "step": 43750 }, { "epoch": 0.85, "learning_rate": 3.5851949614735167e-05, "loss": 0.0, "step": 43752 }, { "epoch": 0.85, "learning_rate": 3.5851302654477936e-05, "loss": 0.0, "step": 43754 }, { "epoch": 0.85, "learning_rate": 3.585065569422071e-05, "loss": 0.0, "step": 43756 }, { "epoch": 0.85, "learning_rate": 3.5850008733963474e-05, "loss": 0.0, "step": 43758 }, { "epoch": 0.85, "learning_rate": 3.584936177370624e-05, "loss": 0.0, "step": 43760 }, { "epoch": 0.85, "learning_rate": 3.584871481344901e-05, "loss": 0.0, "step": 43762 }, { "epoch": 0.85, "learning_rate": 3.584806785319178e-05, "loss": 0.0, "step": 43764 }, { "epoch": 0.85, "learning_rate": 3.584742089293455e-05, "loss": 0.0, "step": 43766 }, { "epoch": 0.85, "learning_rate": 3.584677393267732e-05, "loss": 0.0001, "step": 43768 }, { "epoch": 0.85, "learning_rate": 3.584612697242009e-05, "loss": 0.0001, "step": 43770 }, { "epoch": 0.85, "learning_rate": 3.584548001216285e-05, "loss": 0.0, "step": 43772 }, { "epoch": 0.85, "learning_rate": 3.5844833051905626e-05, "loss": 0.0, "step": 43774 }, { "epoch": 0.85, "learning_rate": 3.584418609164839e-05, "loss": 0.0, "step": 43776 }, { "epoch": 0.85, "learning_rate": 3.5843539131391165e-05, "loss": 0.0, "step": 43778 }, { "epoch": 0.85, "learning_rate": 3.584289217113393e-05, "loss": 0.0, "step": 43780 }, { "epoch": 0.85, "learning_rate": 3.5842245210876696e-05, "loss": 0.0, "step": 43782 }, { "epoch": 0.85, "learning_rate": 3.5841598250619465e-05, "loss": 0.0, "step": 43784 }, { "epoch": 0.85, "learning_rate": 3.5840951290362234e-05, "loss": 0.0012, "step": 43786 }, { "epoch": 0.85, "learning_rate": 3.5840304330105e-05, "loss": 0.0, "step": 43788 }, { "epoch": 0.85, "learning_rate": 3.583965736984777e-05, "loss": 0.0, "step": 43790 }, { "epoch": 0.85, "learning_rate": 3.583901040959054e-05, "loss": 0.0081, "step": 43792 }, { "epoch": 0.85, "learning_rate": 3.5838363449333304e-05, "loss": 0.0, "step": 43794 }, { "epoch": 0.85, "learning_rate": 3.583771648907608e-05, "loss": 0.0145, "step": 43796 }, { "epoch": 0.85, "learning_rate": 3.583706952881884e-05, "loss": 0.0, "step": 43798 }, { "epoch": 0.85, "learning_rate": 3.583642256856162e-05, "loss": 0.0525, "step": 43800 }, { "epoch": 0.85, "learning_rate": 3.583577560830439e-05, "loss": 0.0, "step": 43802 }, { "epoch": 0.85, "learning_rate": 3.583512864804715e-05, "loss": 0.0072, "step": 43804 }, { "epoch": 0.85, "learning_rate": 3.5834481687789925e-05, "loss": 0.0, "step": 43806 }, { "epoch": 0.85, "learning_rate": 3.583383472753269e-05, "loss": 0.0009, "step": 43808 }, { "epoch": 0.85, "learning_rate": 3.5833187767275456e-05, "loss": 0.0127, "step": 43810 }, { "epoch": 0.85, "learning_rate": 3.5832540807018225e-05, "loss": 0.0, "step": 43812 }, { "epoch": 0.85, "learning_rate": 3.5831893846760995e-05, "loss": 0.0, "step": 43814 }, { "epoch": 0.85, "learning_rate": 3.5831246886503764e-05, "loss": 0.0051, "step": 43816 }, { "epoch": 0.85, "learning_rate": 3.583059992624653e-05, "loss": 0.0, "step": 43818 }, { "epoch": 0.85, "learning_rate": 3.58299529659893e-05, "loss": 0.0, "step": 43820 }, { "epoch": 0.85, "learning_rate": 3.582930600573207e-05, "loss": 0.0, "step": 43822 }, { "epoch": 0.85, "learning_rate": 3.582865904547484e-05, "loss": 0.015, "step": 43824 }, { "epoch": 0.85, "learning_rate": 3.58280120852176e-05, "loss": 0.0093, "step": 43826 }, { "epoch": 0.85, "learning_rate": 3.582736512496038e-05, "loss": 0.0, "step": 43828 }, { "epoch": 0.85, "learning_rate": 3.582671816470314e-05, "loss": 0.0, "step": 43830 }, { "epoch": 0.85, "learning_rate": 3.582607120444591e-05, "loss": 0.0004, "step": 43832 }, { "epoch": 0.85, "learning_rate": 3.5825424244188685e-05, "loss": 0.0, "step": 43834 }, { "epoch": 0.85, "learning_rate": 3.582477728393145e-05, "loss": 0.021, "step": 43836 }, { "epoch": 0.85, "learning_rate": 3.5824130323674224e-05, "loss": 0.0111, "step": 43838 }, { "epoch": 0.85, "learning_rate": 3.5823483363416986e-05, "loss": 0.0, "step": 43840 }, { "epoch": 0.85, "learning_rate": 3.5822836403159755e-05, "loss": 0.0124, "step": 43842 }, { "epoch": 0.85, "learning_rate": 3.5822189442902524e-05, "loss": 0.0001, "step": 43844 }, { "epoch": 0.85, "learning_rate": 3.582154248264529e-05, "loss": 0.0, "step": 43846 }, { "epoch": 0.85, "learning_rate": 3.582089552238806e-05, "loss": 0.0001, "step": 43848 }, { "epoch": 0.85, "learning_rate": 3.582024856213083e-05, "loss": 0.0002, "step": 43850 }, { "epoch": 0.85, "learning_rate": 3.58196016018736e-05, "loss": 0.0, "step": 43852 }, { "epoch": 0.85, "learning_rate": 3.581895464161636e-05, "loss": 0.0, "step": 43854 }, { "epoch": 0.85, "learning_rate": 3.581830768135914e-05, "loss": 0.0001, "step": 43856 }, { "epoch": 0.85, "learning_rate": 3.58176607211019e-05, "loss": 0.0, "step": 43858 }, { "epoch": 0.85, "learning_rate": 3.581701376084468e-05, "loss": 0.0104, "step": 43860 }, { "epoch": 0.85, "learning_rate": 3.581636680058744e-05, "loss": 0.0006, "step": 43862 }, { "epoch": 0.85, "learning_rate": 3.581571984033021e-05, "loss": 0.0001, "step": 43864 }, { "epoch": 0.85, "learning_rate": 3.5815072880072984e-05, "loss": 0.0, "step": 43866 }, { "epoch": 0.85, "learning_rate": 3.5814425919815746e-05, "loss": 0.0, "step": 43868 }, { "epoch": 0.85, "learning_rate": 3.5813778959558515e-05, "loss": 0.0, "step": 43870 }, { "epoch": 0.85, "learning_rate": 3.5813131999301284e-05, "loss": 0.0, "step": 43872 }, { "epoch": 0.85, "learning_rate": 3.5812485039044054e-05, "loss": 0.0, "step": 43874 }, { "epoch": 0.85, "learning_rate": 3.5811838078786816e-05, "loss": 0.0, "step": 43876 }, { "epoch": 0.85, "learning_rate": 3.581119111852959e-05, "loss": 0.0, "step": 43878 }, { "epoch": 0.85, "learning_rate": 3.581054415827236e-05, "loss": 0.0, "step": 43880 }, { "epoch": 0.85, "learning_rate": 3.580989719801513e-05, "loss": 0.0002, "step": 43882 }, { "epoch": 0.85, "learning_rate": 3.58092502377579e-05, "loss": 0.0003, "step": 43884 }, { "epoch": 0.85, "learning_rate": 3.580860327750066e-05, "loss": 0.0061, "step": 43886 }, { "epoch": 0.85, "learning_rate": 3.580795631724344e-05, "loss": 0.0, "step": 43888 }, { "epoch": 0.85, "learning_rate": 3.58073093569862e-05, "loss": 0.0, "step": 43890 }, { "epoch": 0.85, "learning_rate": 3.580666239672897e-05, "loss": 0.0, "step": 43892 }, { "epoch": 0.85, "learning_rate": 3.580601543647174e-05, "loss": 0.0, "step": 43894 }, { "epoch": 0.85, "learning_rate": 3.580536847621451e-05, "loss": 0.0002, "step": 43896 }, { "epoch": 0.85, "learning_rate": 3.5804721515957276e-05, "loss": 0.0, "step": 43898 }, { "epoch": 0.85, "learning_rate": 3.5804074555700045e-05, "loss": 0.0001, "step": 43900 }, { "epoch": 0.85, "learning_rate": 3.5803427595442814e-05, "loss": 0.0, "step": 43902 }, { "epoch": 0.85, "learning_rate": 3.580278063518558e-05, "loss": 0.0, "step": 43904 }, { "epoch": 0.85, "learning_rate": 3.580213367492835e-05, "loss": 0.001, "step": 43906 }, { "epoch": 0.85, "learning_rate": 3.5801486714671114e-05, "loss": 0.004, "step": 43908 }, { "epoch": 0.85, "learning_rate": 3.580083975441389e-05, "loss": 0.0084, "step": 43910 }, { "epoch": 0.85, "learning_rate": 3.580019279415666e-05, "loss": 0.0001, "step": 43912 }, { "epoch": 0.85, "learning_rate": 3.579954583389942e-05, "loss": 0.0, "step": 43914 }, { "epoch": 0.85, "learning_rate": 3.57988988736422e-05, "loss": 0.0021, "step": 43916 }, { "epoch": 0.85, "learning_rate": 3.579825191338496e-05, "loss": 0.0, "step": 43918 }, { "epoch": 0.85, "learning_rate": 3.5797604953127736e-05, "loss": 0.0009, "step": 43920 }, { "epoch": 0.85, "learning_rate": 3.57969579928705e-05, "loss": 0.0027, "step": 43922 }, { "epoch": 0.85, "learning_rate": 3.579631103261327e-05, "loss": 0.0, "step": 43924 }, { "epoch": 0.85, "learning_rate": 3.5795664072356036e-05, "loss": 0.0, "step": 43926 }, { "epoch": 0.85, "learning_rate": 3.5795017112098805e-05, "loss": 0.0001, "step": 43928 }, { "epoch": 0.85, "learning_rate": 3.5794370151841574e-05, "loss": 0.0, "step": 43930 }, { "epoch": 0.85, "learning_rate": 3.5793723191584343e-05, "loss": 0.0, "step": 43932 }, { "epoch": 0.85, "learning_rate": 3.579307623132711e-05, "loss": 0.0001, "step": 43934 }, { "epoch": 0.85, "learning_rate": 3.5792429271069875e-05, "loss": 0.0001, "step": 43936 }, { "epoch": 0.85, "learning_rate": 3.579178231081265e-05, "loss": 0.0, "step": 43938 }, { "epoch": 0.85, "learning_rate": 3.579113535055541e-05, "loss": 0.0, "step": 43940 }, { "epoch": 0.85, "learning_rate": 3.579048839029819e-05, "loss": 0.0, "step": 43942 }, { "epoch": 0.85, "learning_rate": 3.578984143004096e-05, "loss": 0.0, "step": 43944 }, { "epoch": 0.85, "learning_rate": 3.578919446978372e-05, "loss": 0.0005, "step": 43946 }, { "epoch": 0.85, "learning_rate": 3.5788547509526496e-05, "loss": 0.0202, "step": 43948 }, { "epoch": 0.85, "learning_rate": 3.578790054926926e-05, "loss": 0.0051, "step": 43950 }, { "epoch": 0.85, "learning_rate": 3.578725358901203e-05, "loss": 0.0, "step": 43952 }, { "epoch": 0.85, "learning_rate": 3.5786606628754797e-05, "loss": 0.0004, "step": 43954 }, { "epoch": 0.85, "learning_rate": 3.5785959668497566e-05, "loss": 0.0056, "step": 43956 }, { "epoch": 0.85, "learning_rate": 3.5785312708240335e-05, "loss": 0.0, "step": 43958 }, { "epoch": 0.85, "learning_rate": 3.5784665747983104e-05, "loss": 0.0, "step": 43960 }, { "epoch": 0.85, "learning_rate": 3.578401878772587e-05, "loss": 0.0, "step": 43962 }, { "epoch": 0.85, "learning_rate": 3.578337182746864e-05, "loss": 0.0, "step": 43964 }, { "epoch": 0.85, "learning_rate": 3.578272486721141e-05, "loss": 0.0, "step": 43966 }, { "epoch": 0.85, "learning_rate": 3.5782077906954173e-05, "loss": 0.0, "step": 43968 }, { "epoch": 0.85, "learning_rate": 3.578143094669695e-05, "loss": 0.0, "step": 43970 }, { "epoch": 0.85, "learning_rate": 3.578078398643971e-05, "loss": 0.0, "step": 43972 }, { "epoch": 0.85, "learning_rate": 3.578013702618248e-05, "loss": 0.0, "step": 43974 }, { "epoch": 0.85, "learning_rate": 3.577949006592525e-05, "loss": 0.0, "step": 43976 }, { "epoch": 0.85, "learning_rate": 3.577884310566802e-05, "loss": 0.0, "step": 43978 }, { "epoch": 0.85, "learning_rate": 3.5778196145410795e-05, "loss": 0.0003, "step": 43980 }, { "epoch": 0.85, "learning_rate": 3.577754918515356e-05, "loss": 0.0, "step": 43982 }, { "epoch": 0.85, "learning_rate": 3.5776902224896326e-05, "loss": 0.0001, "step": 43984 }, { "epoch": 0.85, "learning_rate": 3.5776255264639095e-05, "loss": 0.0, "step": 43986 }, { "epoch": 0.85, "learning_rate": 3.5775608304381864e-05, "loss": 0.0, "step": 43988 }, { "epoch": 0.85, "learning_rate": 3.577496134412463e-05, "loss": 0.0, "step": 43990 }, { "epoch": 0.85, "learning_rate": 3.57743143838674e-05, "loss": 0.0, "step": 43992 }, { "epoch": 0.85, "learning_rate": 3.577366742361017e-05, "loss": 0.0, "step": 43994 }, { "epoch": 0.85, "learning_rate": 3.5773020463352934e-05, "loss": 0.0006, "step": 43996 }, { "epoch": 0.85, "learning_rate": 3.577237350309571e-05, "loss": 0.0001, "step": 43998 }, { "epoch": 0.85, "learning_rate": 3.577172654283847e-05, "loss": 0.0161, "step": 44000 }, { "epoch": 0.85, "learning_rate": 3.577107958258125e-05, "loss": 0.0006, "step": 44002 }, { "epoch": 0.85, "learning_rate": 3.577043262232401e-05, "loss": 0.0, "step": 44004 }, { "epoch": 0.85, "learning_rate": 3.576978566206678e-05, "loss": 0.0009, "step": 44006 }, { "epoch": 0.85, "learning_rate": 3.576913870180955e-05, "loss": 0.0, "step": 44008 }, { "epoch": 0.85, "learning_rate": 3.576849174155232e-05, "loss": 0.0001, "step": 44010 }, { "epoch": 0.85, "learning_rate": 3.5767844781295086e-05, "loss": 0.0, "step": 44012 }, { "epoch": 0.85, "learning_rate": 3.5767197821037856e-05, "loss": 0.0002, "step": 44014 }, { "epoch": 0.85, "learning_rate": 3.5766550860780625e-05, "loss": 0.0018, "step": 44016 }, { "epoch": 0.85, "learning_rate": 3.576590390052339e-05, "loss": 0.0001, "step": 44018 }, { "epoch": 0.85, "learning_rate": 3.576525694026616e-05, "loss": 0.0, "step": 44020 }, { "epoch": 0.85, "learning_rate": 3.5764609980008925e-05, "loss": 0.0, "step": 44022 }, { "epoch": 0.85, "learning_rate": 3.57639630197517e-05, "loss": 0.0, "step": 44024 }, { "epoch": 0.85, "learning_rate": 3.576331605949447e-05, "loss": 0.0, "step": 44026 }, { "epoch": 0.85, "learning_rate": 3.576266909923723e-05, "loss": 0.003, "step": 44028 }, { "epoch": 0.85, "learning_rate": 3.576202213898001e-05, "loss": 0.0026, "step": 44030 }, { "epoch": 0.85, "learning_rate": 3.576137517872277e-05, "loss": 0.0001, "step": 44032 }, { "epoch": 0.85, "learning_rate": 3.576072821846554e-05, "loss": 0.0, "step": 44034 }, { "epoch": 0.85, "learning_rate": 3.576008125820831e-05, "loss": 0.0011, "step": 44036 }, { "epoch": 0.85, "learning_rate": 3.575943429795108e-05, "loss": 0.0011, "step": 44038 }, { "epoch": 0.85, "learning_rate": 3.575878733769385e-05, "loss": 0.0019, "step": 44040 }, { "epoch": 0.85, "learning_rate": 3.5758140377436616e-05, "loss": 0.0, "step": 44042 }, { "epoch": 0.85, "learning_rate": 3.5757493417179385e-05, "loss": 0.0001, "step": 44044 }, { "epoch": 0.85, "learning_rate": 3.5756846456922154e-05, "loss": 0.0, "step": 44046 }, { "epoch": 0.85, "learning_rate": 3.575619949666492e-05, "loss": 0.0, "step": 44048 }, { "epoch": 0.85, "learning_rate": 3.5755552536407686e-05, "loss": 0.0, "step": 44050 }, { "epoch": 0.85, "learning_rate": 3.575490557615046e-05, "loss": 0.004, "step": 44052 }, { "epoch": 0.86, "learning_rate": 3.5754258615893224e-05, "loss": 0.0002, "step": 44054 }, { "epoch": 0.86, "learning_rate": 3.575361165563599e-05, "loss": 0.0, "step": 44056 }, { "epoch": 0.86, "learning_rate": 3.575296469537877e-05, "loss": 0.0313, "step": 44058 }, { "epoch": 0.86, "learning_rate": 3.575231773512153e-05, "loss": 0.0, "step": 44060 }, { "epoch": 0.86, "learning_rate": 3.575167077486431e-05, "loss": 0.0066, "step": 44062 }, { "epoch": 0.86, "learning_rate": 3.575102381460707e-05, "loss": 0.0006, "step": 44064 }, { "epoch": 0.86, "learning_rate": 3.575037685434984e-05, "loss": 0.0, "step": 44066 }, { "epoch": 0.86, "learning_rate": 3.574972989409261e-05, "loss": 0.0018, "step": 44068 }, { "epoch": 0.86, "learning_rate": 3.5749082933835376e-05, "loss": 0.0043, "step": 44070 }, { "epoch": 0.86, "learning_rate": 3.5748435973578145e-05, "loss": 0.0008, "step": 44072 }, { "epoch": 0.86, "learning_rate": 3.5747789013320915e-05, "loss": 0.0, "step": 44074 }, { "epoch": 0.86, "learning_rate": 3.5747142053063684e-05, "loss": 0.0, "step": 44076 }, { "epoch": 0.86, "learning_rate": 3.5746495092806446e-05, "loss": 0.0, "step": 44078 }, { "epoch": 0.86, "learning_rate": 3.574584813254922e-05, "loss": 0.0122, "step": 44080 }, { "epoch": 0.86, "learning_rate": 3.5745201172291984e-05, "loss": 0.0246, "step": 44082 }, { "epoch": 0.86, "learning_rate": 3.574455421203476e-05, "loss": 0.0005, "step": 44084 }, { "epoch": 0.86, "learning_rate": 3.574390725177752e-05, "loss": 0.0, "step": 44086 }, { "epoch": 0.86, "learning_rate": 3.574326029152029e-05, "loss": 0.0146, "step": 44088 }, { "epoch": 0.86, "learning_rate": 3.574261333126307e-05, "loss": 0.0, "step": 44090 }, { "epoch": 0.86, "learning_rate": 3.574196637100583e-05, "loss": 0.0, "step": 44092 }, { "epoch": 0.86, "learning_rate": 3.57413194107486e-05, "loss": 0.0, "step": 44094 }, { "epoch": 0.86, "learning_rate": 3.574067245049137e-05, "loss": 0.0, "step": 44096 }, { "epoch": 0.86, "learning_rate": 3.574002549023414e-05, "loss": 0.0, "step": 44098 }, { "epoch": 0.86, "learning_rate": 3.5739378529976906e-05, "loss": 0.0002, "step": 44100 }, { "epoch": 0.86, "learning_rate": 3.5738731569719675e-05, "loss": 0.0087, "step": 44102 }, { "epoch": 0.86, "learning_rate": 3.5738084609462444e-05, "loss": 0.0, "step": 44104 }, { "epoch": 0.86, "learning_rate": 3.573743764920521e-05, "loss": 0.0001, "step": 44106 }, { "epoch": 0.86, "learning_rate": 3.573679068894798e-05, "loss": 0.0, "step": 44108 }, { "epoch": 0.86, "learning_rate": 3.5736143728690745e-05, "loss": 0.0062, "step": 44110 }, { "epoch": 0.86, "learning_rate": 3.573549676843352e-05, "loss": 0.0017, "step": 44112 }, { "epoch": 0.86, "learning_rate": 3.573484980817628e-05, "loss": 0.0002, "step": 44114 }, { "epoch": 0.86, "learning_rate": 3.573420284791905e-05, "loss": 0.0, "step": 44116 }, { "epoch": 0.86, "learning_rate": 3.573355588766182e-05, "loss": 0.0002, "step": 44118 }, { "epoch": 0.86, "learning_rate": 3.573290892740459e-05, "loss": 0.0002, "step": 44120 }, { "epoch": 0.86, "learning_rate": 3.573226196714736e-05, "loss": 0.0, "step": 44122 }, { "epoch": 0.86, "learning_rate": 3.573161500689013e-05, "loss": 0.0, "step": 44124 }, { "epoch": 0.86, "learning_rate": 3.57309680466329e-05, "loss": 0.0, "step": 44126 }, { "epoch": 0.86, "learning_rate": 3.5730321086375666e-05, "loss": 0.0, "step": 44128 }, { "epoch": 0.86, "learning_rate": 3.5729674126118435e-05, "loss": 0.0002, "step": 44130 }, { "epoch": 0.86, "learning_rate": 3.57290271658612e-05, "loss": 0.0001, "step": 44132 }, { "epoch": 0.86, "learning_rate": 3.5728380205603974e-05, "loss": 0.0069, "step": 44134 }, { "epoch": 0.86, "learning_rate": 3.572773324534674e-05, "loss": 0.0, "step": 44136 }, { "epoch": 0.86, "learning_rate": 3.5727086285089505e-05, "loss": 0.0006, "step": 44138 }, { "epoch": 0.86, "learning_rate": 3.572643932483228e-05, "loss": 0.0081, "step": 44140 }, { "epoch": 0.86, "learning_rate": 3.572579236457504e-05, "loss": 0.0, "step": 44142 }, { "epoch": 0.86, "learning_rate": 3.572514540431782e-05, "loss": 0.0, "step": 44144 }, { "epoch": 0.86, "learning_rate": 3.572449844406058e-05, "loss": 0.0008, "step": 44146 }, { "epoch": 0.86, "learning_rate": 3.572385148380335e-05, "loss": 0.0001, "step": 44148 }, { "epoch": 0.86, "learning_rate": 3.572320452354612e-05, "loss": 0.0002, "step": 44150 }, { "epoch": 0.86, "learning_rate": 3.572255756328889e-05, "loss": 0.0047, "step": 44152 }, { "epoch": 0.86, "learning_rate": 3.572191060303166e-05, "loss": 0.0, "step": 44154 }, { "epoch": 0.86, "learning_rate": 3.572126364277443e-05, "loss": 0.0, "step": 44156 }, { "epoch": 0.86, "learning_rate": 3.5720616682517196e-05, "loss": 0.0, "step": 44158 }, { "epoch": 0.86, "learning_rate": 3.571996972225996e-05, "loss": 0.0006, "step": 44160 }, { "epoch": 0.86, "learning_rate": 3.5719322762002734e-05, "loss": 0.0, "step": 44162 }, { "epoch": 0.86, "learning_rate": 3.5718675801745496e-05, "loss": 0.0115, "step": 44164 }, { "epoch": 0.86, "learning_rate": 3.571802884148827e-05, "loss": 0.0005, "step": 44166 }, { "epoch": 0.86, "learning_rate": 3.571738188123104e-05, "loss": 0.0073, "step": 44168 }, { "epoch": 0.86, "learning_rate": 3.5716734920973803e-05, "loss": 0.013, "step": 44170 }, { "epoch": 0.86, "learning_rate": 3.571608796071658e-05, "loss": 0.0161, "step": 44172 }, { "epoch": 0.86, "learning_rate": 3.571544100045934e-05, "loss": 0.0068, "step": 44174 }, { "epoch": 0.86, "learning_rate": 3.571479404020211e-05, "loss": 0.0002, "step": 44176 }, { "epoch": 0.86, "learning_rate": 3.571414707994488e-05, "loss": 0.0001, "step": 44178 }, { "epoch": 0.86, "learning_rate": 3.571350011968765e-05, "loss": 0.0, "step": 44180 }, { "epoch": 0.86, "learning_rate": 3.571285315943042e-05, "loss": 0.0034, "step": 44182 }, { "epoch": 0.86, "learning_rate": 3.571220619917319e-05, "loss": 0.0008, "step": 44184 }, { "epoch": 0.86, "learning_rate": 3.5711559238915956e-05, "loss": 0.0002, "step": 44186 }, { "epoch": 0.86, "learning_rate": 3.5710912278658725e-05, "loss": 0.0113, "step": 44188 }, { "epoch": 0.86, "learning_rate": 3.5710265318401494e-05, "loss": 0.0011, "step": 44190 }, { "epoch": 0.86, "learning_rate": 3.570961835814426e-05, "loss": 0.0, "step": 44192 }, { "epoch": 0.86, "learning_rate": 3.570897139788703e-05, "loss": 0.0001, "step": 44194 }, { "epoch": 0.86, "learning_rate": 3.5708324437629795e-05, "loss": 0.0, "step": 44196 }, { "epoch": 0.86, "learning_rate": 3.5707677477372564e-05, "loss": 0.0, "step": 44198 }, { "epoch": 0.86, "learning_rate": 3.570703051711533e-05, "loss": 0.0001, "step": 44200 }, { "epoch": 0.86, "learning_rate": 3.57063835568581e-05, "loss": 0.0, "step": 44202 }, { "epoch": 0.86, "learning_rate": 3.570573659660088e-05, "loss": 0.0, "step": 44204 }, { "epoch": 0.86, "learning_rate": 3.570508963634364e-05, "loss": 0.0214, "step": 44206 }, { "epoch": 0.86, "learning_rate": 3.570444267608641e-05, "loss": 0.0066, "step": 44208 }, { "epoch": 0.86, "learning_rate": 3.570379571582918e-05, "loss": 0.0, "step": 44210 }, { "epoch": 0.86, "learning_rate": 3.570314875557195e-05, "loss": 0.0, "step": 44212 }, { "epoch": 0.86, "learning_rate": 3.5702501795314717e-05, "loss": 0.0001, "step": 44214 }, { "epoch": 0.86, "learning_rate": 3.5701854835057486e-05, "loss": 0.0569, "step": 44216 }, { "epoch": 0.86, "learning_rate": 3.5701207874800255e-05, "loss": 0.0, "step": 44218 }, { "epoch": 0.86, "learning_rate": 3.570056091454302e-05, "loss": 0.0, "step": 44220 }, { "epoch": 0.86, "learning_rate": 3.569991395428579e-05, "loss": 0.0, "step": 44222 }, { "epoch": 0.86, "learning_rate": 3.5699266994028555e-05, "loss": 0.0001, "step": 44224 }, { "epoch": 0.86, "learning_rate": 3.569862003377133e-05, "loss": 0.0, "step": 44226 }, { "epoch": 0.86, "learning_rate": 3.5697973073514093e-05, "loss": 0.0, "step": 44228 }, { "epoch": 0.86, "learning_rate": 3.569732611325686e-05, "loss": 0.0, "step": 44230 }, { "epoch": 0.86, "learning_rate": 3.569667915299963e-05, "loss": 0.0, "step": 44232 }, { "epoch": 0.86, "learning_rate": 3.56960321927424e-05, "loss": 0.0, "step": 44234 }, { "epoch": 0.86, "learning_rate": 3.569538523248517e-05, "loss": 0.0, "step": 44236 }, { "epoch": 0.86, "learning_rate": 3.569473827222794e-05, "loss": 0.0, "step": 44238 }, { "epoch": 0.86, "learning_rate": 3.569409131197071e-05, "loss": 0.0, "step": 44240 }, { "epoch": 0.86, "learning_rate": 3.569344435171348e-05, "loss": 0.0001, "step": 44242 }, { "epoch": 0.86, "learning_rate": 3.5692797391456246e-05, "loss": 0.0, "step": 44244 }, { "epoch": 0.86, "learning_rate": 3.5692150431199015e-05, "loss": 0.0049, "step": 44246 }, { "epoch": 0.86, "learning_rate": 3.5691503470941784e-05, "loss": 0.0004, "step": 44248 }, { "epoch": 0.86, "learning_rate": 3.569085651068455e-05, "loss": 0.0, "step": 44250 }, { "epoch": 0.86, "learning_rate": 3.5690209550427316e-05, "loss": 0.0151, "step": 44252 }, { "epoch": 0.86, "learning_rate": 3.568956259017009e-05, "loss": 0.0, "step": 44254 }, { "epoch": 0.86, "learning_rate": 3.5688915629912854e-05, "loss": 0.0, "step": 44256 }, { "epoch": 0.86, "learning_rate": 3.568826866965562e-05, "loss": 0.0001, "step": 44258 }, { "epoch": 0.86, "learning_rate": 3.568762170939839e-05, "loss": 0.0, "step": 44260 }, { "epoch": 0.86, "learning_rate": 3.568697474914116e-05, "loss": 0.0001, "step": 44262 }, { "epoch": 0.86, "learning_rate": 3.568632778888393e-05, "loss": 0.0, "step": 44264 }, { "epoch": 0.86, "learning_rate": 3.56856808286267e-05, "loss": 0.0006, "step": 44266 }, { "epoch": 0.86, "learning_rate": 3.568503386836947e-05, "loss": 0.0, "step": 44268 }, { "epoch": 0.86, "learning_rate": 3.568438690811224e-05, "loss": 0.0, "step": 44270 }, { "epoch": 0.86, "learning_rate": 3.5683739947855006e-05, "loss": 0.0, "step": 44272 }, { "epoch": 0.86, "learning_rate": 3.568309298759777e-05, "loss": 0.0001, "step": 44274 }, { "epoch": 0.86, "learning_rate": 3.5682446027340545e-05, "loss": 0.0001, "step": 44276 }, { "epoch": 0.86, "learning_rate": 3.568179906708331e-05, "loss": 0.0, "step": 44278 }, { "epoch": 0.86, "learning_rate": 3.5681152106826076e-05, "loss": 0.0, "step": 44280 }, { "epoch": 0.86, "learning_rate": 3.568050514656885e-05, "loss": 0.0001, "step": 44282 }, { "epoch": 0.86, "learning_rate": 3.5679858186311614e-05, "loss": 0.0, "step": 44284 }, { "epoch": 0.86, "learning_rate": 3.567921122605439e-05, "loss": 0.0003, "step": 44286 }, { "epoch": 0.86, "learning_rate": 3.567856426579715e-05, "loss": 0.0001, "step": 44288 }, { "epoch": 0.86, "learning_rate": 3.567791730553992e-05, "loss": 0.0, "step": 44290 }, { "epoch": 0.86, "learning_rate": 3.567727034528269e-05, "loss": 0.0, "step": 44292 }, { "epoch": 0.86, "learning_rate": 3.567662338502546e-05, "loss": 0.0048, "step": 44294 }, { "epoch": 0.86, "learning_rate": 3.567597642476823e-05, "loss": 0.002, "step": 44296 }, { "epoch": 0.86, "learning_rate": 3.5675329464511e-05, "loss": 0.0, "step": 44298 }, { "epoch": 0.86, "learning_rate": 3.567468250425377e-05, "loss": 0.0, "step": 44300 }, { "epoch": 0.86, "learning_rate": 3.567403554399653e-05, "loss": 0.0005, "step": 44302 }, { "epoch": 0.86, "learning_rate": 3.5673388583739305e-05, "loss": 0.0, "step": 44304 }, { "epoch": 0.86, "learning_rate": 3.567274162348207e-05, "loss": 0.0, "step": 44306 }, { "epoch": 0.86, "learning_rate": 3.567209466322484e-05, "loss": 0.0013, "step": 44308 }, { "epoch": 0.86, "learning_rate": 3.5671447702967606e-05, "loss": 0.0121, "step": 44310 }, { "epoch": 0.86, "learning_rate": 3.5670800742710375e-05, "loss": 0.0, "step": 44312 }, { "epoch": 0.86, "learning_rate": 3.567015378245315e-05, "loss": 0.0, "step": 44314 }, { "epoch": 0.86, "learning_rate": 3.566950682219591e-05, "loss": 0.0, "step": 44316 }, { "epoch": 0.86, "learning_rate": 3.566885986193868e-05, "loss": 0.0002, "step": 44318 }, { "epoch": 0.86, "learning_rate": 3.566821290168145e-05, "loss": 0.0011, "step": 44320 }, { "epoch": 0.86, "learning_rate": 3.566756594142422e-05, "loss": 0.0, "step": 44322 }, { "epoch": 0.86, "learning_rate": 3.566691898116699e-05, "loss": 0.0001, "step": 44324 }, { "epoch": 0.86, "learning_rate": 3.566627202090976e-05, "loss": 0.007, "step": 44326 }, { "epoch": 0.86, "learning_rate": 3.566562506065253e-05, "loss": 0.0004, "step": 44328 }, { "epoch": 0.86, "learning_rate": 3.5664978100395296e-05, "loss": 0.0, "step": 44330 }, { "epoch": 0.86, "learning_rate": 3.5664331140138065e-05, "loss": 0.0, "step": 44332 }, { "epoch": 0.86, "learning_rate": 3.566368417988083e-05, "loss": 0.0, "step": 44334 }, { "epoch": 0.86, "learning_rate": 3.5663037219623604e-05, "loss": 0.0001, "step": 44336 }, { "epoch": 0.86, "learning_rate": 3.5662390259366366e-05, "loss": 0.0, "step": 44338 }, { "epoch": 0.86, "learning_rate": 3.5661743299109135e-05, "loss": 0.0004, "step": 44340 }, { "epoch": 0.86, "learning_rate": 3.5661096338851904e-05, "loss": 0.001, "step": 44342 }, { "epoch": 0.86, "learning_rate": 3.566044937859467e-05, "loss": 0.0005, "step": 44344 }, { "epoch": 0.86, "learning_rate": 3.565980241833745e-05, "loss": 0.0006, "step": 44346 }, { "epoch": 0.86, "learning_rate": 3.565915545808021e-05, "loss": 0.0, "step": 44348 }, { "epoch": 0.86, "learning_rate": 3.565850849782298e-05, "loss": 0.0001, "step": 44350 }, { "epoch": 0.86, "learning_rate": 3.565786153756575e-05, "loss": 0.0, "step": 44352 }, { "epoch": 0.86, "learning_rate": 3.565721457730852e-05, "loss": 0.0, "step": 44354 }, { "epoch": 0.86, "learning_rate": 3.565656761705128e-05, "loss": 0.0001, "step": 44356 }, { "epoch": 0.86, "learning_rate": 3.565592065679406e-05, "loss": 0.0, "step": 44358 }, { "epoch": 0.86, "learning_rate": 3.5655273696536826e-05, "loss": 0.0, "step": 44360 }, { "epoch": 0.86, "learning_rate": 3.565462673627959e-05, "loss": 0.0002, "step": 44362 }, { "epoch": 0.86, "learning_rate": 3.5653979776022364e-05, "loss": 0.0013, "step": 44364 }, { "epoch": 0.86, "learning_rate": 3.5653332815765126e-05, "loss": 0.0003, "step": 44366 }, { "epoch": 0.86, "learning_rate": 3.56526858555079e-05, "loss": 0.0001, "step": 44368 }, { "epoch": 0.86, "learning_rate": 3.5652038895250664e-05, "loss": 0.0129, "step": 44370 }, { "epoch": 0.86, "learning_rate": 3.5651391934993434e-05, "loss": 0.0, "step": 44372 }, { "epoch": 0.86, "learning_rate": 3.56507449747362e-05, "loss": 0.0, "step": 44374 }, { "epoch": 0.86, "learning_rate": 3.565009801447897e-05, "loss": 0.0, "step": 44376 }, { "epoch": 0.86, "learning_rate": 3.564945105422174e-05, "loss": 0.0021, "step": 44378 }, { "epoch": 0.86, "learning_rate": 3.564880409396451e-05, "loss": 0.0001, "step": 44380 }, { "epoch": 0.86, "learning_rate": 3.564815713370728e-05, "loss": 0.0, "step": 44382 }, { "epoch": 0.86, "learning_rate": 3.564751017345005e-05, "loss": 0.0, "step": 44384 }, { "epoch": 0.86, "learning_rate": 3.564686321319282e-05, "loss": 0.0001, "step": 44386 }, { "epoch": 0.86, "learning_rate": 3.564621625293558e-05, "loss": 0.0002, "step": 44388 }, { "epoch": 0.86, "learning_rate": 3.5645569292678355e-05, "loss": 0.0, "step": 44390 }, { "epoch": 0.86, "learning_rate": 3.5644922332421124e-05, "loss": 0.0, "step": 44392 }, { "epoch": 0.86, "learning_rate": 3.564427537216389e-05, "loss": 0.0091, "step": 44394 }, { "epoch": 0.86, "learning_rate": 3.564362841190666e-05, "loss": 0.0, "step": 44396 }, { "epoch": 0.86, "learning_rate": 3.5642981451649425e-05, "loss": 0.0102, "step": 44398 }, { "epoch": 0.86, "learning_rate": 3.5642334491392194e-05, "loss": 0.0, "step": 44400 }, { "epoch": 0.86, "learning_rate": 3.564168753113496e-05, "loss": 0.0, "step": 44402 }, { "epoch": 0.86, "learning_rate": 3.564104057087773e-05, "loss": 0.0206, "step": 44404 }, { "epoch": 0.86, "learning_rate": 3.56403936106205e-05, "loss": 0.0, "step": 44406 }, { "epoch": 0.86, "learning_rate": 3.563974665036327e-05, "loss": 0.0099, "step": 44408 }, { "epoch": 0.86, "learning_rate": 3.563909969010604e-05, "loss": 0.0006, "step": 44410 }, { "epoch": 0.86, "learning_rate": 3.563845272984881e-05, "loss": 0.0072, "step": 44412 }, { "epoch": 0.86, "learning_rate": 3.563780576959158e-05, "loss": 0.0, "step": 44414 }, { "epoch": 0.86, "learning_rate": 3.563715880933434e-05, "loss": 0.0, "step": 44416 }, { "epoch": 0.86, "learning_rate": 3.5636511849077116e-05, "loss": 0.0, "step": 44418 }, { "epoch": 0.86, "learning_rate": 3.563586488881988e-05, "loss": 0.0001, "step": 44420 }, { "epoch": 0.86, "learning_rate": 3.563521792856265e-05, "loss": 0.0002, "step": 44422 }, { "epoch": 0.86, "learning_rate": 3.563457096830542e-05, "loss": 0.0003, "step": 44424 }, { "epoch": 0.86, "learning_rate": 3.5633924008048185e-05, "loss": 0.0049, "step": 44426 }, { "epoch": 0.86, "learning_rate": 3.563327704779096e-05, "loss": 0.0, "step": 44428 }, { "epoch": 0.86, "learning_rate": 3.5632630087533723e-05, "loss": 0.022, "step": 44430 }, { "epoch": 0.86, "learning_rate": 3.563198312727649e-05, "loss": 0.0435, "step": 44432 }, { "epoch": 0.86, "learning_rate": 3.563133616701926e-05, "loss": 0.0001, "step": 44434 }, { "epoch": 0.86, "learning_rate": 3.563068920676203e-05, "loss": 0.0, "step": 44436 }, { "epoch": 0.86, "learning_rate": 3.56300422465048e-05, "loss": 0.0, "step": 44438 }, { "epoch": 0.86, "learning_rate": 3.562939528624757e-05, "loss": 0.0001, "step": 44440 }, { "epoch": 0.86, "learning_rate": 3.562874832599034e-05, "loss": 0.0001, "step": 44442 }, { "epoch": 0.86, "learning_rate": 3.56281013657331e-05, "loss": 0.0, "step": 44444 }, { "epoch": 0.86, "learning_rate": 3.5627454405475876e-05, "loss": 0.0002, "step": 44446 }, { "epoch": 0.86, "learning_rate": 3.562680744521864e-05, "loss": 0.0, "step": 44448 }, { "epoch": 0.86, "learning_rate": 3.5626160484961414e-05, "loss": 0.0003, "step": 44450 }, { "epoch": 0.86, "learning_rate": 3.562551352470418e-05, "loss": 0.0, "step": 44452 }, { "epoch": 0.86, "learning_rate": 3.5624866564446946e-05, "loss": 0.0, "step": 44454 }, { "epoch": 0.86, "learning_rate": 3.5624219604189715e-05, "loss": 0.0, "step": 44456 }, { "epoch": 0.86, "learning_rate": 3.5623572643932484e-05, "loss": 0.0, "step": 44458 }, { "epoch": 0.86, "learning_rate": 3.562292568367525e-05, "loss": 0.0, "step": 44460 }, { "epoch": 0.86, "learning_rate": 3.562227872341802e-05, "loss": 0.0001, "step": 44462 }, { "epoch": 0.86, "learning_rate": 3.562163176316079e-05, "loss": 0.0311, "step": 44464 }, { "epoch": 0.86, "learning_rate": 3.562098480290356e-05, "loss": 0.0, "step": 44466 }, { "epoch": 0.86, "learning_rate": 3.562033784264633e-05, "loss": 0.0, "step": 44468 }, { "epoch": 0.86, "learning_rate": 3.56196908823891e-05, "loss": 0.0, "step": 44470 }, { "epoch": 0.86, "learning_rate": 3.561904392213187e-05, "loss": 0.0, "step": 44472 }, { "epoch": 0.86, "learning_rate": 3.5618396961874637e-05, "loss": 0.0, "step": 44474 }, { "epoch": 0.86, "learning_rate": 3.56177500016174e-05, "loss": 0.003, "step": 44476 }, { "epoch": 0.86, "learning_rate": 3.5617103041360175e-05, "loss": 0.0007, "step": 44478 }, { "epoch": 0.86, "learning_rate": 3.561645608110294e-05, "loss": 0.0003, "step": 44480 }, { "epoch": 0.86, "learning_rate": 3.5615809120845706e-05, "loss": 0.0001, "step": 44482 }, { "epoch": 0.86, "learning_rate": 3.5615162160588475e-05, "loss": 0.0, "step": 44484 }, { "epoch": 0.86, "learning_rate": 3.5614515200331244e-05, "loss": 0.0, "step": 44486 }, { "epoch": 0.86, "learning_rate": 3.561386824007401e-05, "loss": 0.0, "step": 44488 }, { "epoch": 0.86, "learning_rate": 3.561322127981678e-05, "loss": 0.0, "step": 44490 }, { "epoch": 0.86, "learning_rate": 3.561257431955955e-05, "loss": 0.0019, "step": 44492 }, { "epoch": 0.86, "learning_rate": 3.561192735930232e-05, "loss": 0.0, "step": 44494 }, { "epoch": 0.86, "learning_rate": 3.561128039904509e-05, "loss": 0.0008, "step": 44496 }, { "epoch": 0.86, "learning_rate": 3.561063343878785e-05, "loss": 0.0014, "step": 44498 }, { "epoch": 0.86, "learning_rate": 3.560998647853063e-05, "loss": 0.0027, "step": 44500 }, { "epoch": 0.86, "learning_rate": 3.560933951827339e-05, "loss": 0.0, "step": 44502 }, { "epoch": 0.86, "learning_rate": 3.560869255801616e-05, "loss": 0.0014, "step": 44504 }, { "epoch": 0.86, "learning_rate": 3.5608045597758935e-05, "loss": 0.0, "step": 44506 }, { "epoch": 0.86, "learning_rate": 3.56073986375017e-05, "loss": 0.0037, "step": 44508 }, { "epoch": 0.86, "learning_rate": 3.560675167724447e-05, "loss": 0.0, "step": 44510 }, { "epoch": 0.86, "learning_rate": 3.5606104716987236e-05, "loss": 0.0, "step": 44512 }, { "epoch": 0.86, "learning_rate": 3.5605457756730005e-05, "loss": 0.0098, "step": 44514 }, { "epoch": 0.86, "learning_rate": 3.5604810796472774e-05, "loss": 0.0, "step": 44516 }, { "epoch": 0.86, "learning_rate": 3.560416383621554e-05, "loss": 0.0, "step": 44518 }, { "epoch": 0.86, "learning_rate": 3.560351687595831e-05, "loss": 0.0001, "step": 44520 }, { "epoch": 0.86, "learning_rate": 3.560286991570108e-05, "loss": 0.0001, "step": 44522 }, { "epoch": 0.86, "learning_rate": 3.560222295544385e-05, "loss": 0.0, "step": 44524 }, { "epoch": 0.86, "learning_rate": 3.560157599518662e-05, "loss": 0.0, "step": 44526 }, { "epoch": 0.86, "learning_rate": 3.560092903492939e-05, "loss": 0.0001, "step": 44528 }, { "epoch": 0.86, "learning_rate": 3.560028207467215e-05, "loss": 0.0, "step": 44530 }, { "epoch": 0.86, "learning_rate": 3.5599635114414926e-05, "loss": 0.0005, "step": 44532 }, { "epoch": 0.86, "learning_rate": 3.559898815415769e-05, "loss": 0.0, "step": 44534 }, { "epoch": 0.86, "learning_rate": 3.559834119390046e-05, "loss": 0.0009, "step": 44536 }, { "epoch": 0.86, "learning_rate": 3.5597694233643234e-05, "loss": 0.0002, "step": 44538 }, { "epoch": 0.86, "learning_rate": 3.5597047273385996e-05, "loss": 0.0048, "step": 44540 }, { "epoch": 0.86, "learning_rate": 3.5596400313128765e-05, "loss": 0.0001, "step": 44542 }, { "epoch": 0.86, "learning_rate": 3.5595753352871534e-05, "loss": 0.0, "step": 44544 }, { "epoch": 0.86, "learning_rate": 3.55951063926143e-05, "loss": 0.0002, "step": 44546 }, { "epoch": 0.86, "learning_rate": 3.559445943235707e-05, "loss": 0.0001, "step": 44548 }, { "epoch": 0.86, "learning_rate": 3.559381247209984e-05, "loss": 0.0, "step": 44550 }, { "epoch": 0.86, "learning_rate": 3.559316551184261e-05, "loss": 0.0001, "step": 44552 }, { "epoch": 0.86, "learning_rate": 3.559251855158538e-05, "loss": 0.0, "step": 44554 }, { "epoch": 0.86, "learning_rate": 3.559187159132815e-05, "loss": 0.0024, "step": 44556 }, { "epoch": 0.86, "learning_rate": 3.559122463107091e-05, "loss": 0.0001, "step": 44558 }, { "epoch": 0.86, "learning_rate": 3.559057767081369e-05, "loss": 0.0, "step": 44560 }, { "epoch": 0.86, "learning_rate": 3.558993071055645e-05, "loss": 0.0, "step": 44562 }, { "epoch": 0.86, "learning_rate": 3.558928375029922e-05, "loss": 0.0144, "step": 44564 }, { "epoch": 0.86, "learning_rate": 3.558863679004199e-05, "loss": 0.0, "step": 44566 }, { "epoch": 0.87, "learning_rate": 3.5587989829784756e-05, "loss": 0.0001, "step": 44568 }, { "epoch": 0.87, "learning_rate": 3.558734286952753e-05, "loss": 0.009, "step": 44570 }, { "epoch": 0.87, "learning_rate": 3.5586695909270295e-05, "loss": 0.0, "step": 44572 }, { "epoch": 0.87, "learning_rate": 3.5586048949013064e-05, "loss": 0.0, "step": 44574 }, { "epoch": 0.87, "learning_rate": 3.558540198875583e-05, "loss": 0.0082, "step": 44576 }, { "epoch": 0.87, "learning_rate": 3.55847550284986e-05, "loss": 0.0, "step": 44578 }, { "epoch": 0.87, "learning_rate": 3.5584108068241364e-05, "loss": 0.0, "step": 44580 }, { "epoch": 0.87, "learning_rate": 3.558346110798414e-05, "loss": 0.0002, "step": 44582 }, { "epoch": 0.87, "learning_rate": 3.558281414772691e-05, "loss": 0.0146, "step": 44584 }, { "epoch": 0.87, "learning_rate": 3.558216718746967e-05, "loss": 0.0004, "step": 44586 }, { "epoch": 0.87, "learning_rate": 3.558152022721245e-05, "loss": 0.0, "step": 44588 }, { "epoch": 0.87, "learning_rate": 3.558087326695521e-05, "loss": 0.0, "step": 44590 }, { "epoch": 0.87, "learning_rate": 3.5580226306697985e-05, "loss": 0.0, "step": 44592 }, { "epoch": 0.87, "learning_rate": 3.557957934644075e-05, "loss": 0.0001, "step": 44594 }, { "epoch": 0.87, "learning_rate": 3.557893238618352e-05, "loss": 0.0325, "step": 44596 }, { "epoch": 0.87, "learning_rate": 3.5578285425926286e-05, "loss": 0.0, "step": 44598 }, { "epoch": 0.87, "learning_rate": 3.5577638465669055e-05, "loss": 0.0, "step": 44600 }, { "epoch": 0.87, "learning_rate": 3.5576991505411824e-05, "loss": 0.0, "step": 44602 }, { "epoch": 0.87, "learning_rate": 3.557634454515459e-05, "loss": 0.0001, "step": 44604 }, { "epoch": 0.87, "learning_rate": 3.557569758489736e-05, "loss": 0.0, "step": 44606 }, { "epoch": 0.87, "learning_rate": 3.557505062464013e-05, "loss": 0.0039, "step": 44608 }, { "epoch": 0.87, "learning_rate": 3.55744036643829e-05, "loss": 0.0, "step": 44610 }, { "epoch": 0.87, "learning_rate": 3.557375670412566e-05, "loss": 0.0039, "step": 44612 }, { "epoch": 0.87, "learning_rate": 3.557310974386844e-05, "loss": 0.0001, "step": 44614 }, { "epoch": 0.87, "learning_rate": 3.557246278361121e-05, "loss": 0.0114, "step": 44616 }, { "epoch": 0.87, "learning_rate": 3.557181582335397e-05, "loss": 0.0001, "step": 44618 }, { "epoch": 0.87, "learning_rate": 3.5571168863096746e-05, "loss": 0.0015, "step": 44620 }, { "epoch": 0.87, "learning_rate": 3.557052190283951e-05, "loss": 0.0001, "step": 44622 }, { "epoch": 0.87, "learning_rate": 3.556987494258228e-05, "loss": 0.0081, "step": 44624 }, { "epoch": 0.87, "learning_rate": 3.5569227982325046e-05, "loss": 0.0001, "step": 44626 }, { "epoch": 0.87, "learning_rate": 3.5568581022067815e-05, "loss": 0.0, "step": 44628 }, { "epoch": 0.87, "learning_rate": 3.5567934061810584e-05, "loss": 0.0001, "step": 44630 }, { "epoch": 0.87, "learning_rate": 3.5567287101553354e-05, "loss": 0.0001, "step": 44632 }, { "epoch": 0.87, "learning_rate": 3.556664014129612e-05, "loss": 0.0, "step": 44634 }, { "epoch": 0.87, "learning_rate": 3.556599318103889e-05, "loss": 0.0, "step": 44636 }, { "epoch": 0.87, "learning_rate": 3.556534622078166e-05, "loss": 0.0, "step": 44638 }, { "epoch": 0.87, "learning_rate": 3.556469926052442e-05, "loss": 0.0, "step": 44640 }, { "epoch": 0.87, "learning_rate": 3.55640523002672e-05, "loss": 0.0, "step": 44642 }, { "epoch": 0.87, "learning_rate": 3.556340534000996e-05, "loss": 0.0001, "step": 44644 }, { "epoch": 0.87, "learning_rate": 3.556275837975273e-05, "loss": 0.0001, "step": 44646 }, { "epoch": 0.87, "learning_rate": 3.5562111419495506e-05, "loss": 0.0, "step": 44648 }, { "epoch": 0.87, "learning_rate": 3.556146445923827e-05, "loss": 0.0, "step": 44650 }, { "epoch": 0.87, "learning_rate": 3.5560817498981044e-05, "loss": 0.0, "step": 44652 }, { "epoch": 0.87, "learning_rate": 3.556017053872381e-05, "loss": 0.0, "step": 44654 }, { "epoch": 0.87, "learning_rate": 3.5559523578466576e-05, "loss": 0.0, "step": 44656 }, { "epoch": 0.87, "learning_rate": 3.5558876618209345e-05, "loss": 0.0004, "step": 44658 }, { "epoch": 0.87, "learning_rate": 3.5558229657952114e-05, "loss": 0.0, "step": 44660 }, { "epoch": 0.87, "learning_rate": 3.555758269769488e-05, "loss": 0.0001, "step": 44662 }, { "epoch": 0.87, "learning_rate": 3.555693573743765e-05, "loss": 0.0, "step": 44664 }, { "epoch": 0.87, "learning_rate": 3.555628877718042e-05, "loss": 0.0, "step": 44666 }, { "epoch": 0.87, "learning_rate": 3.555564181692319e-05, "loss": 0.0, "step": 44668 }, { "epoch": 0.87, "learning_rate": 3.555499485666596e-05, "loss": 0.001, "step": 44670 }, { "epoch": 0.87, "learning_rate": 3.555434789640872e-05, "loss": 0.0003, "step": 44672 }, { "epoch": 0.87, "learning_rate": 3.55537009361515e-05, "loss": 0.0065, "step": 44674 }, { "epoch": 0.87, "learning_rate": 3.555305397589426e-05, "loss": 0.0, "step": 44676 }, { "epoch": 0.87, "learning_rate": 3.555240701563703e-05, "loss": 0.0, "step": 44678 }, { "epoch": 0.87, "learning_rate": 3.55517600553798e-05, "loss": 0.0012, "step": 44680 }, { "epoch": 0.87, "learning_rate": 3.555111309512257e-05, "loss": 0.0, "step": 44682 }, { "epoch": 0.87, "learning_rate": 3.5550466134865336e-05, "loss": 0.0031, "step": 44684 }, { "epoch": 0.87, "learning_rate": 3.5549819174608105e-05, "loss": 0.0, "step": 44686 }, { "epoch": 0.87, "learning_rate": 3.5549172214350874e-05, "loss": 0.0, "step": 44688 }, { "epoch": 0.87, "learning_rate": 3.5548525254093643e-05, "loss": 0.0001, "step": 44690 }, { "epoch": 0.87, "learning_rate": 3.554787829383641e-05, "loss": 0.0001, "step": 44692 }, { "epoch": 0.87, "learning_rate": 3.554723133357918e-05, "loss": 0.0081, "step": 44694 }, { "epoch": 0.87, "learning_rate": 3.554658437332195e-05, "loss": 0.0, "step": 44696 }, { "epoch": 0.87, "learning_rate": 3.554593741306472e-05, "loss": 0.0, "step": 44698 }, { "epoch": 0.87, "learning_rate": 3.554529045280748e-05, "loss": 0.0172, "step": 44700 }, { "epoch": 0.87, "learning_rate": 3.554464349255026e-05, "loss": 0.0001, "step": 44702 }, { "epoch": 0.87, "learning_rate": 3.554399653229302e-05, "loss": 0.0001, "step": 44704 }, { "epoch": 0.87, "learning_rate": 3.554334957203579e-05, "loss": 0.0, "step": 44706 }, { "epoch": 0.87, "learning_rate": 3.554270261177856e-05, "loss": 0.0, "step": 44708 }, { "epoch": 0.87, "learning_rate": 3.554205565152133e-05, "loss": 0.0002, "step": 44710 }, { "epoch": 0.87, "learning_rate": 3.5541408691264097e-05, "loss": 0.0, "step": 44712 }, { "epoch": 0.87, "learning_rate": 3.5540761731006866e-05, "loss": 0.0, "step": 44714 }, { "epoch": 0.87, "learning_rate": 3.5540114770749635e-05, "loss": 0.0001, "step": 44716 }, { "epoch": 0.87, "learning_rate": 3.5539467810492404e-05, "loss": 0.0001, "step": 44718 }, { "epoch": 0.87, "learning_rate": 3.553882085023517e-05, "loss": 0.0054, "step": 44720 }, { "epoch": 0.87, "learning_rate": 3.5538173889977935e-05, "loss": 0.0, "step": 44722 }, { "epoch": 0.87, "learning_rate": 3.553752692972071e-05, "loss": 0.0065, "step": 44724 }, { "epoch": 0.87, "learning_rate": 3.553687996946348e-05, "loss": 0.0, "step": 44726 }, { "epoch": 0.87, "learning_rate": 3.553623300920624e-05, "loss": 0.0, "step": 44728 }, { "epoch": 0.87, "learning_rate": 3.553558604894902e-05, "loss": 0.0, "step": 44730 }, { "epoch": 0.87, "learning_rate": 3.553493908869178e-05, "loss": 0.0107, "step": 44732 }, { "epoch": 0.87, "learning_rate": 3.5534292128434557e-05, "loss": 0.0001, "step": 44734 }, { "epoch": 0.87, "learning_rate": 3.553364516817732e-05, "loss": 0.0, "step": 44736 }, { "epoch": 0.87, "learning_rate": 3.553299820792009e-05, "loss": 0.0004, "step": 44738 }, { "epoch": 0.87, "learning_rate": 3.553235124766286e-05, "loss": 0.0, "step": 44740 }, { "epoch": 0.87, "learning_rate": 3.5531704287405626e-05, "loss": 0.0, "step": 44742 }, { "epoch": 0.87, "learning_rate": 3.5531057327148395e-05, "loss": 0.0001, "step": 44744 }, { "epoch": 0.87, "learning_rate": 3.5530410366891164e-05, "loss": 0.0, "step": 44746 }, { "epoch": 0.87, "learning_rate": 3.552976340663393e-05, "loss": 0.0004, "step": 44748 }, { "epoch": 0.87, "learning_rate": 3.55291164463767e-05, "loss": 0.0, "step": 44750 }, { "epoch": 0.87, "learning_rate": 3.552846948611947e-05, "loss": 0.0, "step": 44752 }, { "epoch": 0.87, "learning_rate": 3.5527822525862234e-05, "loss": 0.0, "step": 44754 }, { "epoch": 0.87, "learning_rate": 3.552717556560501e-05, "loss": 0.0, "step": 44756 }, { "epoch": 0.87, "learning_rate": 3.552652860534777e-05, "loss": 0.0034, "step": 44758 }, { "epoch": 0.87, "learning_rate": 3.552588164509054e-05, "loss": 0.0002, "step": 44760 }, { "epoch": 0.87, "learning_rate": 3.552523468483332e-05, "loss": 0.0011, "step": 44762 }, { "epoch": 0.87, "learning_rate": 3.552458772457608e-05, "loss": 0.0, "step": 44764 }, { "epoch": 0.87, "learning_rate": 3.552394076431885e-05, "loss": 0.0002, "step": 44766 }, { "epoch": 0.87, "learning_rate": 3.552329380406162e-05, "loss": 0.0005, "step": 44768 }, { "epoch": 0.87, "learning_rate": 3.5522646843804387e-05, "loss": 0.0, "step": 44770 }, { "epoch": 0.87, "learning_rate": 3.5521999883547156e-05, "loss": 0.0037, "step": 44772 }, { "epoch": 0.87, "learning_rate": 3.5521352923289925e-05, "loss": 0.0, "step": 44774 }, { "epoch": 0.87, "learning_rate": 3.5520705963032694e-05, "loss": 0.0, "step": 44776 }, { "epoch": 0.87, "learning_rate": 3.552005900277546e-05, "loss": 0.001, "step": 44778 }, { "epoch": 0.87, "learning_rate": 3.551941204251823e-05, "loss": 0.0, "step": 44780 }, { "epoch": 0.87, "learning_rate": 3.5518765082260994e-05, "loss": 0.0003, "step": 44782 }, { "epoch": 0.87, "learning_rate": 3.551811812200377e-05, "loss": 0.0, "step": 44784 }, { "epoch": 0.87, "learning_rate": 3.551747116174653e-05, "loss": 0.0, "step": 44786 }, { "epoch": 0.87, "learning_rate": 3.55168242014893e-05, "loss": 0.0035, "step": 44788 }, { "epoch": 0.87, "learning_rate": 3.551617724123207e-05, "loss": 0.0, "step": 44790 }, { "epoch": 0.87, "learning_rate": 3.551553028097484e-05, "loss": 0.0, "step": 44792 }, { "epoch": 0.87, "learning_rate": 3.5514883320717616e-05, "loss": 0.0, "step": 44794 }, { "epoch": 0.87, "learning_rate": 3.551423636046038e-05, "loss": 0.0032, "step": 44796 }, { "epoch": 0.87, "learning_rate": 3.551358940020315e-05, "loss": 0.0, "step": 44798 }, { "epoch": 0.87, "learning_rate": 3.5512942439945916e-05, "loss": 0.0043, "step": 44800 }, { "epoch": 0.87, "learning_rate": 3.5512295479688685e-05, "loss": 0.0017, "step": 44802 }, { "epoch": 0.87, "learning_rate": 3.551164851943145e-05, "loss": 0.0, "step": 44804 }, { "epoch": 0.87, "learning_rate": 3.551100155917422e-05, "loss": 0.0, "step": 44806 }, { "epoch": 0.87, "learning_rate": 3.551035459891699e-05, "loss": 0.0, "step": 44808 }, { "epoch": 0.87, "learning_rate": 3.550970763865976e-05, "loss": 0.0047, "step": 44810 }, { "epoch": 0.87, "learning_rate": 3.550906067840253e-05, "loss": 0.0002, "step": 44812 }, { "epoch": 0.87, "learning_rate": 3.550841371814529e-05, "loss": 0.001, "step": 44814 }, { "epoch": 0.87, "learning_rate": 3.550776675788807e-05, "loss": 0.0, "step": 44816 }, { "epoch": 0.87, "learning_rate": 3.550711979763083e-05, "loss": 0.0025, "step": 44818 }, { "epoch": 0.87, "learning_rate": 3.55064728373736e-05, "loss": 0.0, "step": 44820 }, { "epoch": 0.87, "learning_rate": 3.550582587711637e-05, "loss": 0.0, "step": 44822 }, { "epoch": 0.87, "learning_rate": 3.550517891685914e-05, "loss": 0.0, "step": 44824 }, { "epoch": 0.87, "learning_rate": 3.550453195660191e-05, "loss": 0.0001, "step": 44826 }, { "epoch": 0.87, "learning_rate": 3.5503884996344676e-05, "loss": 0.0, "step": 44828 }, { "epoch": 0.87, "learning_rate": 3.5503238036087445e-05, "loss": 0.0, "step": 44830 }, { "epoch": 0.87, "learning_rate": 3.5502591075830215e-05, "loss": 0.0, "step": 44832 }, { "epoch": 0.87, "learning_rate": 3.5501944115572984e-05, "loss": 0.0, "step": 44834 }, { "epoch": 0.87, "learning_rate": 3.5501297155315746e-05, "loss": 0.0, "step": 44836 }, { "epoch": 0.87, "learning_rate": 3.550065019505852e-05, "loss": 0.0002, "step": 44838 }, { "epoch": 0.87, "learning_rate": 3.550000323480129e-05, "loss": 0.0, "step": 44840 }, { "epoch": 0.87, "learning_rate": 3.549935627454405e-05, "loss": 0.0, "step": 44842 }, { "epoch": 0.87, "learning_rate": 3.549870931428683e-05, "loss": 0.0, "step": 44844 }, { "epoch": 0.87, "learning_rate": 3.549806235402959e-05, "loss": 0.0001, "step": 44846 }, { "epoch": 0.87, "learning_rate": 3.549741539377236e-05, "loss": 0.0002, "step": 44848 }, { "epoch": 0.87, "learning_rate": 3.549676843351513e-05, "loss": 0.0, "step": 44850 }, { "epoch": 0.87, "learning_rate": 3.54961214732579e-05, "loss": 0.0, "step": 44852 }, { "epoch": 0.87, "learning_rate": 3.549547451300067e-05, "loss": 0.021, "step": 44854 }, { "epoch": 0.87, "learning_rate": 3.549482755274344e-05, "loss": 0.0, "step": 44856 }, { "epoch": 0.87, "learning_rate": 3.5494180592486206e-05, "loss": 0.0, "step": 44858 }, { "epoch": 0.87, "learning_rate": 3.5493533632228975e-05, "loss": 0.0003, "step": 44860 }, { "epoch": 0.87, "learning_rate": 3.5492886671971744e-05, "loss": 0.0, "step": 44862 }, { "epoch": 0.87, "learning_rate": 3.5492239711714506e-05, "loss": 0.0001, "step": 44864 }, { "epoch": 0.87, "learning_rate": 3.549159275145728e-05, "loss": 0.0, "step": 44866 }, { "epoch": 0.87, "learning_rate": 3.5490945791200045e-05, "loss": 0.0002, "step": 44868 }, { "epoch": 0.87, "learning_rate": 3.5490298830942814e-05, "loss": 0.0, "step": 44870 }, { "epoch": 0.87, "learning_rate": 3.548965187068559e-05, "loss": 0.0, "step": 44872 }, { "epoch": 0.87, "learning_rate": 3.548900491042835e-05, "loss": 0.0001, "step": 44874 }, { "epoch": 0.87, "learning_rate": 3.548835795017113e-05, "loss": 0.002, "step": 44876 }, { "epoch": 0.87, "learning_rate": 3.548771098991389e-05, "loss": 0.0, "step": 44878 }, { "epoch": 0.87, "learning_rate": 3.548706402965666e-05, "loss": 0.0003, "step": 44880 }, { "epoch": 0.87, "learning_rate": 3.548641706939943e-05, "loss": 0.0187, "step": 44882 }, { "epoch": 0.87, "learning_rate": 3.54857701091422e-05, "loss": 0.0007, "step": 44884 }, { "epoch": 0.87, "learning_rate": 3.5485123148884966e-05, "loss": 0.0, "step": 44886 }, { "epoch": 0.87, "learning_rate": 3.5484476188627735e-05, "loss": 0.0027, "step": 44888 }, { "epoch": 0.87, "learning_rate": 3.5483829228370504e-05, "loss": 0.0, "step": 44890 }, { "epoch": 0.87, "learning_rate": 3.5483182268113274e-05, "loss": 0.0, "step": 44892 }, { "epoch": 0.87, "learning_rate": 3.548253530785604e-05, "loss": 0.0, "step": 44894 }, { "epoch": 0.87, "learning_rate": 3.5481888347598805e-05, "loss": 0.0, "step": 44896 }, { "epoch": 0.87, "learning_rate": 3.548124138734158e-05, "loss": 0.0, "step": 44898 }, { "epoch": 0.87, "learning_rate": 3.548059442708434e-05, "loss": 0.0002, "step": 44900 }, { "epoch": 0.87, "learning_rate": 3.547994746682711e-05, "loss": 0.0014, "step": 44902 }, { "epoch": 0.87, "learning_rate": 3.547930050656989e-05, "loss": 0.0, "step": 44904 }, { "epoch": 0.87, "learning_rate": 3.547865354631265e-05, "loss": 0.0011, "step": 44906 }, { "epoch": 0.87, "learning_rate": 3.547800658605542e-05, "loss": 0.0, "step": 44908 }, { "epoch": 0.87, "learning_rate": 3.547735962579819e-05, "loss": 0.0004, "step": 44910 }, { "epoch": 0.87, "learning_rate": 3.547671266554096e-05, "loss": 0.0, "step": 44912 }, { "epoch": 0.87, "learning_rate": 3.547606570528373e-05, "loss": 0.0, "step": 44914 }, { "epoch": 0.87, "learning_rate": 3.5475418745026496e-05, "loss": 0.0, "step": 44916 }, { "epoch": 0.87, "learning_rate": 3.5474771784769265e-05, "loss": 0.0034, "step": 44918 }, { "epoch": 0.87, "learning_rate": 3.5474124824512034e-05, "loss": 0.0, "step": 44920 }, { "epoch": 0.87, "learning_rate": 3.54734778642548e-05, "loss": 0.0023, "step": 44922 }, { "epoch": 0.87, "learning_rate": 3.5472830903997565e-05, "loss": 0.0, "step": 44924 }, { "epoch": 0.87, "learning_rate": 3.547218394374034e-05, "loss": 0.0, "step": 44926 }, { "epoch": 0.87, "learning_rate": 3.5471536983483104e-05, "loss": 0.0003, "step": 44928 }, { "epoch": 0.87, "learning_rate": 3.547089002322587e-05, "loss": 0.0002, "step": 44930 }, { "epoch": 0.87, "learning_rate": 3.547024306296864e-05, "loss": 0.0107, "step": 44932 }, { "epoch": 0.87, "learning_rate": 3.546959610271141e-05, "loss": 0.0022, "step": 44934 }, { "epoch": 0.87, "learning_rate": 3.546894914245418e-05, "loss": 0.0, "step": 44936 }, { "epoch": 0.87, "learning_rate": 3.546830218219695e-05, "loss": 0.0, "step": 44938 }, { "epoch": 0.87, "learning_rate": 3.546765522193972e-05, "loss": 0.0036, "step": 44940 }, { "epoch": 0.87, "learning_rate": 3.546700826168249e-05, "loss": 0.0001, "step": 44942 }, { "epoch": 0.87, "learning_rate": 3.5466361301425256e-05, "loss": 0.0, "step": 44944 }, { "epoch": 0.87, "learning_rate": 3.546571434116802e-05, "loss": 0.0, "step": 44946 }, { "epoch": 0.87, "learning_rate": 3.5465067380910794e-05, "loss": 0.0035, "step": 44948 }, { "epoch": 0.87, "learning_rate": 3.5464420420653563e-05, "loss": 0.0, "step": 44950 }, { "epoch": 0.87, "learning_rate": 3.546377346039633e-05, "loss": 0.0005, "step": 44952 }, { "epoch": 0.87, "learning_rate": 3.54631265001391e-05, "loss": 0.0001, "step": 44954 }, { "epoch": 0.87, "learning_rate": 3.5462479539881864e-05, "loss": 0.0029, "step": 44956 }, { "epoch": 0.87, "learning_rate": 3.546183257962464e-05, "loss": 0.0, "step": 44958 }, { "epoch": 0.87, "learning_rate": 3.54611856193674e-05, "loss": 0.0001, "step": 44960 }, { "epoch": 0.87, "learning_rate": 3.546053865911017e-05, "loss": 0.0, "step": 44962 }, { "epoch": 0.87, "learning_rate": 3.545989169885294e-05, "loss": 0.0, "step": 44964 }, { "epoch": 0.87, "learning_rate": 3.545924473859571e-05, "loss": 0.0, "step": 44966 }, { "epoch": 0.87, "learning_rate": 3.545859777833848e-05, "loss": 0.0, "step": 44968 }, { "epoch": 0.87, "learning_rate": 3.545795081808125e-05, "loss": 0.0, "step": 44970 }, { "epoch": 0.87, "learning_rate": 3.5457303857824017e-05, "loss": 0.0041, "step": 44972 }, { "epoch": 0.87, "learning_rate": 3.5456656897566786e-05, "loss": 0.0002, "step": 44974 }, { "epoch": 0.87, "learning_rate": 3.5456009937309555e-05, "loss": 0.0, "step": 44976 }, { "epoch": 0.87, "learning_rate": 3.545536297705232e-05, "loss": 0.0042, "step": 44978 }, { "epoch": 0.87, "learning_rate": 3.545471601679509e-05, "loss": 0.0, "step": 44980 }, { "epoch": 0.87, "learning_rate": 3.5454069056537855e-05, "loss": 0.0007, "step": 44982 }, { "epoch": 0.87, "learning_rate": 3.5453422096280624e-05, "loss": 0.0, "step": 44984 }, { "epoch": 0.87, "learning_rate": 3.54527751360234e-05, "loss": 0.0, "step": 44986 }, { "epoch": 0.87, "learning_rate": 3.545212817576616e-05, "loss": 0.0003, "step": 44988 }, { "epoch": 0.87, "learning_rate": 3.545148121550893e-05, "loss": 0.0, "step": 44990 }, { "epoch": 0.87, "learning_rate": 3.54508342552517e-05, "loss": 0.0, "step": 44992 }, { "epoch": 0.87, "learning_rate": 3.545018729499447e-05, "loss": 0.0, "step": 44994 }, { "epoch": 0.87, "learning_rate": 3.544954033473724e-05, "loss": 0.0001, "step": 44996 }, { "epoch": 0.87, "learning_rate": 3.544889337448001e-05, "loss": 0.0024, "step": 44998 }, { "epoch": 0.87, "learning_rate": 3.544824641422278e-05, "loss": 0.0, "step": 45000 }, { "epoch": 0.87, "learning_rate": 3.5447599453965546e-05, "loss": 0.0, "step": 45002 }, { "epoch": 0.87, "learning_rate": 3.5446952493708315e-05, "loss": 0.0002, "step": 45004 }, { "epoch": 0.87, "learning_rate": 3.544630553345108e-05, "loss": 0.001, "step": 45006 }, { "epoch": 0.87, "learning_rate": 3.544565857319385e-05, "loss": 0.0, "step": 45008 }, { "epoch": 0.87, "learning_rate": 3.5445011612936616e-05, "loss": 0.0, "step": 45010 }, { "epoch": 0.87, "learning_rate": 3.5444364652679385e-05, "loss": 0.0033, "step": 45012 }, { "epoch": 0.87, "learning_rate": 3.5443717692422154e-05, "loss": 0.0405, "step": 45014 }, { "epoch": 0.87, "learning_rate": 3.544307073216492e-05, "loss": 0.0014, "step": 45016 }, { "epoch": 0.87, "learning_rate": 3.54424237719077e-05, "loss": 0.0, "step": 45018 }, { "epoch": 0.87, "learning_rate": 3.544177681165046e-05, "loss": 0.0, "step": 45020 }, { "epoch": 0.87, "learning_rate": 3.544112985139323e-05, "loss": 0.0084, "step": 45022 }, { "epoch": 0.87, "learning_rate": 3.5440482891136e-05, "loss": 0.0052, "step": 45024 }, { "epoch": 0.87, "learning_rate": 3.543983593087877e-05, "loss": 0.0, "step": 45026 }, { "epoch": 0.87, "learning_rate": 3.543918897062154e-05, "loss": 0.0, "step": 45028 }, { "epoch": 0.87, "learning_rate": 3.5438542010364306e-05, "loss": 0.0, "step": 45030 }, { "epoch": 0.87, "learning_rate": 3.5437895050107076e-05, "loss": 0.0003, "step": 45032 }, { "epoch": 0.87, "learning_rate": 3.5437248089849845e-05, "loss": 0.0, "step": 45034 }, { "epoch": 0.87, "learning_rate": 3.5436601129592614e-05, "loss": 0.0001, "step": 45036 }, { "epoch": 0.87, "learning_rate": 3.5435954169335376e-05, "loss": 0.0, "step": 45038 }, { "epoch": 0.87, "learning_rate": 3.543530720907815e-05, "loss": 0.0011, "step": 45040 }, { "epoch": 0.87, "learning_rate": 3.5434660248820914e-05, "loss": 0.0, "step": 45042 }, { "epoch": 0.87, "learning_rate": 3.543401328856368e-05, "loss": 0.0003, "step": 45044 }, { "epoch": 0.87, "learning_rate": 3.543336632830645e-05, "loss": 0.0029, "step": 45046 }, { "epoch": 0.87, "learning_rate": 3.543271936804922e-05, "loss": 0.0, "step": 45048 }, { "epoch": 0.87, "learning_rate": 3.543207240779199e-05, "loss": 0.0, "step": 45050 }, { "epoch": 0.87, "learning_rate": 3.543142544753476e-05, "loss": 0.0, "step": 45052 }, { "epoch": 0.87, "learning_rate": 3.543077848727753e-05, "loss": 0.0, "step": 45054 }, { "epoch": 0.87, "learning_rate": 3.54301315270203e-05, "loss": 0.0, "step": 45056 }, { "epoch": 0.87, "learning_rate": 3.542948456676307e-05, "loss": 0.0, "step": 45058 }, { "epoch": 0.87, "learning_rate": 3.542883760650583e-05, "loss": 0.0054, "step": 45060 }, { "epoch": 0.87, "learning_rate": 3.5428190646248605e-05, "loss": 0.0, "step": 45062 }, { "epoch": 0.87, "learning_rate": 3.5427543685991374e-05, "loss": 0.0003, "step": 45064 }, { "epoch": 0.87, "learning_rate": 3.5426896725734136e-05, "loss": 0.0015, "step": 45066 }, { "epoch": 0.87, "learning_rate": 3.542624976547691e-05, "loss": 0.0, "step": 45068 }, { "epoch": 0.87, "learning_rate": 3.5425602805219675e-05, "loss": 0.0177, "step": 45070 }, { "epoch": 0.87, "learning_rate": 3.5424955844962444e-05, "loss": 0.0, "step": 45072 }, { "epoch": 0.87, "learning_rate": 3.542430888470521e-05, "loss": 0.0, "step": 45074 }, { "epoch": 0.87, "learning_rate": 3.542366192444798e-05, "loss": 0.0, "step": 45076 }, { "epoch": 0.87, "learning_rate": 3.542301496419075e-05, "loss": 0.0062, "step": 45078 }, { "epoch": 0.87, "learning_rate": 3.542236800393352e-05, "loss": 0.0, "step": 45080 }, { "epoch": 0.87, "learning_rate": 3.542172104367629e-05, "loss": 0.0015, "step": 45082 }, { "epoch": 0.88, "learning_rate": 3.542107408341906e-05, "loss": 0.0, "step": 45084 }, { "epoch": 0.88, "learning_rate": 3.542042712316183e-05, "loss": 0.0057, "step": 45086 }, { "epoch": 0.88, "learning_rate": 3.541978016290459e-05, "loss": 0.0032, "step": 45088 }, { "epoch": 0.88, "learning_rate": 3.5419133202647365e-05, "loss": 0.0111, "step": 45090 }, { "epoch": 0.88, "learning_rate": 3.541848624239013e-05, "loss": 0.0, "step": 45092 }, { "epoch": 0.88, "learning_rate": 3.5417839282132904e-05, "loss": 0.0001, "step": 45094 }, { "epoch": 0.88, "learning_rate": 3.541719232187567e-05, "loss": 0.0, "step": 45096 }, { "epoch": 0.88, "learning_rate": 3.5416545361618435e-05, "loss": 0.0003, "step": 45098 }, { "epoch": 0.88, "learning_rate": 3.541589840136121e-05, "loss": 0.0, "step": 45100 }, { "epoch": 0.88, "learning_rate": 3.541525144110397e-05, "loss": 0.0, "step": 45102 }, { "epoch": 0.88, "learning_rate": 3.541460448084674e-05, "loss": 0.0, "step": 45104 }, { "epoch": 0.88, "learning_rate": 3.541395752058951e-05, "loss": 0.0, "step": 45106 }, { "epoch": 0.88, "learning_rate": 3.541331056033228e-05, "loss": 0.0, "step": 45108 }, { "epoch": 0.88, "learning_rate": 3.541266360007505e-05, "loss": 0.0122, "step": 45110 }, { "epoch": 0.88, "learning_rate": 3.541201663981782e-05, "loss": 0.0096, "step": 45112 }, { "epoch": 0.88, "learning_rate": 3.541136967956059e-05, "loss": 0.004, "step": 45114 }, { "epoch": 0.88, "learning_rate": 3.541072271930336e-05, "loss": 0.0003, "step": 45116 }, { "epoch": 0.88, "learning_rate": 3.5410075759046126e-05, "loss": 0.0, "step": 45118 }, { "epoch": 0.88, "learning_rate": 3.540942879878889e-05, "loss": 0.0, "step": 45120 }, { "epoch": 0.88, "learning_rate": 3.5408781838531664e-05, "loss": 0.0, "step": 45122 }, { "epoch": 0.88, "learning_rate": 3.5408134878274426e-05, "loss": 0.0, "step": 45124 }, { "epoch": 0.88, "learning_rate": 3.5407487918017195e-05, "loss": 0.0, "step": 45126 }, { "epoch": 0.88, "learning_rate": 3.540684095775997e-05, "loss": 0.0001, "step": 45128 }, { "epoch": 0.88, "learning_rate": 3.5406193997502734e-05, "loss": 0.0, "step": 45130 }, { "epoch": 0.88, "learning_rate": 3.54055470372455e-05, "loss": 0.0, "step": 45132 }, { "epoch": 0.88, "learning_rate": 3.540490007698827e-05, "loss": 0.0001, "step": 45134 }, { "epoch": 0.88, "learning_rate": 3.540425311673104e-05, "loss": 0.0, "step": 45136 }, { "epoch": 0.88, "learning_rate": 3.540360615647381e-05, "loss": 0.0, "step": 45138 }, { "epoch": 0.88, "learning_rate": 3.540295919621658e-05, "loss": 0.0001, "step": 45140 }, { "epoch": 0.88, "learning_rate": 3.540231223595935e-05, "loss": 0.0241, "step": 45142 }, { "epoch": 0.88, "learning_rate": 3.540166527570212e-05, "loss": 0.0, "step": 45144 }, { "epoch": 0.88, "learning_rate": 3.5401018315444886e-05, "loss": 0.0, "step": 45146 }, { "epoch": 0.88, "learning_rate": 3.540037135518765e-05, "loss": 0.0, "step": 45148 }, { "epoch": 0.88, "learning_rate": 3.5399724394930424e-05, "loss": 0.0001, "step": 45150 }, { "epoch": 0.88, "learning_rate": 3.539907743467319e-05, "loss": 0.0, "step": 45152 }, { "epoch": 0.88, "learning_rate": 3.5398430474415956e-05, "loss": 0.0001, "step": 45154 }, { "epoch": 0.88, "learning_rate": 3.5397783514158725e-05, "loss": 0.0, "step": 45156 }, { "epoch": 0.88, "learning_rate": 3.5397136553901494e-05, "loss": 0.0093, "step": 45158 }, { "epoch": 0.88, "learning_rate": 3.539648959364426e-05, "loss": 0.0001, "step": 45160 }, { "epoch": 0.88, "learning_rate": 3.539584263338703e-05, "loss": 0.0009, "step": 45162 }, { "epoch": 0.88, "learning_rate": 3.53951956731298e-05, "loss": 0.0, "step": 45164 }, { "epoch": 0.88, "learning_rate": 3.539454871287257e-05, "loss": 0.0, "step": 45166 }, { "epoch": 0.88, "learning_rate": 3.539390175261534e-05, "loss": 0.0, "step": 45168 }, { "epoch": 0.88, "learning_rate": 3.53932547923581e-05, "loss": 0.0, "step": 45170 }, { "epoch": 0.88, "learning_rate": 3.539260783210088e-05, "loss": 0.0, "step": 45172 }, { "epoch": 0.88, "learning_rate": 3.539196087184365e-05, "loss": 0.0001, "step": 45174 }, { "epoch": 0.88, "learning_rate": 3.5391313911586416e-05, "loss": 0.0231, "step": 45176 }, { "epoch": 0.88, "learning_rate": 3.5390666951329185e-05, "loss": 0.0, "step": 45178 }, { "epoch": 0.88, "learning_rate": 3.539001999107195e-05, "loss": 0.0005, "step": 45180 }, { "epoch": 0.88, "learning_rate": 3.538937303081472e-05, "loss": 0.0, "step": 45182 }, { "epoch": 0.88, "learning_rate": 3.5388726070557485e-05, "loss": 0.0001, "step": 45184 }, { "epoch": 0.88, "learning_rate": 3.5388079110300254e-05, "loss": 0.008, "step": 45186 }, { "epoch": 0.88, "learning_rate": 3.5387432150043023e-05, "loss": 0.0089, "step": 45188 }, { "epoch": 0.88, "learning_rate": 3.538678518978579e-05, "loss": 0.0, "step": 45190 }, { "epoch": 0.88, "learning_rate": 3.538613822952856e-05, "loss": 0.0, "step": 45192 }, { "epoch": 0.88, "learning_rate": 3.538549126927133e-05, "loss": 0.0011, "step": 45194 }, { "epoch": 0.88, "learning_rate": 3.53848443090141e-05, "loss": 0.0, "step": 45196 }, { "epoch": 0.88, "learning_rate": 3.538419734875687e-05, "loss": 0.0, "step": 45198 }, { "epoch": 0.88, "learning_rate": 3.538355038849964e-05, "loss": 0.0004, "step": 45200 }, { "epoch": 0.88, "learning_rate": 3.53829034282424e-05, "loss": 0.0027, "step": 45202 }, { "epoch": 0.88, "learning_rate": 3.5382256467985176e-05, "loss": 0.0008, "step": 45204 }, { "epoch": 0.88, "learning_rate": 3.5381609507727945e-05, "loss": 0.0016, "step": 45206 }, { "epoch": 0.88, "learning_rate": 3.538096254747071e-05, "loss": 0.0045, "step": 45208 }, { "epoch": 0.88, "learning_rate": 3.5380315587213483e-05, "loss": 0.0, "step": 45210 }, { "epoch": 0.88, "learning_rate": 3.5379668626956246e-05, "loss": 0.0004, "step": 45212 }, { "epoch": 0.88, "learning_rate": 3.5379021666699015e-05, "loss": 0.0123, "step": 45214 }, { "epoch": 0.88, "learning_rate": 3.5378374706441784e-05, "loss": 0.0294, "step": 45216 }, { "epoch": 0.88, "learning_rate": 3.537772774618455e-05, "loss": 0.0432, "step": 45218 }, { "epoch": 0.88, "learning_rate": 3.537708078592732e-05, "loss": 0.0, "step": 45220 }, { "epoch": 0.88, "learning_rate": 3.537643382567009e-05, "loss": 0.0, "step": 45222 }, { "epoch": 0.88, "learning_rate": 3.537578686541286e-05, "loss": 0.0, "step": 45224 }, { "epoch": 0.88, "learning_rate": 3.537513990515563e-05, "loss": 0.0114, "step": 45226 }, { "epoch": 0.88, "learning_rate": 3.53744929448984e-05, "loss": 0.0002, "step": 45228 }, { "epoch": 0.88, "learning_rate": 3.537384598464116e-05, "loss": 0.0, "step": 45230 }, { "epoch": 0.88, "learning_rate": 3.5373199024383937e-05, "loss": 0.0, "step": 45232 }, { "epoch": 0.88, "learning_rate": 3.53725520641267e-05, "loss": 0.0, "step": 45234 }, { "epoch": 0.88, "learning_rate": 3.5371905103869475e-05, "loss": 0.0, "step": 45236 }, { "epoch": 0.88, "learning_rate": 3.537125814361224e-05, "loss": 0.0, "step": 45238 }, { "epoch": 0.88, "learning_rate": 3.5370611183355006e-05, "loss": 0.0, "step": 45240 }, { "epoch": 0.88, "learning_rate": 3.536996422309778e-05, "loss": 0.0007, "step": 45242 }, { "epoch": 0.88, "learning_rate": 3.5369317262840544e-05, "loss": 0.0, "step": 45244 }, { "epoch": 0.88, "learning_rate": 3.536867030258331e-05, "loss": 0.0, "step": 45246 }, { "epoch": 0.88, "learning_rate": 3.536802334232608e-05, "loss": 0.0007, "step": 45248 }, { "epoch": 0.88, "learning_rate": 3.536737638206885e-05, "loss": 0.0001, "step": 45250 }, { "epoch": 0.88, "learning_rate": 3.536672942181162e-05, "loss": 0.0, "step": 45252 }, { "epoch": 0.88, "learning_rate": 3.536608246155439e-05, "loss": 0.0032, "step": 45254 }, { "epoch": 0.88, "learning_rate": 3.536543550129716e-05, "loss": 0.0002, "step": 45256 }, { "epoch": 0.88, "learning_rate": 3.536478854103993e-05, "loss": 0.0002, "step": 45258 }, { "epoch": 0.88, "learning_rate": 3.53641415807827e-05, "loss": 0.0009, "step": 45260 }, { "epoch": 0.88, "learning_rate": 3.536349462052546e-05, "loss": 0.0, "step": 45262 }, { "epoch": 0.88, "learning_rate": 3.5362847660268235e-05, "loss": 0.0, "step": 45264 }, { "epoch": 0.88, "learning_rate": 3.5362200700011e-05, "loss": 0.0, "step": 45266 }, { "epoch": 0.88, "learning_rate": 3.5361553739753767e-05, "loss": 0.0002, "step": 45268 }, { "epoch": 0.88, "learning_rate": 3.5360906779496536e-05, "loss": 0.0, "step": 45270 }, { "epoch": 0.88, "learning_rate": 3.5360259819239305e-05, "loss": 0.0, "step": 45272 }, { "epoch": 0.88, "learning_rate": 3.5359612858982074e-05, "loss": 0.0007, "step": 45274 }, { "epoch": 0.88, "learning_rate": 3.535896589872484e-05, "loss": 0.0002, "step": 45276 }, { "epoch": 0.88, "learning_rate": 3.535831893846761e-05, "loss": 0.004, "step": 45278 }, { "epoch": 0.88, "learning_rate": 3.535767197821038e-05, "loss": 0.0, "step": 45280 }, { "epoch": 0.88, "learning_rate": 3.535702501795315e-05, "loss": 0.0, "step": 45282 }, { "epoch": 0.88, "learning_rate": 3.535637805769591e-05, "loss": 0.0, "step": 45284 }, { "epoch": 0.88, "learning_rate": 3.535573109743869e-05, "loss": 0.0001, "step": 45286 }, { "epoch": 0.88, "learning_rate": 3.535508413718146e-05, "loss": 0.0001, "step": 45288 }, { "epoch": 0.88, "learning_rate": 3.535443717692422e-05, "loss": 0.0007, "step": 45290 }, { "epoch": 0.88, "learning_rate": 3.5353790216666996e-05, "loss": 0.0, "step": 45292 }, { "epoch": 0.88, "learning_rate": 3.535314325640976e-05, "loss": 0.0001, "step": 45294 }, { "epoch": 0.88, "learning_rate": 3.535249629615253e-05, "loss": 0.0, "step": 45296 }, { "epoch": 0.88, "learning_rate": 3.5351849335895296e-05, "loss": 0.0205, "step": 45298 }, { "epoch": 0.88, "learning_rate": 3.5351202375638065e-05, "loss": 0.0, "step": 45300 }, { "epoch": 0.88, "learning_rate": 3.5350555415380834e-05, "loss": 0.0, "step": 45302 }, { "epoch": 0.88, "learning_rate": 3.53499084551236e-05, "loss": 0.0, "step": 45304 }, { "epoch": 0.88, "learning_rate": 3.534926149486637e-05, "loss": 0.0, "step": 45306 }, { "epoch": 0.88, "learning_rate": 3.534861453460914e-05, "loss": 0.0, "step": 45308 }, { "epoch": 0.88, "learning_rate": 3.534796757435191e-05, "loss": 0.0001, "step": 45310 }, { "epoch": 0.88, "learning_rate": 3.534732061409467e-05, "loss": 0.0001, "step": 45312 }, { "epoch": 0.88, "learning_rate": 3.534667365383745e-05, "loss": 0.0001, "step": 45314 }, { "epoch": 0.88, "learning_rate": 3.534602669358021e-05, "loss": 0.0, "step": 45316 }, { "epoch": 0.88, "learning_rate": 3.534537973332299e-05, "loss": 0.0, "step": 45318 }, { "epoch": 0.88, "learning_rate": 3.5344732773065756e-05, "loss": 0.0, "step": 45320 }, { "epoch": 0.88, "learning_rate": 3.534408581280852e-05, "loss": 0.0, "step": 45322 }, { "epoch": 0.88, "learning_rate": 3.5343438852551294e-05, "loss": 0.0, "step": 45324 }, { "epoch": 0.88, "learning_rate": 3.5342791892294056e-05, "loss": 0.0, "step": 45326 }, { "epoch": 0.88, "learning_rate": 3.5342144932036826e-05, "loss": 0.0001, "step": 45328 }, { "epoch": 0.88, "learning_rate": 3.5341497971779595e-05, "loss": 0.0, "step": 45330 }, { "epoch": 0.88, "learning_rate": 3.5340851011522364e-05, "loss": 0.0, "step": 45332 }, { "epoch": 0.88, "learning_rate": 3.534020405126513e-05, "loss": 0.0, "step": 45334 }, { "epoch": 0.88, "learning_rate": 3.53395570910079e-05, "loss": 0.0036, "step": 45336 }, { "epoch": 0.88, "learning_rate": 3.533891013075067e-05, "loss": 0.0015, "step": 45338 }, { "epoch": 0.88, "learning_rate": 3.533826317049344e-05, "loss": 0.0, "step": 45340 }, { "epoch": 0.88, "learning_rate": 3.533761621023621e-05, "loss": 0.0, "step": 45342 }, { "epoch": 0.88, "learning_rate": 3.533696924997897e-05, "loss": 0.0, "step": 45344 }, { "epoch": 0.88, "learning_rate": 3.533632228972175e-05, "loss": 0.0, "step": 45346 }, { "epoch": 0.88, "learning_rate": 3.533567532946451e-05, "loss": 0.01, "step": 45348 }, { "epoch": 0.88, "learning_rate": 3.533502836920728e-05, "loss": 0.0001, "step": 45350 }, { "epoch": 0.88, "learning_rate": 3.5334381408950055e-05, "loss": 0.0, "step": 45352 }, { "epoch": 0.88, "learning_rate": 3.533373444869282e-05, "loss": 0.0029, "step": 45354 }, { "epoch": 0.88, "learning_rate": 3.5333087488435586e-05, "loss": 0.0, "step": 45356 }, { "epoch": 0.88, "learning_rate": 3.5332440528178355e-05, "loss": 0.0, "step": 45358 }, { "epoch": 0.88, "learning_rate": 3.5331793567921124e-05, "loss": 0.0, "step": 45360 }, { "epoch": 0.88, "learning_rate": 3.533114660766389e-05, "loss": 0.0, "step": 45362 }, { "epoch": 0.88, "learning_rate": 3.533049964740666e-05, "loss": 0.0064, "step": 45364 }, { "epoch": 0.88, "learning_rate": 3.532985268714943e-05, "loss": 0.0, "step": 45366 }, { "epoch": 0.88, "learning_rate": 3.53292057268922e-05, "loss": 0.0, "step": 45368 }, { "epoch": 0.88, "learning_rate": 3.532855876663497e-05, "loss": 0.0, "step": 45370 }, { "epoch": 0.88, "learning_rate": 3.532791180637773e-05, "loss": 0.0566, "step": 45372 }, { "epoch": 0.88, "learning_rate": 3.532726484612051e-05, "loss": 0.0001, "step": 45374 }, { "epoch": 0.88, "learning_rate": 3.532661788586327e-05, "loss": 0.0002, "step": 45376 }, { "epoch": 0.88, "learning_rate": 3.5325970925606046e-05, "loss": 0.0105, "step": 45378 }, { "epoch": 0.88, "learning_rate": 3.532532396534881e-05, "loss": 0.0167, "step": 45380 }, { "epoch": 0.88, "learning_rate": 3.532467700509158e-05, "loss": 0.0031, "step": 45382 }, { "epoch": 0.88, "learning_rate": 3.532403004483435e-05, "loss": 0.0034, "step": 45384 }, { "epoch": 0.88, "learning_rate": 3.5323383084577115e-05, "loss": 0.0001, "step": 45386 }, { "epoch": 0.88, "learning_rate": 3.5322736124319884e-05, "loss": 0.0, "step": 45388 }, { "epoch": 0.88, "learning_rate": 3.5322089164062654e-05, "loss": 0.0, "step": 45390 }, { "epoch": 0.88, "learning_rate": 3.532144220380542e-05, "loss": 0.0001, "step": 45392 }, { "epoch": 0.88, "learning_rate": 3.5320795243548185e-05, "loss": 0.0029, "step": 45394 }, { "epoch": 0.88, "learning_rate": 3.532014828329096e-05, "loss": 0.0, "step": 45396 }, { "epoch": 0.88, "learning_rate": 3.531950132303373e-05, "loss": 0.0, "step": 45398 }, { "epoch": 0.88, "learning_rate": 3.53188543627765e-05, "loss": 0.0058, "step": 45400 }, { "epoch": 0.88, "learning_rate": 3.531820740251927e-05, "loss": 0.0, "step": 45402 }, { "epoch": 0.88, "learning_rate": 3.531756044226203e-05, "loss": 0.0001, "step": 45404 }, { "epoch": 0.88, "learning_rate": 3.5316913482004806e-05, "loss": 0.0, "step": 45406 }, { "epoch": 0.88, "learning_rate": 3.531626652174757e-05, "loss": 0.0154, "step": 45408 }, { "epoch": 0.88, "learning_rate": 3.531561956149034e-05, "loss": 0.0016, "step": 45410 }, { "epoch": 0.88, "learning_rate": 3.531497260123311e-05, "loss": 0.0, "step": 45412 }, { "epoch": 0.88, "learning_rate": 3.5314325640975876e-05, "loss": 0.0001, "step": 45414 }, { "epoch": 0.88, "learning_rate": 3.5313678680718645e-05, "loss": 0.0, "step": 45416 }, { "epoch": 0.88, "learning_rate": 3.5313031720461414e-05, "loss": 0.0, "step": 45418 }, { "epoch": 0.88, "learning_rate": 3.531238476020418e-05, "loss": 0.0, "step": 45420 }, { "epoch": 0.88, "learning_rate": 3.531173779994695e-05, "loss": 0.0099, "step": 45422 }, { "epoch": 0.88, "learning_rate": 3.531109083968972e-05, "loss": 0.0, "step": 45424 }, { "epoch": 0.88, "learning_rate": 3.5310443879432484e-05, "loss": 0.0002, "step": 45426 }, { "epoch": 0.88, "learning_rate": 3.530979691917526e-05, "loss": 0.0, "step": 45428 }, { "epoch": 0.88, "learning_rate": 3.530914995891803e-05, "loss": 0.0, "step": 45430 }, { "epoch": 0.88, "learning_rate": 3.530850299866079e-05, "loss": 0.0, "step": 45432 }, { "epoch": 0.88, "learning_rate": 3.530785603840357e-05, "loss": 0.0012, "step": 45434 }, { "epoch": 0.88, "learning_rate": 3.530720907814633e-05, "loss": 0.0003, "step": 45436 }, { "epoch": 0.88, "learning_rate": 3.53065621178891e-05, "loss": 0.0111, "step": 45438 }, { "epoch": 0.88, "learning_rate": 3.530591515763187e-05, "loss": 0.0019, "step": 45440 }, { "epoch": 0.88, "learning_rate": 3.5305268197374636e-05, "loss": 0.0, "step": 45442 }, { "epoch": 0.88, "learning_rate": 3.5304621237117405e-05, "loss": 0.0, "step": 45444 }, { "epoch": 0.88, "learning_rate": 3.5303974276860174e-05, "loss": 0.0, "step": 45446 }, { "epoch": 0.88, "learning_rate": 3.5303327316602943e-05, "loss": 0.0, "step": 45448 }, { "epoch": 0.88, "learning_rate": 3.530268035634571e-05, "loss": 0.0, "step": 45450 }, { "epoch": 0.88, "learning_rate": 3.530203339608848e-05, "loss": 0.0001, "step": 45452 }, { "epoch": 0.88, "learning_rate": 3.5301386435831244e-05, "loss": 0.0077, "step": 45454 }, { "epoch": 0.88, "learning_rate": 3.530073947557402e-05, "loss": 0.0, "step": 45456 }, { "epoch": 0.88, "learning_rate": 3.530009251531678e-05, "loss": 0.0, "step": 45458 }, { "epoch": 0.88, "learning_rate": 3.529944555505956e-05, "loss": 0.0001, "step": 45460 }, { "epoch": 0.88, "learning_rate": 3.529879859480232e-05, "loss": 0.0, "step": 45462 }, { "epoch": 0.88, "learning_rate": 3.529815163454509e-05, "loss": 0.0009, "step": 45464 }, { "epoch": 0.88, "learning_rate": 3.5297504674287865e-05, "loss": 0.0, "step": 45466 }, { "epoch": 0.88, "learning_rate": 3.529685771403063e-05, "loss": 0.0003, "step": 45468 }, { "epoch": 0.88, "learning_rate": 3.52962107537734e-05, "loss": 0.0, "step": 45470 }, { "epoch": 0.88, "learning_rate": 3.5295563793516166e-05, "loss": 0.0311, "step": 45472 }, { "epoch": 0.88, "learning_rate": 3.5294916833258935e-05, "loss": 0.0, "step": 45474 }, { "epoch": 0.88, "learning_rate": 3.5294269873001704e-05, "loss": 0.0001, "step": 45476 }, { "epoch": 0.88, "learning_rate": 3.529362291274447e-05, "loss": 0.0001, "step": 45478 }, { "epoch": 0.88, "learning_rate": 3.529297595248724e-05, "loss": 0.0, "step": 45480 }, { "epoch": 0.88, "learning_rate": 3.529232899223001e-05, "loss": 0.0001, "step": 45482 }, { "epoch": 0.88, "learning_rate": 3.529168203197278e-05, "loss": 0.0, "step": 45484 }, { "epoch": 0.88, "learning_rate": 3.529103507171554e-05, "loss": 0.0, "step": 45486 }, { "epoch": 0.88, "learning_rate": 3.529038811145832e-05, "loss": 0.0006, "step": 45488 }, { "epoch": 0.88, "learning_rate": 3.528974115120108e-05, "loss": 0.0, "step": 45490 }, { "epoch": 0.88, "learning_rate": 3.528909419094385e-05, "loss": 0.0023, "step": 45492 }, { "epoch": 0.88, "learning_rate": 3.528844723068662e-05, "loss": 0.0, "step": 45494 }, { "epoch": 0.88, "learning_rate": 3.528780027042939e-05, "loss": 0.0, "step": 45496 }, { "epoch": 0.88, "learning_rate": 3.528715331017216e-05, "loss": 0.0, "step": 45498 }, { "epoch": 0.88, "learning_rate": 3.5286506349914926e-05, "loss": 0.0001, "step": 45500 }, { "epoch": 0.88, "learning_rate": 3.5285859389657695e-05, "loss": 0.0048, "step": 45502 }, { "epoch": 0.88, "learning_rate": 3.5285212429400464e-05, "loss": 0.0, "step": 45504 }, { "epoch": 0.88, "learning_rate": 3.528456546914323e-05, "loss": 0.0, "step": 45506 }, { "epoch": 0.88, "learning_rate": 3.5283918508886e-05, "loss": 0.0, "step": 45508 }, { "epoch": 0.88, "learning_rate": 3.528327154862877e-05, "loss": 0.0002, "step": 45510 }, { "epoch": 0.88, "learning_rate": 3.528262458837154e-05, "loss": 0.0001, "step": 45512 }, { "epoch": 0.88, "learning_rate": 3.52819776281143e-05, "loss": 0.0001, "step": 45514 }, { "epoch": 0.88, "learning_rate": 3.528133066785708e-05, "loss": 0.0034, "step": 45516 }, { "epoch": 0.88, "learning_rate": 3.528068370759984e-05, "loss": 0.0001, "step": 45518 }, { "epoch": 0.88, "learning_rate": 3.528003674734261e-05, "loss": 0.0003, "step": 45520 }, { "epoch": 0.88, "learning_rate": 3.527938978708538e-05, "loss": 0.0, "step": 45522 }, { "epoch": 0.88, "learning_rate": 3.527874282682815e-05, "loss": 0.0001, "step": 45524 }, { "epoch": 0.88, "learning_rate": 3.527809586657092e-05, "loss": 0.0, "step": 45526 }, { "epoch": 0.88, "learning_rate": 3.5277448906313687e-05, "loss": 0.0002, "step": 45528 }, { "epoch": 0.88, "learning_rate": 3.5276801946056456e-05, "loss": 0.0042, "step": 45530 }, { "epoch": 0.88, "learning_rate": 3.5276154985799225e-05, "loss": 0.0022, "step": 45532 }, { "epoch": 0.88, "learning_rate": 3.5275508025541994e-05, "loss": 0.0158, "step": 45534 }, { "epoch": 0.88, "learning_rate": 3.5274861065284756e-05, "loss": 0.0001, "step": 45536 }, { "epoch": 0.88, "learning_rate": 3.527421410502753e-05, "loss": 0.0015, "step": 45538 }, { "epoch": 0.88, "learning_rate": 3.5273567144770294e-05, "loss": 0.0001, "step": 45540 }, { "epoch": 0.88, "learning_rate": 3.527292018451307e-05, "loss": 0.0, "step": 45542 }, { "epoch": 0.88, "learning_rate": 3.527227322425584e-05, "loss": 0.0, "step": 45544 }, { "epoch": 0.88, "learning_rate": 3.52716262639986e-05, "loss": 0.0001, "step": 45546 }, { "epoch": 0.88, "learning_rate": 3.527097930374138e-05, "loss": 0.0, "step": 45548 }, { "epoch": 0.88, "learning_rate": 3.527033234348414e-05, "loss": 0.0001, "step": 45550 }, { "epoch": 0.88, "learning_rate": 3.526968538322691e-05, "loss": 0.0, "step": 45552 }, { "epoch": 0.88, "learning_rate": 3.526903842296968e-05, "loss": 0.0, "step": 45554 }, { "epoch": 0.88, "learning_rate": 3.526839146271245e-05, "loss": 0.0007, "step": 45556 }, { "epoch": 0.88, "learning_rate": 3.5267744502455216e-05, "loss": 0.0023, "step": 45558 }, { "epoch": 0.88, "learning_rate": 3.5267097542197985e-05, "loss": 0.0, "step": 45560 }, { "epoch": 0.88, "learning_rate": 3.5266450581940754e-05, "loss": 0.0, "step": 45562 }, { "epoch": 0.88, "learning_rate": 3.526580362168352e-05, "loss": 0.0, "step": 45564 }, { "epoch": 0.88, "learning_rate": 3.526515666142629e-05, "loss": 0.0004, "step": 45566 }, { "epoch": 0.88, "learning_rate": 3.5264509701169055e-05, "loss": 0.0001, "step": 45568 }, { "epoch": 0.88, "learning_rate": 3.526386274091183e-05, "loss": 0.0, "step": 45570 }, { "epoch": 0.88, "learning_rate": 3.526321578065459e-05, "loss": 0.0148, "step": 45572 }, { "epoch": 0.88, "learning_rate": 3.526256882039736e-05, "loss": 0.0, "step": 45574 }, { "epoch": 0.88, "learning_rate": 3.526192186014014e-05, "loss": 0.0002, "step": 45576 }, { "epoch": 0.88, "learning_rate": 3.52612748998829e-05, "loss": 0.0002, "step": 45578 }, { "epoch": 0.88, "learning_rate": 3.526062793962567e-05, "loss": 0.0, "step": 45580 }, { "epoch": 0.88, "learning_rate": 3.525998097936844e-05, "loss": 0.0001, "step": 45582 }, { "epoch": 0.88, "learning_rate": 3.525933401911121e-05, "loss": 0.0035, "step": 45584 }, { "epoch": 0.88, "learning_rate": 3.5258687058853976e-05, "loss": 0.0, "step": 45586 }, { "epoch": 0.88, "learning_rate": 3.5258040098596746e-05, "loss": 0.0, "step": 45588 }, { "epoch": 0.88, "learning_rate": 3.5257393138339515e-05, "loss": 0.0001, "step": 45590 }, { "epoch": 0.88, "learning_rate": 3.5256746178082284e-05, "loss": 0.0062, "step": 45592 }, { "epoch": 0.88, "learning_rate": 3.525609921782505e-05, "loss": 0.0001, "step": 45594 }, { "epoch": 0.88, "learning_rate": 3.5255452257567815e-05, "loss": 0.0001, "step": 45596 }, { "epoch": 0.89, "learning_rate": 3.525480529731059e-05, "loss": 0.0027, "step": 45598 }, { "epoch": 0.89, "learning_rate": 3.525415833705335e-05, "loss": 0.0001, "step": 45600 }, { "epoch": 0.89, "learning_rate": 3.525351137679613e-05, "loss": 0.0002, "step": 45602 }, { "epoch": 0.89, "learning_rate": 3.525286441653889e-05, "loss": 0.0, "step": 45604 }, { "epoch": 0.89, "learning_rate": 3.525221745628166e-05, "loss": 0.0004, "step": 45606 }, { "epoch": 0.89, "learning_rate": 3.5251570496024436e-05, "loss": 0.0016, "step": 45608 }, { "epoch": 0.89, "learning_rate": 3.52509235357672e-05, "loss": 0.0, "step": 45610 }, { "epoch": 0.89, "learning_rate": 3.525027657550997e-05, "loss": 0.0, "step": 45612 }, { "epoch": 0.89, "learning_rate": 3.524962961525274e-05, "loss": 0.0, "step": 45614 }, { "epoch": 0.89, "learning_rate": 3.5248982654995506e-05, "loss": 0.0, "step": 45616 }, { "epoch": 0.89, "learning_rate": 3.524833569473827e-05, "loss": 0.0001, "step": 45618 }, { "epoch": 0.89, "learning_rate": 3.5247688734481044e-05, "loss": 0.0, "step": 45620 }, { "epoch": 0.89, "learning_rate": 3.524704177422381e-05, "loss": 0.0, "step": 45622 }, { "epoch": 0.89, "learning_rate": 3.524639481396658e-05, "loss": 0.0132, "step": 45624 }, { "epoch": 0.89, "learning_rate": 3.524574785370935e-05, "loss": 0.0121, "step": 45626 }, { "epoch": 0.89, "learning_rate": 3.5245100893452114e-05, "loss": 0.0041, "step": 45628 }, { "epoch": 0.89, "learning_rate": 3.524445393319489e-05, "loss": 0.0, "step": 45630 }, { "epoch": 0.89, "learning_rate": 3.524380697293765e-05, "loss": 0.0109, "step": 45632 }, { "epoch": 0.89, "learning_rate": 3.524316001268042e-05, "loss": 0.0755, "step": 45634 }, { "epoch": 0.89, "learning_rate": 3.524251305242319e-05, "loss": 0.0057, "step": 45636 }, { "epoch": 0.89, "learning_rate": 3.524186609216596e-05, "loss": 0.0, "step": 45638 }, { "epoch": 0.89, "learning_rate": 3.524121913190873e-05, "loss": 0.0, "step": 45640 }, { "epoch": 0.89, "learning_rate": 3.52405721716515e-05, "loss": 0.0021, "step": 45642 }, { "epoch": 0.89, "learning_rate": 3.5239925211394266e-05, "loss": 0.0, "step": 45644 }, { "epoch": 0.89, "learning_rate": 3.5239278251137035e-05, "loss": 0.0015, "step": 45646 }, { "epoch": 0.89, "learning_rate": 3.5238631290879804e-05, "loss": 0.0149, "step": 45648 }, { "epoch": 0.89, "learning_rate": 3.523798433062257e-05, "loss": 0.0015, "step": 45650 }, { "epoch": 0.89, "learning_rate": 3.523733737036534e-05, "loss": 0.0087, "step": 45652 }, { "epoch": 0.89, "learning_rate": 3.523669041010811e-05, "loss": 0.0, "step": 45654 }, { "epoch": 0.89, "learning_rate": 3.5236043449850874e-05, "loss": 0.0, "step": 45656 }, { "epoch": 0.89, "learning_rate": 3.523539648959365e-05, "loss": 0.0015, "step": 45658 }, { "epoch": 0.89, "learning_rate": 3.523474952933641e-05, "loss": 0.0, "step": 45660 }, { "epoch": 0.89, "learning_rate": 3.523410256907918e-05, "loss": 0.0, "step": 45662 }, { "epoch": 0.89, "learning_rate": 3.523345560882195e-05, "loss": 0.0001, "step": 45664 }, { "epoch": 0.89, "learning_rate": 3.523280864856472e-05, "loss": 0.0, "step": 45666 }, { "epoch": 0.89, "learning_rate": 3.523216168830749e-05, "loss": 0.0019, "step": 45668 }, { "epoch": 0.89, "learning_rate": 3.523151472805026e-05, "loss": 0.0, "step": 45670 }, { "epoch": 0.89, "learning_rate": 3.523086776779303e-05, "loss": 0.0049, "step": 45672 }, { "epoch": 0.89, "learning_rate": 3.5230220807535796e-05, "loss": 0.0002, "step": 45674 }, { "epoch": 0.89, "learning_rate": 3.5229573847278565e-05, "loss": 0.0, "step": 45676 }, { "epoch": 0.89, "learning_rate": 3.522892688702133e-05, "loss": 0.0003, "step": 45678 }, { "epoch": 0.89, "learning_rate": 3.52282799267641e-05, "loss": 0.0002, "step": 45680 }, { "epoch": 0.89, "learning_rate": 3.5227632966506865e-05, "loss": 0.0, "step": 45682 }, { "epoch": 0.89, "learning_rate": 3.522698600624964e-05, "loss": 0.0, "step": 45684 }, { "epoch": 0.89, "learning_rate": 3.522633904599241e-05, "loss": 0.0, "step": 45686 }, { "epoch": 0.89, "learning_rate": 3.522569208573517e-05, "loss": 0.0, "step": 45688 }, { "epoch": 0.89, "learning_rate": 3.522504512547795e-05, "loss": 0.0, "step": 45690 }, { "epoch": 0.89, "learning_rate": 3.522439816522071e-05, "loss": 0.0001, "step": 45692 }, { "epoch": 0.89, "learning_rate": 3.522375120496348e-05, "loss": 0.0, "step": 45694 }, { "epoch": 0.89, "learning_rate": 3.522310424470625e-05, "loss": 0.0, "step": 45696 }, { "epoch": 0.89, "learning_rate": 3.522245728444902e-05, "loss": 0.0, "step": 45698 }, { "epoch": 0.89, "learning_rate": 3.522181032419179e-05, "loss": 0.0003, "step": 45700 }, { "epoch": 0.89, "learning_rate": 3.5221163363934556e-05, "loss": 0.0, "step": 45702 }, { "epoch": 0.89, "learning_rate": 3.5220516403677325e-05, "loss": 0.0, "step": 45704 }, { "epoch": 0.89, "learning_rate": 3.5219869443420094e-05, "loss": 0.0, "step": 45706 }, { "epoch": 0.89, "learning_rate": 3.5219222483162863e-05, "loss": 0.0099, "step": 45708 }, { "epoch": 0.89, "learning_rate": 3.5218575522905626e-05, "loss": 0.0229, "step": 45710 }, { "epoch": 0.89, "learning_rate": 3.52179285626484e-05, "loss": 0.002, "step": 45712 }, { "epoch": 0.89, "learning_rate": 3.5217281602391164e-05, "loss": 0.0, "step": 45714 }, { "epoch": 0.89, "learning_rate": 3.521663464213393e-05, "loss": 0.0016, "step": 45716 }, { "epoch": 0.89, "learning_rate": 3.52159876818767e-05, "loss": 0.0001, "step": 45718 }, { "epoch": 0.89, "learning_rate": 3.521534072161947e-05, "loss": 0.0, "step": 45720 }, { "epoch": 0.89, "learning_rate": 3.521469376136224e-05, "loss": 0.0, "step": 45722 }, { "epoch": 0.89, "learning_rate": 3.521404680110501e-05, "loss": 0.0, "step": 45724 }, { "epoch": 0.89, "learning_rate": 3.521339984084778e-05, "loss": 0.0001, "step": 45726 }, { "epoch": 0.89, "learning_rate": 3.521275288059055e-05, "loss": 0.0001, "step": 45728 }, { "epoch": 0.89, "learning_rate": 3.5212105920333317e-05, "loss": 0.0, "step": 45730 }, { "epoch": 0.89, "learning_rate": 3.5211458960076086e-05, "loss": 0.0, "step": 45732 }, { "epoch": 0.89, "learning_rate": 3.5210811999818855e-05, "loss": 0.0002, "step": 45734 }, { "epoch": 0.89, "learning_rate": 3.5210165039561624e-05, "loss": 0.0005, "step": 45736 }, { "epoch": 0.89, "learning_rate": 3.5209518079304386e-05, "loss": 0.007, "step": 45738 }, { "epoch": 0.89, "learning_rate": 3.520887111904716e-05, "loss": 0.0049, "step": 45740 }, { "epoch": 0.89, "learning_rate": 3.5208224158789924e-05, "loss": 0.0014, "step": 45742 }, { "epoch": 0.89, "learning_rate": 3.52075771985327e-05, "loss": 0.0109, "step": 45744 }, { "epoch": 0.89, "learning_rate": 3.520693023827546e-05, "loss": 0.0133, "step": 45746 }, { "epoch": 0.89, "learning_rate": 3.520628327801823e-05, "loss": 0.0, "step": 45748 }, { "epoch": 0.89, "learning_rate": 3.5205636317761e-05, "loss": 0.0002, "step": 45750 }, { "epoch": 0.89, "learning_rate": 3.520498935750377e-05, "loss": 0.0003, "step": 45752 }, { "epoch": 0.89, "learning_rate": 3.520434239724654e-05, "loss": 0.0008, "step": 45754 }, { "epoch": 0.89, "learning_rate": 3.520369543698931e-05, "loss": 0.0, "step": 45756 }, { "epoch": 0.89, "learning_rate": 3.520304847673208e-05, "loss": 0.0, "step": 45758 }, { "epoch": 0.89, "learning_rate": 3.520240151647484e-05, "loss": 0.0003, "step": 45760 }, { "epoch": 0.89, "learning_rate": 3.5201754556217615e-05, "loss": 0.0002, "step": 45762 }, { "epoch": 0.89, "learning_rate": 3.5201107595960384e-05, "loss": 0.0, "step": 45764 }, { "epoch": 0.89, "learning_rate": 3.520046063570315e-05, "loss": 0.0, "step": 45766 }, { "epoch": 0.89, "learning_rate": 3.519981367544592e-05, "loss": 0.0, "step": 45768 }, { "epoch": 0.89, "learning_rate": 3.5199166715188685e-05, "loss": 0.0, "step": 45770 }, { "epoch": 0.89, "learning_rate": 3.519851975493146e-05, "loss": 0.0, "step": 45772 }, { "epoch": 0.89, "learning_rate": 3.519787279467422e-05, "loss": 0.0, "step": 45774 }, { "epoch": 0.89, "learning_rate": 3.519722583441699e-05, "loss": 0.0, "step": 45776 }, { "epoch": 0.89, "learning_rate": 3.519657887415976e-05, "loss": 0.0, "step": 45778 }, { "epoch": 0.89, "learning_rate": 3.519593191390253e-05, "loss": 0.0, "step": 45780 }, { "epoch": 0.89, "learning_rate": 3.51952849536453e-05, "loss": 0.0001, "step": 45782 }, { "epoch": 0.89, "learning_rate": 3.519463799338807e-05, "loss": 0.0, "step": 45784 }, { "epoch": 0.89, "learning_rate": 3.519399103313084e-05, "loss": 0.0001, "step": 45786 }, { "epoch": 0.89, "learning_rate": 3.5193344072873607e-05, "loss": 0.0355, "step": 45788 }, { "epoch": 0.89, "learning_rate": 3.5192697112616376e-05, "loss": 0.0, "step": 45790 }, { "epoch": 0.89, "learning_rate": 3.519205015235914e-05, "loss": 0.0, "step": 45792 }, { "epoch": 0.89, "learning_rate": 3.5191403192101914e-05, "loss": 0.0002, "step": 45794 }, { "epoch": 0.89, "learning_rate": 3.5190756231844676e-05, "loss": 0.0, "step": 45796 }, { "epoch": 0.89, "learning_rate": 3.5190109271587445e-05, "loss": 0.0, "step": 45798 }, { "epoch": 0.89, "learning_rate": 3.518946231133022e-05, "loss": 0.0142, "step": 45800 }, { "epoch": 0.89, "learning_rate": 3.518881535107298e-05, "loss": 0.0, "step": 45802 }, { "epoch": 0.89, "learning_rate": 3.518816839081575e-05, "loss": 0.0, "step": 45804 }, { "epoch": 0.89, "learning_rate": 3.518752143055852e-05, "loss": 0.0, "step": 45806 }, { "epoch": 0.89, "learning_rate": 3.518687447030129e-05, "loss": 0.0059, "step": 45808 }, { "epoch": 0.89, "learning_rate": 3.518622751004406e-05, "loss": 0.0, "step": 45810 }, { "epoch": 0.89, "learning_rate": 3.518558054978683e-05, "loss": 0.0263, "step": 45812 }, { "epoch": 0.89, "learning_rate": 3.51849335895296e-05, "loss": 0.0, "step": 45814 }, { "epoch": 0.89, "learning_rate": 3.518428662927237e-05, "loss": 0.001, "step": 45816 }, { "epoch": 0.89, "learning_rate": 3.5183639669015136e-05, "loss": 0.0006, "step": 45818 }, { "epoch": 0.89, "learning_rate": 3.51829927087579e-05, "loss": 0.0002, "step": 45820 }, { "epoch": 0.89, "learning_rate": 3.5182345748500674e-05, "loss": 0.0076, "step": 45822 }, { "epoch": 0.89, "learning_rate": 3.5181698788243436e-05, "loss": 0.0002, "step": 45824 }, { "epoch": 0.89, "learning_rate": 3.518105182798621e-05, "loss": 0.0001, "step": 45826 }, { "epoch": 0.89, "learning_rate": 3.5180404867728975e-05, "loss": 0.0108, "step": 45828 }, { "epoch": 0.89, "learning_rate": 3.5179757907471744e-05, "loss": 0.0, "step": 45830 }, { "epoch": 0.89, "learning_rate": 3.517911094721452e-05, "loss": 0.0002, "step": 45832 }, { "epoch": 0.89, "learning_rate": 3.517846398695728e-05, "loss": 0.0001, "step": 45834 }, { "epoch": 0.89, "learning_rate": 3.517781702670005e-05, "loss": 0.0, "step": 45836 }, { "epoch": 0.89, "learning_rate": 3.517717006644282e-05, "loss": 0.0061, "step": 45838 }, { "epoch": 0.89, "learning_rate": 3.517652310618559e-05, "loss": 0.0, "step": 45840 }, { "epoch": 0.89, "learning_rate": 3.517587614592835e-05, "loss": 0.0, "step": 45842 }, { "epoch": 0.89, "learning_rate": 3.517522918567113e-05, "loss": 0.0001, "step": 45844 }, { "epoch": 0.89, "learning_rate": 3.5174582225413896e-05, "loss": 0.0, "step": 45846 }, { "epoch": 0.89, "learning_rate": 3.5173935265156665e-05, "loss": 0.0009, "step": 45848 }, { "epoch": 0.89, "learning_rate": 3.5173288304899435e-05, "loss": 0.0002, "step": 45850 }, { "epoch": 0.89, "learning_rate": 3.51726413446422e-05, "loss": 0.0001, "step": 45852 }, { "epoch": 0.89, "learning_rate": 3.517199438438497e-05, "loss": 0.0001, "step": 45854 }, { "epoch": 0.89, "learning_rate": 3.5171347424127735e-05, "loss": 0.0, "step": 45856 }, { "epoch": 0.89, "learning_rate": 3.5170700463870504e-05, "loss": 0.0, "step": 45858 }, { "epoch": 0.89, "learning_rate": 3.517005350361327e-05, "loss": 0.0, "step": 45860 }, { "epoch": 0.89, "learning_rate": 3.516940654335604e-05, "loss": 0.0027, "step": 45862 }, { "epoch": 0.89, "learning_rate": 3.516875958309881e-05, "loss": 0.0, "step": 45864 }, { "epoch": 0.89, "learning_rate": 3.516811262284158e-05, "loss": 0.0002, "step": 45866 }, { "epoch": 0.89, "learning_rate": 3.516746566258435e-05, "loss": 0.0, "step": 45868 }, { "epoch": 0.89, "learning_rate": 3.516681870232712e-05, "loss": 0.0027, "step": 45870 }, { "epoch": 0.89, "learning_rate": 3.516617174206989e-05, "loss": 0.0, "step": 45872 }, { "epoch": 0.89, "learning_rate": 3.516552478181265e-05, "loss": 0.0, "step": 45874 }, { "epoch": 0.89, "learning_rate": 3.5164877821555426e-05, "loss": 0.0, "step": 45876 }, { "epoch": 0.89, "learning_rate": 3.5164230861298195e-05, "loss": 0.0001, "step": 45878 }, { "epoch": 0.89, "learning_rate": 3.516358390104096e-05, "loss": 0.0005, "step": 45880 }, { "epoch": 0.89, "learning_rate": 3.516293694078373e-05, "loss": 0.0, "step": 45882 }, { "epoch": 0.89, "learning_rate": 3.5162289980526495e-05, "loss": 0.0, "step": 45884 }, { "epoch": 0.89, "learning_rate": 3.516164302026927e-05, "loss": 0.0104, "step": 45886 }, { "epoch": 0.89, "learning_rate": 3.5160996060012034e-05, "loss": 0.0, "step": 45888 }, { "epoch": 0.89, "learning_rate": 3.51603490997548e-05, "loss": 0.0, "step": 45890 }, { "epoch": 0.89, "learning_rate": 3.515970213949757e-05, "loss": 0.0, "step": 45892 }, { "epoch": 0.89, "learning_rate": 3.515905517924034e-05, "loss": 0.0, "step": 45894 }, { "epoch": 0.89, "learning_rate": 3.515840821898311e-05, "loss": 0.0007, "step": 45896 }, { "epoch": 0.89, "learning_rate": 3.515776125872588e-05, "loss": 0.0, "step": 45898 }, { "epoch": 0.89, "learning_rate": 3.515711429846865e-05, "loss": 0.0001, "step": 45900 }, { "epoch": 0.89, "learning_rate": 3.515646733821141e-05, "loss": 0.0, "step": 45902 }, { "epoch": 0.89, "learning_rate": 3.5155820377954186e-05, "loss": 0.0001, "step": 45904 }, { "epoch": 0.89, "learning_rate": 3.515517341769695e-05, "loss": 0.0001, "step": 45906 }, { "epoch": 0.89, "learning_rate": 3.5154526457439724e-05, "loss": 0.0, "step": 45908 }, { "epoch": 0.89, "learning_rate": 3.5153879497182494e-05, "loss": 0.0, "step": 45910 }, { "epoch": 0.89, "learning_rate": 3.5153232536925256e-05, "loss": 0.0, "step": 45912 }, { "epoch": 0.89, "learning_rate": 3.515258557666803e-05, "loss": 0.0, "step": 45914 }, { "epoch": 0.89, "learning_rate": 3.5151938616410794e-05, "loss": 0.0, "step": 45916 }, { "epoch": 0.89, "learning_rate": 3.515129165615356e-05, "loss": 0.0001, "step": 45918 }, { "epoch": 0.89, "learning_rate": 3.515064469589633e-05, "loss": 0.011, "step": 45920 }, { "epoch": 0.89, "learning_rate": 3.51499977356391e-05, "loss": 0.0, "step": 45922 }, { "epoch": 0.89, "learning_rate": 3.514935077538187e-05, "loss": 0.0001, "step": 45924 }, { "epoch": 0.89, "learning_rate": 3.514870381512464e-05, "loss": 0.0, "step": 45926 }, { "epoch": 0.89, "learning_rate": 3.514805685486741e-05, "loss": 0.0001, "step": 45928 }, { "epoch": 0.89, "learning_rate": 3.514740989461018e-05, "loss": 0.0051, "step": 45930 }, { "epoch": 0.89, "learning_rate": 3.514676293435295e-05, "loss": 0.0002, "step": 45932 }, { "epoch": 0.89, "learning_rate": 3.514611597409571e-05, "loss": 0.0, "step": 45934 }, { "epoch": 0.89, "learning_rate": 3.5145469013838485e-05, "loss": 0.009, "step": 45936 }, { "epoch": 0.89, "learning_rate": 3.514482205358125e-05, "loss": 0.0022, "step": 45938 }, { "epoch": 0.89, "learning_rate": 3.5144175093324016e-05, "loss": 0.0092, "step": 45940 }, { "epoch": 0.89, "learning_rate": 3.5143528133066785e-05, "loss": 0.0, "step": 45942 }, { "epoch": 0.89, "learning_rate": 3.5142881172809554e-05, "loss": 0.0001, "step": 45944 }, { "epoch": 0.89, "learning_rate": 3.5142234212552324e-05, "loss": 0.0, "step": 45946 }, { "epoch": 0.89, "learning_rate": 3.514158725229509e-05, "loss": 0.0, "step": 45948 }, { "epoch": 0.89, "learning_rate": 3.514094029203786e-05, "loss": 0.0, "step": 45950 }, { "epoch": 0.89, "learning_rate": 3.514029333178063e-05, "loss": 0.0, "step": 45952 }, { "epoch": 0.89, "learning_rate": 3.51396463715234e-05, "loss": 0.0003, "step": 45954 }, { "epoch": 0.89, "learning_rate": 3.513899941126617e-05, "loss": 0.0, "step": 45956 }, { "epoch": 0.89, "learning_rate": 3.513835245100894e-05, "loss": 0.0001, "step": 45958 }, { "epoch": 0.89, "learning_rate": 3.513770549075171e-05, "loss": 0.0, "step": 45960 }, { "epoch": 0.89, "learning_rate": 3.513705853049447e-05, "loss": 0.0, "step": 45962 }, { "epoch": 0.89, "learning_rate": 3.5136411570237245e-05, "loss": 0.0, "step": 45964 }, { "epoch": 0.89, "learning_rate": 3.513576460998001e-05, "loss": 0.0067, "step": 45966 }, { "epoch": 0.89, "learning_rate": 3.5135117649722783e-05, "loss": 0.0, "step": 45968 }, { "epoch": 0.89, "learning_rate": 3.5134470689465546e-05, "loss": 0.0, "step": 45970 }, { "epoch": 0.89, "learning_rate": 3.5133823729208315e-05, "loss": 0.0, "step": 45972 }, { "epoch": 0.89, "learning_rate": 3.5133176768951084e-05, "loss": 0.0, "step": 45974 }, { "epoch": 0.89, "learning_rate": 3.513252980869385e-05, "loss": 0.0, "step": 45976 }, { "epoch": 0.89, "learning_rate": 3.513188284843662e-05, "loss": 0.0, "step": 45978 }, { "epoch": 0.89, "learning_rate": 3.513123588817939e-05, "loss": 0.0001, "step": 45980 }, { "epoch": 0.89, "learning_rate": 3.513058892792216e-05, "loss": 0.0003, "step": 45982 }, { "epoch": 0.89, "learning_rate": 3.512994196766492e-05, "loss": 0.0, "step": 45984 }, { "epoch": 0.89, "learning_rate": 3.51292950074077e-05, "loss": 0.014, "step": 45986 }, { "epoch": 0.89, "learning_rate": 3.512864804715047e-05, "loss": 0.0001, "step": 45988 }, { "epoch": 0.89, "learning_rate": 3.5128001086893237e-05, "loss": 0.0, "step": 45990 }, { "epoch": 0.89, "learning_rate": 3.5127354126636006e-05, "loss": 0.0001, "step": 45992 }, { "epoch": 0.89, "learning_rate": 3.512670716637877e-05, "loss": 0.0, "step": 45994 }, { "epoch": 0.89, "learning_rate": 3.5126060206121544e-05, "loss": 0.0, "step": 45996 }, { "epoch": 0.89, "learning_rate": 3.5125413245864306e-05, "loss": 0.0, "step": 45998 }, { "epoch": 0.89, "learning_rate": 3.5124766285607075e-05, "loss": 0.0, "step": 46000 }, { "epoch": 0.89, "learning_rate": 3.5124119325349844e-05, "loss": 0.0, "step": 46002 }, { "epoch": 0.89, "learning_rate": 3.5123472365092613e-05, "loss": 0.0001, "step": 46004 }, { "epoch": 0.89, "learning_rate": 3.512282540483538e-05, "loss": 0.0, "step": 46006 }, { "epoch": 0.89, "learning_rate": 3.512217844457815e-05, "loss": 0.0001, "step": 46008 }, { "epoch": 0.89, "learning_rate": 3.512153148432092e-05, "loss": 0.0, "step": 46010 }, { "epoch": 0.89, "learning_rate": 3.512088452406369e-05, "loss": 0.0, "step": 46012 }, { "epoch": 0.89, "learning_rate": 3.512023756380646e-05, "loss": 0.0, "step": 46014 }, { "epoch": 0.89, "learning_rate": 3.511959060354922e-05, "loss": 0.0148, "step": 46016 }, { "epoch": 0.89, "learning_rate": 3.5118943643292e-05, "loss": 0.0001, "step": 46018 }, { "epoch": 0.89, "learning_rate": 3.511829668303476e-05, "loss": 0.0, "step": 46020 }, { "epoch": 0.89, "learning_rate": 3.511764972277753e-05, "loss": 0.0001, "step": 46022 }, { "epoch": 0.89, "learning_rate": 3.5117002762520304e-05, "loss": 0.0002, "step": 46024 }, { "epoch": 0.89, "learning_rate": 3.5116355802263067e-05, "loss": 0.0, "step": 46026 }, { "epoch": 0.89, "learning_rate": 3.511570884200584e-05, "loss": 0.0, "step": 46028 }, { "epoch": 0.89, "learning_rate": 3.5115061881748605e-05, "loss": 0.0024, "step": 46030 }, { "epoch": 0.89, "learning_rate": 3.5114414921491374e-05, "loss": 0.0001, "step": 46032 }, { "epoch": 0.89, "learning_rate": 3.511376796123414e-05, "loss": 0.0124, "step": 46034 }, { "epoch": 0.89, "learning_rate": 3.511312100097691e-05, "loss": 0.0027, "step": 46036 }, { "epoch": 0.89, "learning_rate": 3.511247404071968e-05, "loss": 0.0001, "step": 46038 }, { "epoch": 0.89, "learning_rate": 3.511182708046245e-05, "loss": 0.0, "step": 46040 }, { "epoch": 0.89, "learning_rate": 3.511118012020522e-05, "loss": 0.0161, "step": 46042 }, { "epoch": 0.89, "learning_rate": 3.511053315994798e-05, "loss": 0.0, "step": 46044 }, { "epoch": 0.89, "learning_rate": 3.510988619969076e-05, "loss": 0.0272, "step": 46046 }, { "epoch": 0.89, "learning_rate": 3.510923923943352e-05, "loss": 0.0, "step": 46048 }, { "epoch": 0.89, "learning_rate": 3.5108592279176296e-05, "loss": 0.0001, "step": 46050 }, { "epoch": 0.89, "learning_rate": 3.510794531891906e-05, "loss": 0.0058, "step": 46052 }, { "epoch": 0.89, "learning_rate": 3.510729835866183e-05, "loss": 0.0, "step": 46054 }, { "epoch": 0.89, "learning_rate": 3.51066513984046e-05, "loss": 0.0, "step": 46056 }, { "epoch": 0.89, "learning_rate": 3.5106004438147365e-05, "loss": 0.0191, "step": 46058 }, { "epoch": 0.89, "learning_rate": 3.5105357477890134e-05, "loss": 0.0157, "step": 46060 }, { "epoch": 0.89, "learning_rate": 3.51047105176329e-05, "loss": 0.0, "step": 46062 }, { "epoch": 0.89, "learning_rate": 3.510406355737567e-05, "loss": 0.0001, "step": 46064 }, { "epoch": 0.89, "learning_rate": 3.510341659711844e-05, "loss": 0.0, "step": 46066 }, { "epoch": 0.89, "learning_rate": 3.510276963686121e-05, "loss": 0.0, "step": 46068 }, { "epoch": 0.89, "learning_rate": 3.510212267660398e-05, "loss": 0.0001, "step": 46070 }, { "epoch": 0.89, "learning_rate": 3.510147571634675e-05, "loss": 0.0001, "step": 46072 }, { "epoch": 0.89, "learning_rate": 3.510082875608952e-05, "loss": 0.0, "step": 46074 }, { "epoch": 0.89, "learning_rate": 3.510018179583228e-05, "loss": 0.0, "step": 46076 }, { "epoch": 0.89, "learning_rate": 3.5099534835575056e-05, "loss": 0.0, "step": 46078 }, { "epoch": 0.89, "learning_rate": 3.509888787531782e-05, "loss": 0.0, "step": 46080 }, { "epoch": 0.89, "learning_rate": 3.509824091506059e-05, "loss": 0.0, "step": 46082 }, { "epoch": 0.89, "learning_rate": 3.5097593954803356e-05, "loss": 0.0001, "step": 46084 }, { "epoch": 0.89, "learning_rate": 3.5096946994546126e-05, "loss": 0.0001, "step": 46086 }, { "epoch": 0.89, "learning_rate": 3.5096300034288895e-05, "loss": 0.0, "step": 46088 }, { "epoch": 0.89, "learning_rate": 3.5095653074031664e-05, "loss": 0.0001, "step": 46090 }, { "epoch": 0.89, "learning_rate": 3.509500611377443e-05, "loss": 0.0, "step": 46092 }, { "epoch": 0.89, "learning_rate": 3.50943591535172e-05, "loss": 0.0004, "step": 46094 }, { "epoch": 0.89, "learning_rate": 3.509371219325997e-05, "loss": 0.0001, "step": 46096 }, { "epoch": 0.89, "learning_rate": 3.509306523300273e-05, "loss": 0.0, "step": 46098 }, { "epoch": 0.89, "learning_rate": 3.509241827274551e-05, "loss": 0.0, "step": 46100 }, { "epoch": 0.89, "learning_rate": 3.509177131248828e-05, "loss": 0.0, "step": 46102 }, { "epoch": 0.89, "learning_rate": 3.509112435223104e-05, "loss": 0.0, "step": 46104 }, { "epoch": 0.89, "learning_rate": 3.5090477391973816e-05, "loss": 0.0, "step": 46106 }, { "epoch": 0.89, "learning_rate": 3.508983043171658e-05, "loss": 0.0012, "step": 46108 }, { "epoch": 0.89, "learning_rate": 3.5089183471459355e-05, "loss": 0.0, "step": 46110 }, { "epoch": 0.89, "learning_rate": 3.508853651120212e-05, "loss": 0.0001, "step": 46112 }, { "epoch": 0.9, "learning_rate": 3.5087889550944886e-05, "loss": 0.0001, "step": 46114 }, { "epoch": 0.9, "learning_rate": 3.5087242590687655e-05, "loss": 0.0001, "step": 46116 }, { "epoch": 0.9, "learning_rate": 3.5086595630430424e-05, "loss": 0.0, "step": 46118 }, { "epoch": 0.9, "learning_rate": 3.508594867017319e-05, "loss": 0.0, "step": 46120 }, { "epoch": 0.9, "learning_rate": 3.508530170991596e-05, "loss": 0.0, "step": 46122 }, { "epoch": 0.9, "learning_rate": 3.508465474965873e-05, "loss": 0.0, "step": 46124 }, { "epoch": 0.9, "learning_rate": 3.5084007789401494e-05, "loss": 0.0001, "step": 46126 }, { "epoch": 0.9, "learning_rate": 3.508336082914427e-05, "loss": 0.0, "step": 46128 }, { "epoch": 0.9, "learning_rate": 3.508271386888703e-05, "loss": 0.0001, "step": 46130 }, { "epoch": 0.9, "learning_rate": 3.508206690862981e-05, "loss": 0.0001, "step": 46132 }, { "epoch": 0.9, "learning_rate": 3.508141994837258e-05, "loss": 0.002, "step": 46134 }, { "epoch": 0.9, "learning_rate": 3.508077298811534e-05, "loss": 0.0, "step": 46136 }, { "epoch": 0.9, "learning_rate": 3.5080126027858115e-05, "loss": 0.0, "step": 46138 }, { "epoch": 0.9, "learning_rate": 3.507947906760088e-05, "loss": 0.0086, "step": 46140 }, { "epoch": 0.9, "learning_rate": 3.5078832107343646e-05, "loss": 0.002, "step": 46142 }, { "epoch": 0.9, "learning_rate": 3.5078185147086415e-05, "loss": 0.0086, "step": 46144 }, { "epoch": 0.9, "learning_rate": 3.5077538186829185e-05, "loss": 0.0, "step": 46146 }, { "epoch": 0.9, "learning_rate": 3.5076891226571954e-05, "loss": 0.0012, "step": 46148 }, { "epoch": 0.9, "learning_rate": 3.507624426631472e-05, "loss": 0.0, "step": 46150 }, { "epoch": 0.9, "learning_rate": 3.507559730605749e-05, "loss": 0.0, "step": 46152 }, { "epoch": 0.9, "learning_rate": 3.507495034580026e-05, "loss": 0.0034, "step": 46154 }, { "epoch": 0.9, "learning_rate": 3.507430338554303e-05, "loss": 0.0, "step": 46156 }, { "epoch": 0.9, "learning_rate": 3.507365642528579e-05, "loss": 0.0, "step": 46158 }, { "epoch": 0.9, "learning_rate": 3.507300946502857e-05, "loss": 0.0075, "step": 46160 }, { "epoch": 0.9, "learning_rate": 3.507236250477133e-05, "loss": 0.0, "step": 46162 }, { "epoch": 0.9, "learning_rate": 3.50717155445141e-05, "loss": 0.0, "step": 46164 }, { "epoch": 0.9, "learning_rate": 3.5071068584256875e-05, "loss": 0.0, "step": 46166 }, { "epoch": 0.9, "learning_rate": 3.507042162399964e-05, "loss": 0.0, "step": 46168 }, { "epoch": 0.9, "learning_rate": 3.5069774663742414e-05, "loss": 0.0, "step": 46170 }, { "epoch": 0.9, "learning_rate": 3.5069127703485176e-05, "loss": 0.0003, "step": 46172 }, { "epoch": 0.9, "learning_rate": 3.5068480743227945e-05, "loss": 0.0014, "step": 46174 }, { "epoch": 0.9, "learning_rate": 3.5067833782970714e-05, "loss": 0.0008, "step": 46176 }, { "epoch": 0.9, "learning_rate": 3.506718682271348e-05, "loss": 0.0001, "step": 46178 }, { "epoch": 0.9, "learning_rate": 3.506653986245625e-05, "loss": 0.0, "step": 46180 }, { "epoch": 0.9, "learning_rate": 3.506589290219902e-05, "loss": 0.0001, "step": 46182 }, { "epoch": 0.9, "learning_rate": 3.506524594194179e-05, "loss": 0.0, "step": 46184 }, { "epoch": 0.9, "learning_rate": 3.506459898168455e-05, "loss": 0.0, "step": 46186 }, { "epoch": 0.9, "learning_rate": 3.506395202142733e-05, "loss": 0.0001, "step": 46188 }, { "epoch": 0.9, "learning_rate": 3.506330506117009e-05, "loss": 0.0, "step": 46190 }, { "epoch": 0.9, "learning_rate": 3.506265810091287e-05, "loss": 0.0004, "step": 46192 }, { "epoch": 0.9, "learning_rate": 3.506201114065563e-05, "loss": 0.0185, "step": 46194 }, { "epoch": 0.9, "learning_rate": 3.50613641803984e-05, "loss": 0.0003, "step": 46196 }, { "epoch": 0.9, "learning_rate": 3.506071722014117e-05, "loss": 0.0002, "step": 46198 }, { "epoch": 0.9, "learning_rate": 3.5060070259883936e-05, "loss": 0.0069, "step": 46200 }, { "epoch": 0.9, "learning_rate": 3.5059423299626705e-05, "loss": 0.0001, "step": 46202 }, { "epoch": 0.9, "learning_rate": 3.5058776339369474e-05, "loss": 0.0, "step": 46204 }, { "epoch": 0.9, "learning_rate": 3.5058129379112243e-05, "loss": 0.0, "step": 46206 }, { "epoch": 0.9, "learning_rate": 3.5057482418855006e-05, "loss": 0.0, "step": 46208 }, { "epoch": 0.9, "learning_rate": 3.505683545859778e-05, "loss": 0.0, "step": 46210 }, { "epoch": 0.9, "learning_rate": 3.505618849834055e-05, "loss": 0.0001, "step": 46212 }, { "epoch": 0.9, "learning_rate": 3.505554153808332e-05, "loss": 0.0, "step": 46214 }, { "epoch": 0.9, "learning_rate": 3.505489457782609e-05, "loss": 0.0011, "step": 46216 }, { "epoch": 0.9, "learning_rate": 3.505424761756885e-05, "loss": 0.0, "step": 46218 }, { "epoch": 0.9, "learning_rate": 3.505360065731163e-05, "loss": 0.0099, "step": 46220 }, { "epoch": 0.9, "learning_rate": 3.505295369705439e-05, "loss": 0.0, "step": 46222 }, { "epoch": 0.9, "learning_rate": 3.505230673679716e-05, "loss": 0.0002, "step": 46224 }, { "epoch": 0.9, "learning_rate": 3.505165977653993e-05, "loss": 0.0001, "step": 46226 }, { "epoch": 0.9, "learning_rate": 3.50510128162827e-05, "loss": 0.0, "step": 46228 }, { "epoch": 0.9, "learning_rate": 3.5050365856025466e-05, "loss": 0.0, "step": 46230 }, { "epoch": 0.9, "learning_rate": 3.5049718895768235e-05, "loss": 0.0004, "step": 46232 }, { "epoch": 0.9, "learning_rate": 3.5049071935511004e-05, "loss": 0.0002, "step": 46234 }, { "epoch": 0.9, "learning_rate": 3.504842497525377e-05, "loss": 0.0, "step": 46236 }, { "epoch": 0.9, "learning_rate": 3.504777801499654e-05, "loss": 0.0, "step": 46238 }, { "epoch": 0.9, "learning_rate": 3.5047131054739304e-05, "loss": 0.0, "step": 46240 }, { "epoch": 0.9, "learning_rate": 3.504648409448208e-05, "loss": 0.0001, "step": 46242 }, { "epoch": 0.9, "learning_rate": 3.504583713422485e-05, "loss": 0.0001, "step": 46244 }, { "epoch": 0.9, "learning_rate": 3.504519017396761e-05, "loss": 0.0002, "step": 46246 }, { "epoch": 0.9, "learning_rate": 3.504454321371039e-05, "loss": 0.0014, "step": 46248 }, { "epoch": 0.9, "learning_rate": 3.504389625345315e-05, "loss": 0.0, "step": 46250 }, { "epoch": 0.9, "learning_rate": 3.5043249293195926e-05, "loss": 0.0066, "step": 46252 }, { "epoch": 0.9, "learning_rate": 3.504260233293869e-05, "loss": 0.0, "step": 46254 }, { "epoch": 0.9, "learning_rate": 3.504195537268146e-05, "loss": 0.0, "step": 46256 }, { "epoch": 0.9, "learning_rate": 3.5041308412424226e-05, "loss": 0.0001, "step": 46258 }, { "epoch": 0.9, "learning_rate": 3.5040661452166995e-05, "loss": 0.0, "step": 46260 }, { "epoch": 0.9, "learning_rate": 3.5040014491909764e-05, "loss": 0.0126, "step": 46262 }, { "epoch": 0.9, "learning_rate": 3.503936753165253e-05, "loss": 0.0, "step": 46264 }, { "epoch": 0.9, "learning_rate": 3.50387205713953e-05, "loss": 0.0031, "step": 46266 }, { "epoch": 0.9, "learning_rate": 3.5038073611138065e-05, "loss": 0.0002, "step": 46268 }, { "epoch": 0.9, "learning_rate": 3.503742665088084e-05, "loss": 0.0029, "step": 46270 }, { "epoch": 0.9, "learning_rate": 3.50367796906236e-05, "loss": 0.0, "step": 46272 }, { "epoch": 0.9, "learning_rate": 3.503613273036638e-05, "loss": 0.0, "step": 46274 }, { "epoch": 0.9, "learning_rate": 3.503548577010914e-05, "loss": 0.0, "step": 46276 }, { "epoch": 0.9, "learning_rate": 3.503483880985191e-05, "loss": 0.0001, "step": 46278 }, { "epoch": 0.9, "learning_rate": 3.5034191849594686e-05, "loss": 0.0001, "step": 46280 }, { "epoch": 0.9, "learning_rate": 3.503354488933745e-05, "loss": 0.0, "step": 46282 }, { "epoch": 0.9, "learning_rate": 3.503289792908022e-05, "loss": 0.0, "step": 46284 }, { "epoch": 0.9, "learning_rate": 3.5032250968822987e-05, "loss": 0.001, "step": 46286 }, { "epoch": 0.9, "learning_rate": 3.5031604008565756e-05, "loss": 0.0001, "step": 46288 }, { "epoch": 0.9, "learning_rate": 3.5030957048308525e-05, "loss": 0.0, "step": 46290 }, { "epoch": 0.9, "learning_rate": 3.5030310088051294e-05, "loss": 0.0, "step": 46292 }, { "epoch": 0.9, "learning_rate": 3.502966312779406e-05, "loss": 0.0068, "step": 46294 }, { "epoch": 0.9, "learning_rate": 3.502901616753683e-05, "loss": 0.0, "step": 46296 }, { "epoch": 0.9, "learning_rate": 3.50283692072796e-05, "loss": 0.0003, "step": 46298 }, { "epoch": 0.9, "learning_rate": 3.502772224702236e-05, "loss": 0.0, "step": 46300 }, { "epoch": 0.9, "learning_rate": 3.502707528676514e-05, "loss": 0.0049, "step": 46302 }, { "epoch": 0.9, "learning_rate": 3.50264283265079e-05, "loss": 0.0, "step": 46304 }, { "epoch": 0.9, "learning_rate": 3.502578136625067e-05, "loss": 0.0119, "step": 46306 }, { "epoch": 0.9, "learning_rate": 3.502513440599344e-05, "loss": 0.0, "step": 46308 }, { "epoch": 0.9, "learning_rate": 3.502448744573621e-05, "loss": 0.0, "step": 46310 }, { "epoch": 0.9, "learning_rate": 3.5023840485478985e-05, "loss": 0.0, "step": 46312 }, { "epoch": 0.9, "learning_rate": 3.502319352522175e-05, "loss": 0.0, "step": 46314 }, { "epoch": 0.9, "learning_rate": 3.5022546564964516e-05, "loss": 0.0003, "step": 46316 }, { "epoch": 0.9, "learning_rate": 3.5021899604707285e-05, "loss": 0.0, "step": 46318 }, { "epoch": 0.9, "learning_rate": 3.5021252644450054e-05, "loss": 0.0017, "step": 46320 }, { "epoch": 0.9, "learning_rate": 3.5020605684192817e-05, "loss": 0.0, "step": 46322 }, { "epoch": 0.9, "learning_rate": 3.501995872393559e-05, "loss": 0.0001, "step": 46324 }, { "epoch": 0.9, "learning_rate": 3.501931176367836e-05, "loss": 0.0, "step": 46326 }, { "epoch": 0.9, "learning_rate": 3.5018664803421124e-05, "loss": 0.0, "step": 46328 }, { "epoch": 0.9, "learning_rate": 3.50180178431639e-05, "loss": 0.0, "step": 46330 }, { "epoch": 0.9, "learning_rate": 3.501737088290666e-05, "loss": 0.0, "step": 46332 }, { "epoch": 0.9, "learning_rate": 3.501672392264944e-05, "loss": 0.0039, "step": 46334 }, { "epoch": 0.9, "learning_rate": 3.50160769623922e-05, "loss": 0.0, "step": 46336 }, { "epoch": 0.9, "learning_rate": 3.501543000213497e-05, "loss": 0.0, "step": 46338 }, { "epoch": 0.9, "learning_rate": 3.501478304187774e-05, "loss": 0.0, "step": 46340 }, { "epoch": 0.9, "learning_rate": 3.501413608162051e-05, "loss": 0.002, "step": 46342 }, { "epoch": 0.9, "learning_rate": 3.5013489121363276e-05, "loss": 0.0001, "step": 46344 }, { "epoch": 0.9, "learning_rate": 3.5012842161106046e-05, "loss": 0.0001, "step": 46346 }, { "epoch": 0.9, "learning_rate": 3.5012195200848815e-05, "loss": 0.0, "step": 46348 }, { "epoch": 0.9, "learning_rate": 3.501154824059158e-05, "loss": 0.0005, "step": 46350 }, { "epoch": 0.9, "learning_rate": 3.501090128033435e-05, "loss": 0.0269, "step": 46352 }, { "epoch": 0.9, "learning_rate": 3.5010254320077115e-05, "loss": 0.0009, "step": 46354 }, { "epoch": 0.9, "learning_rate": 3.500960735981989e-05, "loss": 0.0, "step": 46356 }, { "epoch": 0.9, "learning_rate": 3.500896039956266e-05, "loss": 0.0, "step": 46358 }, { "epoch": 0.9, "learning_rate": 3.500831343930542e-05, "loss": 0.0, "step": 46360 }, { "epoch": 0.9, "learning_rate": 3.50076664790482e-05, "loss": 0.0003, "step": 46362 }, { "epoch": 0.9, "learning_rate": 3.500701951879096e-05, "loss": 0.0, "step": 46364 }, { "epoch": 0.9, "learning_rate": 3.500637255853373e-05, "loss": 0.0137, "step": 46366 }, { "epoch": 0.9, "learning_rate": 3.50057255982765e-05, "loss": 0.0002, "step": 46368 }, { "epoch": 0.9, "learning_rate": 3.500507863801927e-05, "loss": 0.0, "step": 46370 }, { "epoch": 0.9, "learning_rate": 3.500443167776204e-05, "loss": 0.0, "step": 46372 }, { "epoch": 0.9, "learning_rate": 3.5003784717504806e-05, "loss": 0.0036, "step": 46374 }, { "epoch": 0.9, "learning_rate": 3.5003137757247575e-05, "loss": 0.0, "step": 46376 }, { "epoch": 0.9, "learning_rate": 3.5002490796990344e-05, "loss": 0.0005, "step": 46378 }, { "epoch": 0.9, "learning_rate": 3.500184383673311e-05, "loss": 0.0, "step": 46380 }, { "epoch": 0.9, "learning_rate": 3.5001196876475875e-05, "loss": 0.0, "step": 46382 }, { "epoch": 0.9, "learning_rate": 3.500054991621865e-05, "loss": 0.0424, "step": 46384 }, { "epoch": 0.9, "learning_rate": 3.4999902955961414e-05, "loss": 0.0, "step": 46386 }, { "epoch": 0.9, "learning_rate": 3.499925599570418e-05, "loss": 0.0, "step": 46388 }, { "epoch": 0.9, "learning_rate": 3.499860903544696e-05, "loss": 0.0013, "step": 46390 }, { "epoch": 0.9, "learning_rate": 3.499796207518972e-05, "loss": 0.0001, "step": 46392 }, { "epoch": 0.9, "learning_rate": 3.49973151149325e-05, "loss": 0.0, "step": 46394 }, { "epoch": 0.9, "learning_rate": 3.499666815467526e-05, "loss": 0.0, "step": 46396 }, { "epoch": 0.9, "learning_rate": 3.499602119441803e-05, "loss": 0.0, "step": 46398 }, { "epoch": 0.9, "learning_rate": 3.49953742341608e-05, "loss": 0.0028, "step": 46400 }, { "epoch": 0.9, "learning_rate": 3.4994727273903566e-05, "loss": 0.0, "step": 46402 }, { "epoch": 0.9, "learning_rate": 3.4994080313646335e-05, "loss": 0.0, "step": 46404 }, { "epoch": 0.9, "learning_rate": 3.4993433353389104e-05, "loss": 0.0, "step": 46406 }, { "epoch": 0.9, "learning_rate": 3.4992786393131874e-05, "loss": 0.0, "step": 46408 }, { "epoch": 0.9, "learning_rate": 3.4992139432874636e-05, "loss": 0.0, "step": 46410 }, { "epoch": 0.9, "learning_rate": 3.499149247261741e-05, "loss": 0.0027, "step": 46412 }, { "epoch": 0.9, "learning_rate": 3.4990845512360174e-05, "loss": 0.0021, "step": 46414 }, { "epoch": 0.9, "learning_rate": 3.499019855210295e-05, "loss": 0.0, "step": 46416 }, { "epoch": 0.9, "learning_rate": 3.498955159184571e-05, "loss": 0.0001, "step": 46418 }, { "epoch": 0.9, "learning_rate": 3.498890463158848e-05, "loss": 0.0011, "step": 46420 }, { "epoch": 0.9, "learning_rate": 3.498825767133125e-05, "loss": 0.0001, "step": 46422 }, { "epoch": 0.9, "learning_rate": 3.498761071107402e-05, "loss": 0.0001, "step": 46424 }, { "epoch": 0.9, "learning_rate": 3.498696375081679e-05, "loss": 0.0019, "step": 46426 }, { "epoch": 0.9, "learning_rate": 3.498631679055956e-05, "loss": 0.0, "step": 46428 }, { "epoch": 0.9, "learning_rate": 3.498566983030233e-05, "loss": 0.0002, "step": 46430 }, { "epoch": 0.9, "learning_rate": 3.498502287004509e-05, "loss": 0.0091, "step": 46432 }, { "epoch": 0.9, "learning_rate": 3.4984375909787865e-05, "loss": 0.0007, "step": 46434 }, { "epoch": 0.9, "learning_rate": 3.4983728949530634e-05, "loss": 0.0, "step": 46436 }, { "epoch": 0.9, "learning_rate": 3.49830819892734e-05, "loss": 0.0001, "step": 46438 }, { "epoch": 0.9, "learning_rate": 3.498243502901617e-05, "loss": 0.0, "step": 46440 }, { "epoch": 0.9, "learning_rate": 3.4981788068758934e-05, "loss": 0.0021, "step": 46442 }, { "epoch": 0.9, "learning_rate": 3.498114110850171e-05, "loss": 0.0, "step": 46444 }, { "epoch": 0.9, "learning_rate": 3.498049414824447e-05, "loss": 0.001, "step": 46446 }, { "epoch": 0.9, "learning_rate": 3.497984718798724e-05, "loss": 0.0003, "step": 46448 }, { "epoch": 0.9, "learning_rate": 3.497920022773001e-05, "loss": 0.0, "step": 46450 }, { "epoch": 0.9, "learning_rate": 3.497855326747278e-05, "loss": 0.0, "step": 46452 }, { "epoch": 0.9, "learning_rate": 3.497790630721555e-05, "loss": 0.0, "step": 46454 }, { "epoch": 0.9, "learning_rate": 3.497725934695832e-05, "loss": 0.0, "step": 46456 }, { "epoch": 0.9, "learning_rate": 3.497661238670109e-05, "loss": 0.0, "step": 46458 }, { "epoch": 0.9, "learning_rate": 3.4975965426443856e-05, "loss": 0.0, "step": 46460 }, { "epoch": 0.9, "learning_rate": 3.4975318466186625e-05, "loss": 0.0003, "step": 46462 }, { "epoch": 0.9, "learning_rate": 3.497467150592939e-05, "loss": 0.0, "step": 46464 }, { "epoch": 0.9, "learning_rate": 3.4974024545672163e-05, "loss": 0.0237, "step": 46466 }, { "epoch": 0.9, "learning_rate": 3.497337758541493e-05, "loss": 0.0, "step": 46468 }, { "epoch": 0.9, "learning_rate": 3.4972730625157695e-05, "loss": 0.008, "step": 46470 }, { "epoch": 0.9, "learning_rate": 3.497208366490047e-05, "loss": 0.0, "step": 46472 }, { "epoch": 0.9, "learning_rate": 3.497143670464323e-05, "loss": 0.0, "step": 46474 }, { "epoch": 0.9, "learning_rate": 3.497078974438601e-05, "loss": 0.0004, "step": 46476 }, { "epoch": 0.9, "learning_rate": 3.497014278412877e-05, "loss": 0.0, "step": 46478 }, { "epoch": 0.9, "learning_rate": 3.496949582387154e-05, "loss": 0.0, "step": 46480 }, { "epoch": 0.9, "learning_rate": 3.496884886361431e-05, "loss": 0.0, "step": 46482 }, { "epoch": 0.9, "learning_rate": 3.496820190335708e-05, "loss": 0.0, "step": 46484 }, { "epoch": 0.9, "learning_rate": 3.496755494309985e-05, "loss": 0.0, "step": 46486 }, { "epoch": 0.9, "learning_rate": 3.496690798284262e-05, "loss": 0.0, "step": 46488 }, { "epoch": 0.9, "learning_rate": 3.4966261022585386e-05, "loss": 0.059, "step": 46490 }, { "epoch": 0.9, "learning_rate": 3.496561406232815e-05, "loss": 0.0005, "step": 46492 }, { "epoch": 0.9, "learning_rate": 3.4964967102070924e-05, "loss": 0.0, "step": 46494 }, { "epoch": 0.9, "learning_rate": 3.4964320141813686e-05, "loss": 0.0001, "step": 46496 }, { "epoch": 0.9, "learning_rate": 3.496367318155646e-05, "loss": 0.0097, "step": 46498 }, { "epoch": 0.9, "learning_rate": 3.4963026221299224e-05, "loss": 0.0001, "step": 46500 }, { "epoch": 0.9, "learning_rate": 3.4962379261041993e-05, "loss": 0.0005, "step": 46502 }, { "epoch": 0.9, "learning_rate": 3.496173230078477e-05, "loss": 0.0002, "step": 46504 }, { "epoch": 0.9, "learning_rate": 3.496108534052753e-05, "loss": 0.0, "step": 46506 }, { "epoch": 0.9, "learning_rate": 3.49604383802703e-05, "loss": 0.0, "step": 46508 }, { "epoch": 0.9, "learning_rate": 3.495979142001307e-05, "loss": 0.0, "step": 46510 }, { "epoch": 0.9, "learning_rate": 3.495914445975584e-05, "loss": 0.0005, "step": 46512 }, { "epoch": 0.9, "learning_rate": 3.495849749949861e-05, "loss": 0.0, "step": 46514 }, { "epoch": 0.9, "learning_rate": 3.495785053924138e-05, "loss": 0.0, "step": 46516 }, { "epoch": 0.9, "learning_rate": 3.4957203578984146e-05, "loss": 0.0, "step": 46518 }, { "epoch": 0.9, "learning_rate": 3.4956556618726915e-05, "loss": 0.009, "step": 46520 }, { "epoch": 0.9, "learning_rate": 3.4955909658469684e-05, "loss": 0.0042, "step": 46522 }, { "epoch": 0.9, "learning_rate": 3.4955262698212447e-05, "loss": 0.0, "step": 46524 }, { "epoch": 0.9, "learning_rate": 3.495461573795522e-05, "loss": 0.0, "step": 46526 }, { "epoch": 0.9, "learning_rate": 3.4953968777697985e-05, "loss": 0.0001, "step": 46528 }, { "epoch": 0.9, "learning_rate": 3.4953321817440754e-05, "loss": 0.0, "step": 46530 }, { "epoch": 0.9, "learning_rate": 3.495267485718352e-05, "loss": 0.0, "step": 46532 }, { "epoch": 0.9, "learning_rate": 3.495202789692629e-05, "loss": 0.0, "step": 46534 }, { "epoch": 0.9, "learning_rate": 3.495138093666907e-05, "loss": 0.0, "step": 46536 }, { "epoch": 0.9, "learning_rate": 3.495073397641183e-05, "loss": 0.0, "step": 46538 }, { "epoch": 0.9, "learning_rate": 3.49500870161546e-05, "loss": 0.0, "step": 46540 }, { "epoch": 0.9, "learning_rate": 3.494944005589737e-05, "loss": 0.0003, "step": 46542 }, { "epoch": 0.9, "learning_rate": 3.494879309564014e-05, "loss": 0.0, "step": 46544 }, { "epoch": 0.9, "learning_rate": 3.4948146135382907e-05, "loss": 0.0, "step": 46546 }, { "epoch": 0.9, "learning_rate": 3.4947499175125676e-05, "loss": 0.0, "step": 46548 }, { "epoch": 0.9, "learning_rate": 3.4946852214868445e-05, "loss": 0.0, "step": 46550 }, { "epoch": 0.9, "learning_rate": 3.494620525461121e-05, "loss": 0.0001, "step": 46552 }, { "epoch": 0.9, "learning_rate": 3.494555829435398e-05, "loss": 0.0, "step": 46554 }, { "epoch": 0.9, "learning_rate": 3.4944911334096745e-05, "loss": 0.0, "step": 46556 }, { "epoch": 0.9, "learning_rate": 3.494426437383952e-05, "loss": 0.0007, "step": 46558 }, { "epoch": 0.9, "learning_rate": 3.494361741358228e-05, "loss": 0.0, "step": 46560 }, { "epoch": 0.9, "learning_rate": 3.494297045332505e-05, "loss": 0.0, "step": 46562 }, { "epoch": 0.9, "learning_rate": 3.494232349306782e-05, "loss": 0.0, "step": 46564 }, { "epoch": 0.9, "learning_rate": 3.494167653281059e-05, "loss": 0.0, "step": 46566 }, { "epoch": 0.9, "learning_rate": 3.494102957255336e-05, "loss": 0.0001, "step": 46568 }, { "epoch": 0.9, "learning_rate": 3.494038261229613e-05, "loss": 0.0, "step": 46570 }, { "epoch": 0.9, "learning_rate": 3.49397356520389e-05, "loss": 0.0002, "step": 46572 }, { "epoch": 0.9, "learning_rate": 3.493908869178166e-05, "loss": 0.0026, "step": 46574 }, { "epoch": 0.9, "learning_rate": 3.4938441731524436e-05, "loss": 0.0, "step": 46576 }, { "epoch": 0.9, "learning_rate": 3.49377947712672e-05, "loss": 0.0001, "step": 46578 }, { "epoch": 0.9, "learning_rate": 3.4937147811009974e-05, "loss": 0.0329, "step": 46580 }, { "epoch": 0.9, "learning_rate": 3.493650085075274e-05, "loss": 0.0002, "step": 46582 }, { "epoch": 0.9, "learning_rate": 3.4935853890495506e-05, "loss": 0.0, "step": 46584 }, { "epoch": 0.9, "learning_rate": 3.493520693023828e-05, "loss": 0.0, "step": 46586 }, { "epoch": 0.9, "learning_rate": 3.4934559969981044e-05, "loss": 0.0001, "step": 46588 }, { "epoch": 0.9, "learning_rate": 3.493391300972381e-05, "loss": 0.0001, "step": 46590 }, { "epoch": 0.9, "learning_rate": 3.493326604946658e-05, "loss": 0.0, "step": 46592 }, { "epoch": 0.9, "learning_rate": 3.493261908920935e-05, "loss": 0.0001, "step": 46594 }, { "epoch": 0.9, "learning_rate": 3.493197212895212e-05, "loss": 0.0001, "step": 46596 }, { "epoch": 0.9, "learning_rate": 3.493132516869489e-05, "loss": 0.0, "step": 46598 }, { "epoch": 0.9, "learning_rate": 3.493067820843766e-05, "loss": 0.0, "step": 46600 }, { "epoch": 0.9, "learning_rate": 3.493003124818043e-05, "loss": 0.0, "step": 46602 }, { "epoch": 0.9, "learning_rate": 3.4929384287923196e-05, "loss": 0.0001, "step": 46604 }, { "epoch": 0.9, "learning_rate": 3.492873732766596e-05, "loss": 0.0, "step": 46606 }, { "epoch": 0.9, "learning_rate": 3.4928090367408735e-05, "loss": 0.0001, "step": 46608 }, { "epoch": 0.9, "learning_rate": 3.49274434071515e-05, "loss": 0.0, "step": 46610 }, { "epoch": 0.9, "learning_rate": 3.4926796446894266e-05, "loss": 0.0, "step": 46612 }, { "epoch": 0.9, "learning_rate": 3.492614948663704e-05, "loss": 0.0, "step": 46614 }, { "epoch": 0.9, "learning_rate": 3.4925502526379804e-05, "loss": 0.0208, "step": 46616 }, { "epoch": 0.9, "learning_rate": 3.492485556612258e-05, "loss": 0.0, "step": 46618 }, { "epoch": 0.9, "learning_rate": 3.492420860586534e-05, "loss": 0.0007, "step": 46620 }, { "epoch": 0.9, "learning_rate": 3.492356164560811e-05, "loss": 0.0, "step": 46622 }, { "epoch": 0.9, "learning_rate": 3.492291468535088e-05, "loss": 0.0008, "step": 46624 }, { "epoch": 0.9, "learning_rate": 3.492226772509365e-05, "loss": 0.0003, "step": 46626 }, { "epoch": 0.9, "learning_rate": 3.492162076483642e-05, "loss": 0.008, "step": 46628 }, { "epoch": 0.91, "learning_rate": 3.492097380457919e-05, "loss": 0.0013, "step": 46630 }, { "epoch": 0.91, "learning_rate": 3.492032684432196e-05, "loss": 0.0, "step": 46632 }, { "epoch": 0.91, "learning_rate": 3.491967988406472e-05, "loss": 0.0, "step": 46634 }, { "epoch": 0.91, "learning_rate": 3.4919032923807495e-05, "loss": 0.0, "step": 46636 }, { "epoch": 0.91, "learning_rate": 3.491838596355026e-05, "loss": 0.0102, "step": 46638 }, { "epoch": 0.91, "learning_rate": 3.491773900329303e-05, "loss": 0.0, "step": 46640 }, { "epoch": 0.91, "learning_rate": 3.4917092043035795e-05, "loss": 0.0, "step": 46642 }, { "epoch": 0.91, "learning_rate": 3.4916445082778565e-05, "loss": 0.0, "step": 46644 }, { "epoch": 0.91, "learning_rate": 3.491579812252134e-05, "loss": 0.0, "step": 46646 }, { "epoch": 0.91, "learning_rate": 3.49151511622641e-05, "loss": 0.0001, "step": 46648 }, { "epoch": 0.91, "learning_rate": 3.491450420200687e-05, "loss": 0.0002, "step": 46650 }, { "epoch": 0.91, "learning_rate": 3.491385724174964e-05, "loss": 0.0001, "step": 46652 }, { "epoch": 0.91, "learning_rate": 3.491321028149241e-05, "loss": 0.0009, "step": 46654 }, { "epoch": 0.91, "learning_rate": 3.491256332123517e-05, "loss": 0.0, "step": 46656 }, { "epoch": 0.91, "learning_rate": 3.491191636097795e-05, "loss": 0.0131, "step": 46658 }, { "epoch": 0.91, "learning_rate": 3.491126940072072e-05, "loss": 0.0003, "step": 46660 }, { "epoch": 0.91, "learning_rate": 3.4910622440463486e-05, "loss": 0.0, "step": 46662 }, { "epoch": 0.91, "learning_rate": 3.4909975480206255e-05, "loss": 0.0, "step": 46664 }, { "epoch": 0.91, "learning_rate": 3.490932851994902e-05, "loss": 0.0001, "step": 46666 }, { "epoch": 0.91, "learning_rate": 3.4908681559691794e-05, "loss": 0.0001, "step": 46668 }, { "epoch": 0.91, "learning_rate": 3.4908034599434556e-05, "loss": 0.0009, "step": 46670 }, { "epoch": 0.91, "learning_rate": 3.4907387639177325e-05, "loss": 0.0006, "step": 46672 }, { "epoch": 0.91, "learning_rate": 3.4906740678920094e-05, "loss": 0.0201, "step": 46674 }, { "epoch": 0.91, "learning_rate": 3.490609371866286e-05, "loss": 0.0, "step": 46676 }, { "epoch": 0.91, "learning_rate": 3.490544675840563e-05, "loss": 0.0, "step": 46678 }, { "epoch": 0.91, "learning_rate": 3.49047997981484e-05, "loss": 0.0, "step": 46680 }, { "epoch": 0.91, "learning_rate": 3.490415283789117e-05, "loss": 0.0, "step": 46682 }, { "epoch": 0.91, "learning_rate": 3.490350587763394e-05, "loss": 0.0093, "step": 46684 }, { "epoch": 0.91, "learning_rate": 3.490285891737671e-05, "loss": 0.0115, "step": 46686 }, { "epoch": 0.91, "learning_rate": 3.490221195711947e-05, "loss": 0.0001, "step": 46688 }, { "epoch": 0.91, "learning_rate": 3.490156499686225e-05, "loss": 0.0, "step": 46690 }, { "epoch": 0.91, "learning_rate": 3.4900918036605016e-05, "loss": 0.0053, "step": 46692 }, { "epoch": 0.91, "learning_rate": 3.490027107634778e-05, "loss": 0.0, "step": 46694 }, { "epoch": 0.91, "learning_rate": 3.4899624116090554e-05, "loss": 0.0, "step": 46696 }, { "epoch": 0.91, "learning_rate": 3.4898977155833316e-05, "loss": 0.0037, "step": 46698 }, { "epoch": 0.91, "learning_rate": 3.489833019557609e-05, "loss": 0.003, "step": 46700 }, { "epoch": 0.91, "learning_rate": 3.4897683235318854e-05, "loss": 0.0, "step": 46702 }, { "epoch": 0.91, "learning_rate": 3.4897036275061624e-05, "loss": 0.0, "step": 46704 }, { "epoch": 0.91, "learning_rate": 3.489638931480439e-05, "loss": 0.0, "step": 46706 }, { "epoch": 0.91, "learning_rate": 3.489574235454716e-05, "loss": 0.0, "step": 46708 }, { "epoch": 0.91, "learning_rate": 3.489509539428993e-05, "loss": 0.0, "step": 46710 }, { "epoch": 0.91, "learning_rate": 3.48944484340327e-05, "loss": 0.0001, "step": 46712 }, { "epoch": 0.91, "learning_rate": 3.489380147377547e-05, "loss": 0.0, "step": 46714 }, { "epoch": 0.91, "learning_rate": 3.489315451351823e-05, "loss": 0.0002, "step": 46716 }, { "epoch": 0.91, "learning_rate": 3.489250755326101e-05, "loss": 0.0, "step": 46718 }, { "epoch": 0.91, "learning_rate": 3.489186059300377e-05, "loss": 0.0016, "step": 46720 }, { "epoch": 0.91, "learning_rate": 3.4891213632746545e-05, "loss": 0.0017, "step": 46722 }, { "epoch": 0.91, "learning_rate": 3.4890566672489314e-05, "loss": 0.0, "step": 46724 }, { "epoch": 0.91, "learning_rate": 3.488991971223208e-05, "loss": 0.0, "step": 46726 }, { "epoch": 0.91, "learning_rate": 3.488927275197485e-05, "loss": 0.0002, "step": 46728 }, { "epoch": 0.91, "learning_rate": 3.4888625791717615e-05, "loss": 0.0, "step": 46730 }, { "epoch": 0.91, "learning_rate": 3.4887978831460384e-05, "loss": 0.0002, "step": 46732 }, { "epoch": 0.91, "learning_rate": 3.488733187120315e-05, "loss": 0.0, "step": 46734 }, { "epoch": 0.91, "learning_rate": 3.488668491094592e-05, "loss": 0.0016, "step": 46736 }, { "epoch": 0.91, "learning_rate": 3.488603795068869e-05, "loss": 0.0, "step": 46738 }, { "epoch": 0.91, "learning_rate": 3.488539099043146e-05, "loss": 0.0001, "step": 46740 }, { "epoch": 0.91, "learning_rate": 3.488474403017423e-05, "loss": 0.0, "step": 46742 }, { "epoch": 0.91, "learning_rate": 3.4884097069917e-05, "loss": 0.0, "step": 46744 }, { "epoch": 0.91, "learning_rate": 3.488345010965977e-05, "loss": 0.0006, "step": 46746 }, { "epoch": 0.91, "learning_rate": 3.488280314940253e-05, "loss": 0.0, "step": 46748 }, { "epoch": 0.91, "learning_rate": 3.4882156189145306e-05, "loss": 0.0001, "step": 46750 }, { "epoch": 0.91, "learning_rate": 3.488150922888807e-05, "loss": 0.0, "step": 46752 }, { "epoch": 0.91, "learning_rate": 3.488086226863084e-05, "loss": 0.0213, "step": 46754 }, { "epoch": 0.91, "learning_rate": 3.4880215308373606e-05, "loss": 0.0, "step": 46756 }, { "epoch": 0.91, "learning_rate": 3.4879568348116375e-05, "loss": 0.0, "step": 46758 }, { "epoch": 0.91, "learning_rate": 3.487892138785915e-05, "loss": 0.0009, "step": 46760 }, { "epoch": 0.91, "learning_rate": 3.4878274427601913e-05, "loss": 0.0001, "step": 46762 }, { "epoch": 0.91, "learning_rate": 3.487762746734468e-05, "loss": 0.0007, "step": 46764 }, { "epoch": 0.91, "learning_rate": 3.487698050708745e-05, "loss": 0.0003, "step": 46766 }, { "epoch": 0.91, "learning_rate": 3.487633354683022e-05, "loss": 0.0, "step": 46768 }, { "epoch": 0.91, "learning_rate": 3.487568658657299e-05, "loss": 0.0002, "step": 46770 }, { "epoch": 0.91, "learning_rate": 3.487503962631576e-05, "loss": 0.0059, "step": 46772 }, { "epoch": 0.91, "learning_rate": 3.487439266605853e-05, "loss": 0.0, "step": 46774 }, { "epoch": 0.91, "learning_rate": 3.487374570580129e-05, "loss": 0.0116, "step": 46776 }, { "epoch": 0.91, "learning_rate": 3.4873098745544066e-05, "loss": 0.0, "step": 46778 }, { "epoch": 0.91, "learning_rate": 3.487245178528683e-05, "loss": 0.0, "step": 46780 }, { "epoch": 0.91, "learning_rate": 3.4871804825029604e-05, "loss": 0.0, "step": 46782 }, { "epoch": 0.91, "learning_rate": 3.4871157864772367e-05, "loss": 0.0095, "step": 46784 }, { "epoch": 0.91, "learning_rate": 3.4870510904515136e-05, "loss": 0.0091, "step": 46786 }, { "epoch": 0.91, "learning_rate": 3.4869863944257905e-05, "loss": 0.0, "step": 46788 }, { "epoch": 0.91, "learning_rate": 3.4869216984000674e-05, "loss": 0.0008, "step": 46790 }, { "epoch": 0.91, "learning_rate": 3.486857002374344e-05, "loss": 0.0183, "step": 46792 }, { "epoch": 0.91, "learning_rate": 3.486792306348621e-05, "loss": 0.0005, "step": 46794 }, { "epoch": 0.91, "learning_rate": 3.486727610322898e-05, "loss": 0.0001, "step": 46796 }, { "epoch": 0.91, "learning_rate": 3.486662914297174e-05, "loss": 0.0001, "step": 46798 }, { "epoch": 0.91, "learning_rate": 3.486598218271452e-05, "loss": 0.0, "step": 46800 }, { "epoch": 0.91, "learning_rate": 3.486533522245728e-05, "loss": 0.0, "step": 46802 }, { "epoch": 0.91, "learning_rate": 3.486468826220006e-05, "loss": 0.0, "step": 46804 }, { "epoch": 0.91, "learning_rate": 3.4864041301942827e-05, "loss": 0.0001, "step": 46806 }, { "epoch": 0.91, "learning_rate": 3.486339434168559e-05, "loss": 0.0046, "step": 46808 }, { "epoch": 0.91, "learning_rate": 3.4862747381428365e-05, "loss": 0.0076, "step": 46810 }, { "epoch": 0.91, "learning_rate": 3.486210042117113e-05, "loss": 0.0275, "step": 46812 }, { "epoch": 0.91, "learning_rate": 3.4861453460913896e-05, "loss": 0.0, "step": 46814 }, { "epoch": 0.91, "learning_rate": 3.4860806500656665e-05, "loss": 0.0, "step": 46816 }, { "epoch": 0.91, "learning_rate": 3.4860159540399434e-05, "loss": 0.0, "step": 46818 }, { "epoch": 0.91, "learning_rate": 3.48595125801422e-05, "loss": 0.0252, "step": 46820 }, { "epoch": 0.91, "learning_rate": 3.485886561988497e-05, "loss": 0.0, "step": 46822 }, { "epoch": 0.91, "learning_rate": 3.485821865962774e-05, "loss": 0.0, "step": 46824 }, { "epoch": 0.91, "learning_rate": 3.485757169937051e-05, "loss": 0.0, "step": 46826 }, { "epoch": 0.91, "learning_rate": 3.485692473911328e-05, "loss": 0.0001, "step": 46828 }, { "epoch": 0.91, "learning_rate": 3.485627777885604e-05, "loss": 0.0, "step": 46830 }, { "epoch": 0.91, "learning_rate": 3.485563081859882e-05, "loss": 0.0015, "step": 46832 }, { "epoch": 0.91, "learning_rate": 3.485498385834158e-05, "loss": 0.0, "step": 46834 }, { "epoch": 0.91, "learning_rate": 3.485433689808435e-05, "loss": 0.0, "step": 46836 }, { "epoch": 0.91, "learning_rate": 3.4853689937827125e-05, "loss": 0.0, "step": 46838 }, { "epoch": 0.91, "learning_rate": 3.485304297756989e-05, "loss": 0.0, "step": 46840 }, { "epoch": 0.91, "learning_rate": 3.485239601731266e-05, "loss": 0.0, "step": 46842 }, { "epoch": 0.91, "learning_rate": 3.4851749057055426e-05, "loss": 0.0001, "step": 46844 }, { "epoch": 0.91, "learning_rate": 3.4851102096798195e-05, "loss": 0.0011, "step": 46846 }, { "epoch": 0.91, "learning_rate": 3.4850455136540964e-05, "loss": 0.0, "step": 46848 }, { "epoch": 0.91, "learning_rate": 3.484980817628373e-05, "loss": 0.0001, "step": 46850 }, { "epoch": 0.91, "learning_rate": 3.48491612160265e-05, "loss": 0.0055, "step": 46852 }, { "epoch": 0.91, "learning_rate": 3.484851425576927e-05, "loss": 0.0001, "step": 46854 }, { "epoch": 0.91, "learning_rate": 3.484786729551204e-05, "loss": 0.0, "step": 46856 }, { "epoch": 0.91, "learning_rate": 3.48472203352548e-05, "loss": 0.0018, "step": 46858 }, { "epoch": 0.91, "learning_rate": 3.484657337499758e-05, "loss": 0.0, "step": 46860 }, { "epoch": 0.91, "learning_rate": 3.484592641474034e-05, "loss": 0.0007, "step": 46862 }, { "epoch": 0.91, "learning_rate": 3.4845279454483116e-05, "loss": 0.0, "step": 46864 }, { "epoch": 0.91, "learning_rate": 3.484463249422588e-05, "loss": 0.0, "step": 46866 }, { "epoch": 0.91, "learning_rate": 3.484398553396865e-05, "loss": 0.0001, "step": 46868 }, { "epoch": 0.91, "learning_rate": 3.4843338573711424e-05, "loss": 0.0017, "step": 46870 }, { "epoch": 0.91, "learning_rate": 3.4842691613454186e-05, "loss": 0.0, "step": 46872 }, { "epoch": 0.91, "learning_rate": 3.4842044653196955e-05, "loss": 0.0066, "step": 46874 }, { "epoch": 0.91, "learning_rate": 3.4841397692939724e-05, "loss": 0.0048, "step": 46876 }, { "epoch": 0.91, "learning_rate": 3.484075073268249e-05, "loss": 0.0002, "step": 46878 }, { "epoch": 0.91, "learning_rate": 3.484010377242526e-05, "loss": 0.0, "step": 46880 }, { "epoch": 0.91, "learning_rate": 3.483945681216803e-05, "loss": 0.0008, "step": 46882 }, { "epoch": 0.91, "learning_rate": 3.48388098519108e-05, "loss": 0.0, "step": 46884 }, { "epoch": 0.91, "learning_rate": 3.483816289165357e-05, "loss": 0.0027, "step": 46886 }, { "epoch": 0.91, "learning_rate": 3.483751593139634e-05, "loss": 0.0, "step": 46888 }, { "epoch": 0.91, "learning_rate": 3.48368689711391e-05, "loss": 0.0, "step": 46890 }, { "epoch": 0.91, "learning_rate": 3.483622201088188e-05, "loss": 0.0, "step": 46892 }, { "epoch": 0.91, "learning_rate": 3.483557505062464e-05, "loss": 0.0001, "step": 46894 }, { "epoch": 0.91, "learning_rate": 3.483492809036741e-05, "loss": 0.0, "step": 46896 }, { "epoch": 0.91, "learning_rate": 3.483428113011018e-05, "loss": 0.0001, "step": 46898 }, { "epoch": 0.91, "learning_rate": 3.4833634169852946e-05, "loss": 0.0, "step": 46900 }, { "epoch": 0.91, "learning_rate": 3.4832987209595715e-05, "loss": 0.0008, "step": 46902 }, { "epoch": 0.91, "learning_rate": 3.4832340249338485e-05, "loss": 0.0, "step": 46904 }, { "epoch": 0.91, "learning_rate": 3.4831693289081254e-05, "loss": 0.0, "step": 46906 }, { "epoch": 0.91, "learning_rate": 3.483104632882402e-05, "loss": 0.0, "step": 46908 }, { "epoch": 0.91, "learning_rate": 3.483039936856679e-05, "loss": 0.0001, "step": 46910 }, { "epoch": 0.91, "learning_rate": 3.4829752408309554e-05, "loss": 0.0, "step": 46912 }, { "epoch": 0.91, "learning_rate": 3.482910544805233e-05, "loss": 0.0, "step": 46914 }, { "epoch": 0.91, "learning_rate": 3.48284584877951e-05, "loss": 0.0, "step": 46916 }, { "epoch": 0.91, "learning_rate": 3.482781152753786e-05, "loss": 0.0, "step": 46918 }, { "epoch": 0.91, "learning_rate": 3.482716456728064e-05, "loss": 0.0003, "step": 46920 }, { "epoch": 0.91, "learning_rate": 3.48265176070234e-05, "loss": 0.0, "step": 46922 }, { "epoch": 0.91, "learning_rate": 3.4825870646766175e-05, "loss": 0.0, "step": 46924 }, { "epoch": 0.91, "learning_rate": 3.482522368650894e-05, "loss": 0.0, "step": 46926 }, { "epoch": 0.91, "learning_rate": 3.482457672625171e-05, "loss": 0.0001, "step": 46928 }, { "epoch": 0.91, "learning_rate": 3.4823929765994476e-05, "loss": 0.0, "step": 46930 }, { "epoch": 0.91, "learning_rate": 3.4823282805737245e-05, "loss": 0.002, "step": 46932 }, { "epoch": 0.91, "learning_rate": 3.4822635845480014e-05, "loss": 0.0, "step": 46934 }, { "epoch": 0.91, "learning_rate": 3.482198888522278e-05, "loss": 0.0, "step": 46936 }, { "epoch": 0.91, "learning_rate": 3.482134192496555e-05, "loss": 0.0, "step": 46938 }, { "epoch": 0.91, "learning_rate": 3.4820694964708314e-05, "loss": 0.0, "step": 46940 }, { "epoch": 0.91, "learning_rate": 3.482004800445109e-05, "loss": 0.0, "step": 46942 }, { "epoch": 0.91, "learning_rate": 3.481940104419385e-05, "loss": 0.0, "step": 46944 }, { "epoch": 0.91, "learning_rate": 3.481875408393663e-05, "loss": 0.0, "step": 46946 }, { "epoch": 0.91, "learning_rate": 3.48181071236794e-05, "loss": 0.0001, "step": 46948 }, { "epoch": 0.91, "learning_rate": 3.481746016342216e-05, "loss": 0.0, "step": 46950 }, { "epoch": 0.91, "learning_rate": 3.4816813203164936e-05, "loss": 0.0001, "step": 46952 }, { "epoch": 0.91, "learning_rate": 3.48161662429077e-05, "loss": 0.0, "step": 46954 }, { "epoch": 0.91, "learning_rate": 3.481551928265047e-05, "loss": 0.0, "step": 46956 }, { "epoch": 0.91, "learning_rate": 3.4814872322393236e-05, "loss": 0.0, "step": 46958 }, { "epoch": 0.91, "learning_rate": 3.4814225362136005e-05, "loss": 0.0002, "step": 46960 }, { "epoch": 0.91, "learning_rate": 3.4813578401878774e-05, "loss": 0.0001, "step": 46962 }, { "epoch": 0.91, "learning_rate": 3.4812931441621544e-05, "loss": 0.0003, "step": 46964 }, { "epoch": 0.91, "learning_rate": 3.481228448136431e-05, "loss": 0.0022, "step": 46966 }, { "epoch": 0.91, "learning_rate": 3.481163752110708e-05, "loss": 0.0, "step": 46968 }, { "epoch": 0.91, "learning_rate": 3.481099056084985e-05, "loss": 0.0, "step": 46970 }, { "epoch": 0.91, "learning_rate": 3.481034360059261e-05, "loss": 0.0, "step": 46972 }, { "epoch": 0.91, "learning_rate": 3.480969664033539e-05, "loss": 0.0027, "step": 46974 }, { "epoch": 0.91, "learning_rate": 3.480904968007815e-05, "loss": 0.0, "step": 46976 }, { "epoch": 0.91, "learning_rate": 3.480840271982092e-05, "loss": 0.0, "step": 46978 }, { "epoch": 0.91, "learning_rate": 3.480775575956369e-05, "loss": 0.0008, "step": 46980 }, { "epoch": 0.91, "learning_rate": 3.480710879930646e-05, "loss": 0.01, "step": 46982 }, { "epoch": 0.91, "learning_rate": 3.4806461839049234e-05, "loss": 0.017, "step": 46984 }, { "epoch": 0.91, "learning_rate": 3.4805814878792e-05, "loss": 0.0081, "step": 46986 }, { "epoch": 0.91, "learning_rate": 3.4805167918534766e-05, "loss": 0.0001, "step": 46988 }, { "epoch": 0.91, "learning_rate": 3.4804520958277535e-05, "loss": 0.0, "step": 46990 }, { "epoch": 0.91, "learning_rate": 3.4803873998020304e-05, "loss": 0.0001, "step": 46992 }, { "epoch": 0.91, "learning_rate": 3.480322703776307e-05, "loss": 0.0, "step": 46994 }, { "epoch": 0.91, "learning_rate": 3.480258007750584e-05, "loss": 0.0255, "step": 46996 }, { "epoch": 0.91, "learning_rate": 3.480193311724861e-05, "loss": 0.0, "step": 46998 }, { "epoch": 0.91, "learning_rate": 3.4801286156991373e-05, "loss": 0.0, "step": 47000 }, { "epoch": 0.91, "learning_rate": 3.480063919673415e-05, "loss": 0.0001, "step": 47002 }, { "epoch": 0.91, "learning_rate": 3.479999223647691e-05, "loss": 0.0014, "step": 47004 }, { "epoch": 0.91, "learning_rate": 3.479934527621969e-05, "loss": 0.0005, "step": 47006 }, { "epoch": 0.91, "learning_rate": 3.479869831596245e-05, "loss": 0.0, "step": 47008 }, { "epoch": 0.91, "learning_rate": 3.479805135570522e-05, "loss": 0.0001, "step": 47010 }, { "epoch": 0.91, "learning_rate": 3.479740439544799e-05, "loss": 0.0, "step": 47012 }, { "epoch": 0.91, "learning_rate": 3.479675743519076e-05, "loss": 0.0001, "step": 47014 }, { "epoch": 0.91, "learning_rate": 3.4796110474933526e-05, "loss": 0.0, "step": 47016 }, { "epoch": 0.91, "learning_rate": 3.4795463514676295e-05, "loss": 0.0, "step": 47018 }, { "epoch": 0.91, "learning_rate": 3.4794816554419064e-05, "loss": 0.0001, "step": 47020 }, { "epoch": 0.91, "learning_rate": 3.4794169594161833e-05, "loss": 0.0027, "step": 47022 }, { "epoch": 0.91, "learning_rate": 3.47935226339046e-05, "loss": 0.0, "step": 47024 }, { "epoch": 0.91, "learning_rate": 3.479287567364737e-05, "loss": 0.0001, "step": 47026 }, { "epoch": 0.91, "learning_rate": 3.479222871339014e-05, "loss": 0.0056, "step": 47028 }, { "epoch": 0.91, "learning_rate": 3.479158175313291e-05, "loss": 0.0, "step": 47030 }, { "epoch": 0.91, "learning_rate": 3.479093479287567e-05, "loss": 0.0001, "step": 47032 }, { "epoch": 0.91, "learning_rate": 3.479028783261845e-05, "loss": 0.0, "step": 47034 }, { "epoch": 0.91, "learning_rate": 3.478964087236121e-05, "loss": 0.0, "step": 47036 }, { "epoch": 0.91, "learning_rate": 3.478899391210398e-05, "loss": 0.0001, "step": 47038 }, { "epoch": 0.91, "learning_rate": 3.478834695184675e-05, "loss": 0.0004, "step": 47040 }, { "epoch": 0.91, "learning_rate": 3.478769999158952e-05, "loss": 0.0, "step": 47042 }, { "epoch": 0.91, "learning_rate": 3.4787053031332287e-05, "loss": 0.0, "step": 47044 }, { "epoch": 0.91, "learning_rate": 3.4786406071075056e-05, "loss": 0.0, "step": 47046 }, { "epoch": 0.91, "learning_rate": 3.4785759110817825e-05, "loss": 0.0001, "step": 47048 }, { "epoch": 0.91, "learning_rate": 3.4785112150560594e-05, "loss": 0.0005, "step": 47050 }, { "epoch": 0.91, "learning_rate": 3.478446519030336e-05, "loss": 0.0, "step": 47052 }, { "epoch": 0.91, "learning_rate": 3.4783818230046125e-05, "loss": 0.0, "step": 47054 }, { "epoch": 0.91, "learning_rate": 3.47831712697889e-05, "loss": 0.0, "step": 47056 }, { "epoch": 0.91, "learning_rate": 3.478252430953166e-05, "loss": 0.0095, "step": 47058 }, { "epoch": 0.91, "learning_rate": 3.478187734927443e-05, "loss": 0.0, "step": 47060 }, { "epoch": 0.91, "learning_rate": 3.478123038901721e-05, "loss": 0.0001, "step": 47062 }, { "epoch": 0.91, "learning_rate": 3.478058342875997e-05, "loss": 0.0011, "step": 47064 }, { "epoch": 0.91, "learning_rate": 3.4779936468502746e-05, "loss": 0.0, "step": 47066 }, { "epoch": 0.91, "learning_rate": 3.477928950824551e-05, "loss": 0.0, "step": 47068 }, { "epoch": 0.91, "learning_rate": 3.477864254798828e-05, "loss": 0.0, "step": 47070 }, { "epoch": 0.91, "learning_rate": 3.477799558773105e-05, "loss": 0.0223, "step": 47072 }, { "epoch": 0.91, "learning_rate": 3.4777348627473816e-05, "loss": 0.0, "step": 47074 }, { "epoch": 0.91, "learning_rate": 3.4776701667216585e-05, "loss": 0.0, "step": 47076 }, { "epoch": 0.91, "learning_rate": 3.4776054706959354e-05, "loss": 0.0001, "step": 47078 }, { "epoch": 0.91, "learning_rate": 3.477540774670212e-05, "loss": 0.0, "step": 47080 }, { "epoch": 0.91, "learning_rate": 3.4774760786444886e-05, "loss": 0.0001, "step": 47082 }, { "epoch": 0.91, "learning_rate": 3.477411382618766e-05, "loss": 0.0, "step": 47084 }, { "epoch": 0.91, "learning_rate": 3.4773466865930424e-05, "loss": 0.0, "step": 47086 }, { "epoch": 0.91, "learning_rate": 3.47728199056732e-05, "loss": 0.0015, "step": 47088 }, { "epoch": 0.91, "learning_rate": 3.477217294541596e-05, "loss": 0.0001, "step": 47090 }, { "epoch": 0.91, "learning_rate": 3.477152598515873e-05, "loss": 0.0, "step": 47092 }, { "epoch": 0.91, "learning_rate": 3.477087902490151e-05, "loss": 0.0, "step": 47094 }, { "epoch": 0.91, "learning_rate": 3.477023206464427e-05, "loss": 0.0, "step": 47096 }, { "epoch": 0.91, "learning_rate": 3.476958510438704e-05, "loss": 0.0, "step": 47098 }, { "epoch": 0.91, "learning_rate": 3.476893814412981e-05, "loss": 0.0, "step": 47100 }, { "epoch": 0.91, "learning_rate": 3.4768291183872576e-05, "loss": 0.0002, "step": 47102 }, { "epoch": 0.91, "learning_rate": 3.4767644223615346e-05, "loss": 0.0, "step": 47104 }, { "epoch": 0.91, "learning_rate": 3.4766997263358115e-05, "loss": 0.0, "step": 47106 }, { "epoch": 0.91, "learning_rate": 3.4766350303100884e-05, "loss": 0.0, "step": 47108 }, { "epoch": 0.91, "learning_rate": 3.476570334284365e-05, "loss": 0.0, "step": 47110 }, { "epoch": 0.91, "learning_rate": 3.476505638258642e-05, "loss": 0.0, "step": 47112 }, { "epoch": 0.91, "learning_rate": 3.4764409422329184e-05, "loss": 0.0, "step": 47114 }, { "epoch": 0.91, "learning_rate": 3.476376246207196e-05, "loss": 0.0001, "step": 47116 }, { "epoch": 0.91, "learning_rate": 3.476311550181472e-05, "loss": 0.0, "step": 47118 }, { "epoch": 0.91, "learning_rate": 3.476246854155749e-05, "loss": 0.0001, "step": 47120 }, { "epoch": 0.91, "learning_rate": 3.476182158130026e-05, "loss": 0.0182, "step": 47122 }, { "epoch": 0.91, "learning_rate": 3.476117462104303e-05, "loss": 0.0, "step": 47124 }, { "epoch": 0.91, "learning_rate": 3.4760527660785805e-05, "loss": 0.0, "step": 47126 }, { "epoch": 0.91, "learning_rate": 3.475988070052857e-05, "loss": 0.0, "step": 47128 }, { "epoch": 0.91, "learning_rate": 3.475923374027134e-05, "loss": 0.0, "step": 47130 }, { "epoch": 0.91, "learning_rate": 3.4758586780014106e-05, "loss": 0.0016, "step": 47132 }, { "epoch": 0.91, "learning_rate": 3.4757939819756875e-05, "loss": 0.0, "step": 47134 }, { "epoch": 0.91, "learning_rate": 3.475729285949964e-05, "loss": 0.0001, "step": 47136 }, { "epoch": 0.91, "learning_rate": 3.475664589924241e-05, "loss": 0.0, "step": 47138 }, { "epoch": 0.91, "learning_rate": 3.475599893898518e-05, "loss": 0.0001, "step": 47140 }, { "epoch": 0.91, "learning_rate": 3.4755351978727945e-05, "loss": 0.0003, "step": 47142 }, { "epoch": 0.92, "learning_rate": 3.475470501847072e-05, "loss": 0.0, "step": 47144 }, { "epoch": 0.92, "learning_rate": 3.475405805821348e-05, "loss": 0.0, "step": 47146 }, { "epoch": 0.92, "learning_rate": 3.475341109795626e-05, "loss": 0.0, "step": 47148 }, { "epoch": 0.92, "learning_rate": 3.475276413769902e-05, "loss": 0.0, "step": 47150 }, { "epoch": 0.92, "learning_rate": 3.475211717744179e-05, "loss": 0.0, "step": 47152 }, { "epoch": 0.92, "learning_rate": 3.475147021718456e-05, "loss": 0.0, "step": 47154 }, { "epoch": 0.92, "learning_rate": 3.475082325692733e-05, "loss": 0.0, "step": 47156 }, { "epoch": 0.92, "learning_rate": 3.47501762966701e-05, "loss": 0.0001, "step": 47158 }, { "epoch": 0.92, "learning_rate": 3.4749529336412866e-05, "loss": 0.0, "step": 47160 }, { "epoch": 0.92, "learning_rate": 3.4748882376155635e-05, "loss": 0.0, "step": 47162 }, { "epoch": 0.92, "learning_rate": 3.47482354158984e-05, "loss": 0.0, "step": 47164 }, { "epoch": 0.92, "learning_rate": 3.4747588455641174e-05, "loss": 0.0, "step": 47166 }, { "epoch": 0.92, "learning_rate": 3.4746941495383936e-05, "loss": 0.0001, "step": 47168 }, { "epoch": 0.92, "learning_rate": 3.474629453512671e-05, "loss": 0.0, "step": 47170 }, { "epoch": 0.92, "learning_rate": 3.474564757486948e-05, "loss": 0.0096, "step": 47172 }, { "epoch": 0.92, "learning_rate": 3.474500061461224e-05, "loss": 0.0, "step": 47174 }, { "epoch": 0.92, "learning_rate": 3.474435365435502e-05, "loss": 0.0002, "step": 47176 }, { "epoch": 0.92, "learning_rate": 3.474370669409778e-05, "loss": 0.0058, "step": 47178 }, { "epoch": 0.92, "learning_rate": 3.474305973384055e-05, "loss": 0.0, "step": 47180 }, { "epoch": 0.92, "learning_rate": 3.474241277358332e-05, "loss": 0.0001, "step": 47182 }, { "epoch": 0.92, "learning_rate": 3.474176581332609e-05, "loss": 0.0013, "step": 47184 }, { "epoch": 0.92, "learning_rate": 3.474111885306886e-05, "loss": 0.0, "step": 47186 }, { "epoch": 0.92, "learning_rate": 3.474047189281163e-05, "loss": 0.0006, "step": 47188 }, { "epoch": 0.92, "learning_rate": 3.4739824932554396e-05, "loss": 0.0, "step": 47190 }, { "epoch": 0.92, "learning_rate": 3.4739177972297165e-05, "loss": 0.0, "step": 47192 }, { "epoch": 0.92, "learning_rate": 3.4738531012039934e-05, "loss": 0.0, "step": 47194 }, { "epoch": 0.92, "learning_rate": 3.4737884051782696e-05, "loss": 0.001, "step": 47196 }, { "epoch": 0.92, "learning_rate": 3.473723709152547e-05, "loss": 0.009, "step": 47198 }, { "epoch": 0.92, "learning_rate": 3.4736590131268234e-05, "loss": 0.0002, "step": 47200 }, { "epoch": 0.92, "learning_rate": 3.4735943171011004e-05, "loss": 0.0, "step": 47202 }, { "epoch": 0.92, "learning_rate": 3.473529621075378e-05, "loss": 0.0842, "step": 47204 }, { "epoch": 0.92, "learning_rate": 3.473464925049654e-05, "loss": 0.0, "step": 47206 }, { "epoch": 0.92, "learning_rate": 3.473400229023932e-05, "loss": 0.0, "step": 47208 }, { "epoch": 0.92, "learning_rate": 3.473335532998208e-05, "loss": 0.0106, "step": 47210 }, { "epoch": 0.92, "learning_rate": 3.473270836972485e-05, "loss": 0.0, "step": 47212 }, { "epoch": 0.92, "learning_rate": 3.473206140946762e-05, "loss": 0.0, "step": 47214 }, { "epoch": 0.92, "learning_rate": 3.473141444921039e-05, "loss": 0.0, "step": 47216 }, { "epoch": 0.92, "learning_rate": 3.4730767488953156e-05, "loss": 0.0, "step": 47218 }, { "epoch": 0.92, "learning_rate": 3.4730120528695925e-05, "loss": 0.0161, "step": 47220 }, { "epoch": 0.92, "learning_rate": 3.4729473568438694e-05, "loss": 0.0, "step": 47222 }, { "epoch": 0.92, "learning_rate": 3.472882660818146e-05, "loss": 0.0, "step": 47224 }, { "epoch": 0.92, "learning_rate": 3.472817964792423e-05, "loss": 0.0001, "step": 47226 }, { "epoch": 0.92, "learning_rate": 3.4727532687666995e-05, "loss": 0.0001, "step": 47228 }, { "epoch": 0.92, "learning_rate": 3.472688572740977e-05, "loss": 0.0078, "step": 47230 }, { "epoch": 0.92, "learning_rate": 3.472623876715253e-05, "loss": 0.0, "step": 47232 }, { "epoch": 0.92, "learning_rate": 3.47255918068953e-05, "loss": 0.0, "step": 47234 }, { "epoch": 0.92, "learning_rate": 3.472494484663807e-05, "loss": 0.0, "step": 47236 }, { "epoch": 0.92, "learning_rate": 3.472429788638084e-05, "loss": 0.0, "step": 47238 }, { "epoch": 0.92, "learning_rate": 3.472365092612361e-05, "loss": 0.0, "step": 47240 }, { "epoch": 0.92, "learning_rate": 3.472300396586638e-05, "loss": 0.0, "step": 47242 }, { "epoch": 0.92, "learning_rate": 3.472235700560915e-05, "loss": 0.0, "step": 47244 }, { "epoch": 0.92, "learning_rate": 3.472171004535192e-05, "loss": 0.0, "step": 47246 }, { "epoch": 0.92, "learning_rate": 3.4721063085094686e-05, "loss": 0.0, "step": 47248 }, { "epoch": 0.92, "learning_rate": 3.4720416124837455e-05, "loss": 0.0002, "step": 47250 }, { "epoch": 0.92, "learning_rate": 3.4719769164580224e-05, "loss": 0.0, "step": 47252 }, { "epoch": 0.92, "learning_rate": 3.471912220432299e-05, "loss": 0.0001, "step": 47254 }, { "epoch": 0.92, "learning_rate": 3.4718475244065755e-05, "loss": 0.0048, "step": 47256 }, { "epoch": 0.92, "learning_rate": 3.471782828380853e-05, "loss": 0.0, "step": 47258 }, { "epoch": 0.92, "learning_rate": 3.4717181323551293e-05, "loss": 0.0, "step": 47260 }, { "epoch": 0.92, "learning_rate": 3.471653436329406e-05, "loss": 0.0056, "step": 47262 }, { "epoch": 0.92, "learning_rate": 3.471588740303683e-05, "loss": 0.0, "step": 47264 }, { "epoch": 0.92, "learning_rate": 3.47152404427796e-05, "loss": 0.0, "step": 47266 }, { "epoch": 0.92, "learning_rate": 3.471459348252237e-05, "loss": 0.0, "step": 47268 }, { "epoch": 0.92, "learning_rate": 3.471394652226514e-05, "loss": 0.0, "step": 47270 }, { "epoch": 0.92, "learning_rate": 3.471329956200791e-05, "loss": 0.0, "step": 47272 }, { "epoch": 0.92, "learning_rate": 3.471265260175068e-05, "loss": 0.0001, "step": 47274 }, { "epoch": 0.92, "learning_rate": 3.4712005641493446e-05, "loss": 0.0, "step": 47276 }, { "epoch": 0.92, "learning_rate": 3.471135868123621e-05, "loss": 0.0002, "step": 47278 }, { "epoch": 0.92, "learning_rate": 3.4710711720978984e-05, "loss": 0.0, "step": 47280 }, { "epoch": 0.92, "learning_rate": 3.4710064760721747e-05, "loss": 0.0, "step": 47282 }, { "epoch": 0.92, "learning_rate": 3.4709417800464516e-05, "loss": 0.0001, "step": 47284 }, { "epoch": 0.92, "learning_rate": 3.470877084020729e-05, "loss": 0.0079, "step": 47286 }, { "epoch": 0.92, "learning_rate": 3.4708123879950054e-05, "loss": 0.0, "step": 47288 }, { "epoch": 0.92, "learning_rate": 3.470747691969283e-05, "loss": 0.0157, "step": 47290 }, { "epoch": 0.92, "learning_rate": 3.470682995943559e-05, "loss": 0.0, "step": 47292 }, { "epoch": 0.92, "learning_rate": 3.470618299917836e-05, "loss": 0.0, "step": 47294 }, { "epoch": 0.92, "learning_rate": 3.470553603892113e-05, "loss": 0.0, "step": 47296 }, { "epoch": 0.92, "learning_rate": 3.47048890786639e-05, "loss": 0.0001, "step": 47298 }, { "epoch": 0.92, "learning_rate": 3.470424211840667e-05, "loss": 0.0011, "step": 47300 }, { "epoch": 0.92, "learning_rate": 3.470359515814944e-05, "loss": 0.0, "step": 47302 }, { "epoch": 0.92, "learning_rate": 3.4702948197892207e-05, "loss": 0.0092, "step": 47304 }, { "epoch": 0.92, "learning_rate": 3.470230123763497e-05, "loss": 0.0053, "step": 47306 }, { "epoch": 0.92, "learning_rate": 3.4701654277377745e-05, "loss": 0.0, "step": 47308 }, { "epoch": 0.92, "learning_rate": 3.470100731712051e-05, "loss": 0.0014, "step": 47310 }, { "epoch": 0.92, "learning_rate": 3.470036035686328e-05, "loss": 0.0001, "step": 47312 }, { "epoch": 0.92, "learning_rate": 3.4699713396606045e-05, "loss": 0.0, "step": 47314 }, { "epoch": 0.92, "learning_rate": 3.4699066436348814e-05, "loss": 0.0001, "step": 47316 }, { "epoch": 0.92, "learning_rate": 3.469841947609159e-05, "loss": 0.0, "step": 47318 }, { "epoch": 0.92, "learning_rate": 3.469777251583435e-05, "loss": 0.0002, "step": 47320 }, { "epoch": 0.92, "learning_rate": 3.469712555557712e-05, "loss": 0.0, "step": 47322 }, { "epoch": 0.92, "learning_rate": 3.469647859531989e-05, "loss": 0.0077, "step": 47324 }, { "epoch": 0.92, "learning_rate": 3.469583163506266e-05, "loss": 0.0002, "step": 47326 }, { "epoch": 0.92, "learning_rate": 3.469518467480543e-05, "loss": 0.0001, "step": 47328 }, { "epoch": 0.92, "learning_rate": 3.46945377145482e-05, "loss": 0.0001, "step": 47330 }, { "epoch": 0.92, "learning_rate": 3.469389075429097e-05, "loss": 0.0001, "step": 47332 }, { "epoch": 0.92, "learning_rate": 3.4693243794033736e-05, "loss": 0.0, "step": 47334 }, { "epoch": 0.92, "learning_rate": 3.4692596833776505e-05, "loss": 0.0087, "step": 47336 }, { "epoch": 0.92, "learning_rate": 3.469194987351927e-05, "loss": 0.0001, "step": 47338 }, { "epoch": 0.92, "learning_rate": 3.469130291326204e-05, "loss": 0.0, "step": 47340 }, { "epoch": 0.92, "learning_rate": 3.4690655953004806e-05, "loss": 0.0014, "step": 47342 }, { "epoch": 0.92, "learning_rate": 3.4690008992747575e-05, "loss": 0.0, "step": 47344 }, { "epoch": 0.92, "learning_rate": 3.4689362032490344e-05, "loss": 0.0, "step": 47346 }, { "epoch": 0.92, "learning_rate": 3.468871507223311e-05, "loss": 0.0813, "step": 47348 }, { "epoch": 0.92, "learning_rate": 3.468806811197589e-05, "loss": 0.0, "step": 47350 }, { "epoch": 0.92, "learning_rate": 3.468742115171865e-05, "loss": 0.0, "step": 47352 }, { "epoch": 0.92, "learning_rate": 3.468677419146142e-05, "loss": 0.0002, "step": 47354 }, { "epoch": 0.92, "learning_rate": 3.468612723120419e-05, "loss": 0.0001, "step": 47356 }, { "epoch": 0.92, "learning_rate": 3.468548027094696e-05, "loss": 0.0042, "step": 47358 }, { "epoch": 0.92, "learning_rate": 3.468483331068972e-05, "loss": 0.0, "step": 47360 }, { "epoch": 0.92, "learning_rate": 3.4684186350432496e-05, "loss": 0.0, "step": 47362 }, { "epoch": 0.92, "learning_rate": 3.4683539390175266e-05, "loss": 0.0, "step": 47364 }, { "epoch": 0.92, "learning_rate": 3.468289242991803e-05, "loss": 0.0, "step": 47366 }, { "epoch": 0.92, "learning_rate": 3.4682245469660804e-05, "loss": 0.0, "step": 47368 }, { "epoch": 0.92, "learning_rate": 3.4681598509403566e-05, "loss": 0.0, "step": 47370 }, { "epoch": 0.92, "learning_rate": 3.468095154914634e-05, "loss": 0.0, "step": 47372 }, { "epoch": 0.92, "learning_rate": 3.4680304588889104e-05, "loss": 0.0, "step": 47374 }, { "epoch": 0.92, "learning_rate": 3.467965762863187e-05, "loss": 0.0, "step": 47376 }, { "epoch": 0.92, "learning_rate": 3.467901066837464e-05, "loss": 0.0089, "step": 47378 }, { "epoch": 0.92, "learning_rate": 3.467836370811741e-05, "loss": 0.0, "step": 47380 }, { "epoch": 0.92, "learning_rate": 3.467771674786018e-05, "loss": 0.0001, "step": 47382 }, { "epoch": 0.92, "learning_rate": 3.467706978760295e-05, "loss": 0.0, "step": 47384 }, { "epoch": 0.92, "learning_rate": 3.467642282734572e-05, "loss": 0.0003, "step": 47386 }, { "epoch": 0.92, "learning_rate": 3.467577586708849e-05, "loss": 0.0, "step": 47388 }, { "epoch": 0.92, "learning_rate": 3.467512890683126e-05, "loss": 0.0001, "step": 47390 }, { "epoch": 0.92, "learning_rate": 3.467448194657402e-05, "loss": 0.0003, "step": 47392 }, { "epoch": 0.92, "learning_rate": 3.4673834986316795e-05, "loss": 0.0024, "step": 47394 }, { "epoch": 0.92, "learning_rate": 3.4673188026059564e-05, "loss": 0.0, "step": 47396 }, { "epoch": 0.92, "learning_rate": 3.4672541065802326e-05, "loss": 0.0, "step": 47398 }, { "epoch": 0.92, "learning_rate": 3.46718941055451e-05, "loss": 0.0, "step": 47400 }, { "epoch": 0.92, "learning_rate": 3.4671247145287865e-05, "loss": 0.0, "step": 47402 }, { "epoch": 0.92, "learning_rate": 3.4670600185030634e-05, "loss": 0.0, "step": 47404 }, { "epoch": 0.92, "learning_rate": 3.46699532247734e-05, "loss": 0.0, "step": 47406 }, { "epoch": 0.92, "learning_rate": 3.466930626451617e-05, "loss": 0.0, "step": 47408 }, { "epoch": 0.92, "learning_rate": 3.466865930425894e-05, "loss": 0.0, "step": 47410 }, { "epoch": 0.92, "learning_rate": 3.466801234400171e-05, "loss": 0.0001, "step": 47412 }, { "epoch": 0.92, "learning_rate": 3.466736538374448e-05, "loss": 0.0036, "step": 47414 }, { "epoch": 0.92, "learning_rate": 3.466671842348725e-05, "loss": 0.0, "step": 47416 }, { "epoch": 0.92, "learning_rate": 3.466607146323002e-05, "loss": 0.0, "step": 47418 }, { "epoch": 0.92, "learning_rate": 3.466542450297278e-05, "loss": 0.0008, "step": 47420 }, { "epoch": 0.92, "learning_rate": 3.4664777542715555e-05, "loss": 0.0, "step": 47422 }, { "epoch": 0.92, "learning_rate": 3.466413058245832e-05, "loss": 0.0, "step": 47424 }, { "epoch": 0.92, "learning_rate": 3.466348362220109e-05, "loss": 0.0003, "step": 47426 }, { "epoch": 0.92, "learning_rate": 3.466283666194386e-05, "loss": 0.0005, "step": 47428 }, { "epoch": 0.92, "learning_rate": 3.4662189701686625e-05, "loss": 0.0, "step": 47430 }, { "epoch": 0.92, "learning_rate": 3.46615427414294e-05, "loss": 0.0002, "step": 47432 }, { "epoch": 0.92, "learning_rate": 3.466089578117216e-05, "loss": 0.0003, "step": 47434 }, { "epoch": 0.92, "learning_rate": 3.466024882091493e-05, "loss": 0.0024, "step": 47436 }, { "epoch": 0.92, "learning_rate": 3.46596018606577e-05, "loss": 0.0, "step": 47438 }, { "epoch": 0.92, "learning_rate": 3.465895490040047e-05, "loss": 0.0092, "step": 47440 }, { "epoch": 0.92, "learning_rate": 3.465830794014324e-05, "loss": 0.0, "step": 47442 }, { "epoch": 0.92, "learning_rate": 3.465766097988601e-05, "loss": 0.0, "step": 47444 }, { "epoch": 0.92, "learning_rate": 3.465701401962878e-05, "loss": 0.0, "step": 47446 }, { "epoch": 0.92, "learning_rate": 3.465636705937154e-05, "loss": 0.0, "step": 47448 }, { "epoch": 0.92, "learning_rate": 3.4655720099114316e-05, "loss": 0.0, "step": 47450 }, { "epoch": 0.92, "learning_rate": 3.465507313885708e-05, "loss": 0.0, "step": 47452 }, { "epoch": 0.92, "learning_rate": 3.4654426178599854e-05, "loss": 0.0183, "step": 47454 }, { "epoch": 0.92, "learning_rate": 3.4653779218342616e-05, "loss": 0.0, "step": 47456 }, { "epoch": 0.92, "learning_rate": 3.4653132258085385e-05, "loss": 0.0, "step": 47458 }, { "epoch": 0.92, "learning_rate": 3.4652485297828154e-05, "loss": 0.0013, "step": 47460 }, { "epoch": 0.92, "learning_rate": 3.4651838337570924e-05, "loss": 0.0, "step": 47462 }, { "epoch": 0.92, "learning_rate": 3.465119137731369e-05, "loss": 0.0, "step": 47464 }, { "epoch": 0.92, "learning_rate": 3.465054441705646e-05, "loss": 0.0002, "step": 47466 }, { "epoch": 0.92, "learning_rate": 3.464989745679923e-05, "loss": 0.0001, "step": 47468 }, { "epoch": 0.92, "learning_rate": 3.4649250496542e-05, "loss": 0.0049, "step": 47470 }, { "epoch": 0.92, "learning_rate": 3.464860353628477e-05, "loss": 0.0, "step": 47472 }, { "epoch": 0.92, "learning_rate": 3.464795657602754e-05, "loss": 0.0, "step": 47474 }, { "epoch": 0.92, "learning_rate": 3.464730961577031e-05, "loss": 0.0, "step": 47476 }, { "epoch": 0.92, "learning_rate": 3.4646662655513076e-05, "loss": 0.0256, "step": 47478 }, { "epoch": 0.92, "learning_rate": 3.464601569525584e-05, "loss": 0.004, "step": 47480 }, { "epoch": 0.92, "learning_rate": 3.4645368734998614e-05, "loss": 0.0, "step": 47482 }, { "epoch": 0.92, "learning_rate": 3.464472177474138e-05, "loss": 0.0005, "step": 47484 }, { "epoch": 0.92, "learning_rate": 3.4644074814484146e-05, "loss": 0.0025, "step": 47486 }, { "epoch": 0.92, "learning_rate": 3.4643427854226915e-05, "loss": 0.0, "step": 47488 }, { "epoch": 0.92, "learning_rate": 3.4642780893969684e-05, "loss": 0.0, "step": 47490 }, { "epoch": 0.92, "learning_rate": 3.464213393371245e-05, "loss": 0.0054, "step": 47492 }, { "epoch": 0.92, "learning_rate": 3.464148697345522e-05, "loss": 0.0, "step": 47494 }, { "epoch": 0.92, "learning_rate": 3.464084001319799e-05, "loss": 0.0077, "step": 47496 }, { "epoch": 0.92, "learning_rate": 3.464019305294076e-05, "loss": 0.0, "step": 47498 }, { "epoch": 0.92, "learning_rate": 3.463954609268353e-05, "loss": 0.0001, "step": 47500 }, { "epoch": 0.92, "learning_rate": 3.463889913242629e-05, "loss": 0.0014, "step": 47502 }, { "epoch": 0.92, "learning_rate": 3.463825217216907e-05, "loss": 0.0, "step": 47504 }, { "epoch": 0.92, "learning_rate": 3.463760521191184e-05, "loss": 0.0, "step": 47506 }, { "epoch": 0.92, "learning_rate": 3.46369582516546e-05, "loss": 0.0, "step": 47508 }, { "epoch": 0.92, "learning_rate": 3.4636311291397375e-05, "loss": 0.0, "step": 47510 }, { "epoch": 0.92, "learning_rate": 3.463566433114014e-05, "loss": 0.0, "step": 47512 }, { "epoch": 0.92, "learning_rate": 3.463501737088291e-05, "loss": 0.0, "step": 47514 }, { "epoch": 0.92, "learning_rate": 3.4634370410625675e-05, "loss": 0.0, "step": 47516 }, { "epoch": 0.92, "learning_rate": 3.4633723450368444e-05, "loss": 0.0, "step": 47518 }, { "epoch": 0.92, "learning_rate": 3.4633076490111213e-05, "loss": 0.0005, "step": 47520 }, { "epoch": 0.92, "learning_rate": 3.463242952985398e-05, "loss": 0.0, "step": 47522 }, { "epoch": 0.92, "learning_rate": 3.463178256959675e-05, "loss": 0.0001, "step": 47524 }, { "epoch": 0.92, "learning_rate": 3.463113560933952e-05, "loss": 0.0002, "step": 47526 }, { "epoch": 0.92, "learning_rate": 3.463048864908229e-05, "loss": 0.0, "step": 47528 }, { "epoch": 0.92, "learning_rate": 3.462984168882506e-05, "loss": 0.0001, "step": 47530 }, { "epoch": 0.92, "learning_rate": 3.462919472856783e-05, "loss": 0.0028, "step": 47532 }, { "epoch": 0.92, "learning_rate": 3.462854776831059e-05, "loss": 0.0, "step": 47534 }, { "epoch": 0.92, "learning_rate": 3.4627900808053366e-05, "loss": 0.0022, "step": 47536 }, { "epoch": 0.92, "learning_rate": 3.462725384779613e-05, "loss": 0.0173, "step": 47538 }, { "epoch": 0.92, "learning_rate": 3.46266068875389e-05, "loss": 0.0, "step": 47540 }, { "epoch": 0.92, "learning_rate": 3.462595992728167e-05, "loss": 0.0055, "step": 47542 }, { "epoch": 0.92, "learning_rate": 3.4625312967024436e-05, "loss": 0.0, "step": 47544 }, { "epoch": 0.92, "learning_rate": 3.4624666006767205e-05, "loss": 0.0005, "step": 47546 }, { "epoch": 0.92, "learning_rate": 3.4624019046509974e-05, "loss": 0.0001, "step": 47548 }, { "epoch": 0.92, "learning_rate": 3.462337208625274e-05, "loss": 0.0, "step": 47550 }, { "epoch": 0.92, "learning_rate": 3.462272512599551e-05, "loss": 0.0, "step": 47552 }, { "epoch": 0.92, "learning_rate": 3.462207816573828e-05, "loss": 0.0, "step": 47554 }, { "epoch": 0.92, "learning_rate": 3.462143120548105e-05, "loss": 0.0, "step": 47556 }, { "epoch": 0.92, "learning_rate": 3.462078424522382e-05, "loss": 0.0004, "step": 47558 }, { "epoch": 0.92, "learning_rate": 3.462013728496659e-05, "loss": 0.0065, "step": 47560 }, { "epoch": 0.92, "learning_rate": 3.461949032470935e-05, "loss": 0.0001, "step": 47562 }, { "epoch": 0.92, "learning_rate": 3.4618843364452127e-05, "loss": 0.0005, "step": 47564 }, { "epoch": 0.92, "learning_rate": 3.461819640419489e-05, "loss": 0.0, "step": 47566 }, { "epoch": 0.92, "learning_rate": 3.461754944393766e-05, "loss": 0.0, "step": 47568 }, { "epoch": 0.92, "learning_rate": 3.461690248368043e-05, "loss": 0.0007, "step": 47570 }, { "epoch": 0.92, "learning_rate": 3.4616255523423196e-05, "loss": 0.0, "step": 47572 }, { "epoch": 0.92, "learning_rate": 3.461560856316597e-05, "loss": 0.0, "step": 47574 }, { "epoch": 0.92, "learning_rate": 3.4614961602908734e-05, "loss": 0.0, "step": 47576 }, { "epoch": 0.92, "learning_rate": 3.46143146426515e-05, "loss": 0.0, "step": 47578 }, { "epoch": 0.92, "learning_rate": 3.461366768239427e-05, "loss": 0.0012, "step": 47580 }, { "epoch": 0.92, "learning_rate": 3.461302072213704e-05, "loss": 0.0, "step": 47582 }, { "epoch": 0.92, "learning_rate": 3.4612373761879804e-05, "loss": 0.0, "step": 47584 }, { "epoch": 0.92, "learning_rate": 3.461172680162258e-05, "loss": 0.0178, "step": 47586 }, { "epoch": 0.92, "learning_rate": 3.461107984136535e-05, "loss": 0.0004, "step": 47588 }, { "epoch": 0.92, "learning_rate": 3.461043288110811e-05, "loss": 0.0, "step": 47590 }, { "epoch": 0.92, "learning_rate": 3.460978592085089e-05, "loss": 0.0, "step": 47592 }, { "epoch": 0.92, "learning_rate": 3.460913896059365e-05, "loss": 0.0072, "step": 47594 }, { "epoch": 0.92, "learning_rate": 3.4608492000336425e-05, "loss": 0.0, "step": 47596 }, { "epoch": 0.92, "learning_rate": 3.460784504007919e-05, "loss": 0.0003, "step": 47598 }, { "epoch": 0.92, "learning_rate": 3.4607198079821956e-05, "loss": 0.0, "step": 47600 }, { "epoch": 0.92, "learning_rate": 3.4606551119564726e-05, "loss": 0.0, "step": 47602 }, { "epoch": 0.92, "learning_rate": 3.4605904159307495e-05, "loss": 0.0005, "step": 47604 }, { "epoch": 0.92, "learning_rate": 3.4605257199050264e-05, "loss": 0.0001, "step": 47606 }, { "epoch": 0.92, "learning_rate": 3.460461023879303e-05, "loss": 0.0025, "step": 47608 }, { "epoch": 0.92, "learning_rate": 3.46039632785358e-05, "loss": 0.0, "step": 47610 }, { "epoch": 0.92, "learning_rate": 3.460331631827857e-05, "loss": 0.0003, "step": 47612 }, { "epoch": 0.92, "learning_rate": 3.460266935802134e-05, "loss": 0.0019, "step": 47614 }, { "epoch": 0.92, "learning_rate": 3.46020223977641e-05, "loss": 0.0021, "step": 47616 }, { "epoch": 0.92, "learning_rate": 3.460137543750688e-05, "loss": 0.0002, "step": 47618 }, { "epoch": 0.92, "learning_rate": 3.460072847724965e-05, "loss": 0.0, "step": 47620 }, { "epoch": 0.92, "learning_rate": 3.460008151699241e-05, "loss": 0.0047, "step": 47622 }, { "epoch": 0.92, "learning_rate": 3.4599434556735185e-05, "loss": 0.0, "step": 47624 }, { "epoch": 0.92, "learning_rate": 3.459878759647795e-05, "loss": 0.0, "step": 47626 }, { "epoch": 0.92, "learning_rate": 3.459814063622072e-05, "loss": 0.0013, "step": 47628 }, { "epoch": 0.92, "learning_rate": 3.4597493675963486e-05, "loss": 0.0, "step": 47630 }, { "epoch": 0.92, "learning_rate": 3.4596846715706255e-05, "loss": 0.0, "step": 47632 }, { "epoch": 0.92, "learning_rate": 3.4596199755449024e-05, "loss": 0.0, "step": 47634 }, { "epoch": 0.92, "learning_rate": 3.459555279519179e-05, "loss": 0.0, "step": 47636 }, { "epoch": 0.92, "learning_rate": 3.459490583493456e-05, "loss": 0.0001, "step": 47638 }, { "epoch": 0.92, "learning_rate": 3.459425887467733e-05, "loss": 0.0, "step": 47640 }, { "epoch": 0.92, "learning_rate": 3.45936119144201e-05, "loss": 0.0048, "step": 47642 }, { "epoch": 0.92, "learning_rate": 3.459296495416286e-05, "loss": 0.0, "step": 47644 }, { "epoch": 0.92, "learning_rate": 3.459231799390564e-05, "loss": 0.0, "step": 47646 }, { "epoch": 0.92, "learning_rate": 3.45916710336484e-05, "loss": 0.0002, "step": 47648 }, { "epoch": 0.92, "learning_rate": 3.459102407339117e-05, "loss": 0.0003, "step": 47650 }, { "epoch": 0.92, "learning_rate": 3.4590377113133946e-05, "loss": 0.0, "step": 47652 }, { "epoch": 0.92, "learning_rate": 3.458973015287671e-05, "loss": 0.0069, "step": 47654 }, { "epoch": 0.92, "learning_rate": 3.4589083192619484e-05, "loss": 0.0001, "step": 47656 }, { "epoch": 0.92, "learning_rate": 3.4588436232362246e-05, "loss": 0.0017, "step": 47658 }, { "epoch": 0.93, "learning_rate": 3.4587789272105015e-05, "loss": 0.0, "step": 47660 }, { "epoch": 0.93, "learning_rate": 3.4587142311847785e-05, "loss": 0.0, "step": 47662 }, { "epoch": 0.93, "learning_rate": 3.4586495351590554e-05, "loss": 0.0151, "step": 47664 }, { "epoch": 0.93, "learning_rate": 3.458584839133332e-05, "loss": 0.0106, "step": 47666 }, { "epoch": 0.93, "learning_rate": 3.458520143107609e-05, "loss": 0.0051, "step": 47668 }, { "epoch": 0.93, "learning_rate": 3.458455447081886e-05, "loss": 0.0028, "step": 47670 }, { "epoch": 0.93, "learning_rate": 3.458390751056163e-05, "loss": 0.0, "step": 47672 }, { "epoch": 0.93, "learning_rate": 3.45832605503044e-05, "loss": 0.0042, "step": 47674 }, { "epoch": 0.93, "learning_rate": 3.458261359004716e-05, "loss": 0.0009, "step": 47676 }, { "epoch": 0.93, "learning_rate": 3.458196662978994e-05, "loss": 0.003, "step": 47678 }, { "epoch": 0.93, "learning_rate": 3.45813196695327e-05, "loss": 0.0001, "step": 47680 }, { "epoch": 0.93, "learning_rate": 3.458067270927547e-05, "loss": 0.0, "step": 47682 }, { "epoch": 0.93, "learning_rate": 3.4580025749018244e-05, "loss": 0.0, "step": 47684 }, { "epoch": 0.93, "learning_rate": 3.457937878876101e-05, "loss": 0.0042, "step": 47686 }, { "epoch": 0.93, "learning_rate": 3.4578731828503776e-05, "loss": 0.0001, "step": 47688 }, { "epoch": 0.93, "learning_rate": 3.4578084868246545e-05, "loss": 0.0133, "step": 47690 }, { "epoch": 0.93, "learning_rate": 3.4577437907989314e-05, "loss": 0.0046, "step": 47692 }, { "epoch": 0.93, "learning_rate": 3.457679094773208e-05, "loss": 0.0, "step": 47694 }, { "epoch": 0.93, "learning_rate": 3.457614398747485e-05, "loss": 0.0, "step": 47696 }, { "epoch": 0.93, "learning_rate": 3.457549702721762e-05, "loss": 0.0004, "step": 47698 }, { "epoch": 0.93, "learning_rate": 3.457485006696039e-05, "loss": 0.0, "step": 47700 }, { "epoch": 0.93, "learning_rate": 3.457420310670316e-05, "loss": 0.0, "step": 47702 }, { "epoch": 0.93, "learning_rate": 3.457355614644592e-05, "loss": 0.0, "step": 47704 }, { "epoch": 0.93, "learning_rate": 3.45729091861887e-05, "loss": 0.0005, "step": 47706 }, { "epoch": 0.93, "learning_rate": 3.457226222593146e-05, "loss": 0.0, "step": 47708 }, { "epoch": 0.93, "learning_rate": 3.457161526567423e-05, "loss": 0.0001, "step": 47710 }, { "epoch": 0.93, "learning_rate": 3.4570968305417e-05, "loss": 0.0001, "step": 47712 }, { "epoch": 0.93, "learning_rate": 3.457032134515977e-05, "loss": 0.0, "step": 47714 }, { "epoch": 0.93, "learning_rate": 3.4569674384902536e-05, "loss": 0.0, "step": 47716 }, { "epoch": 0.93, "learning_rate": 3.4569027424645305e-05, "loss": 0.0, "step": 47718 }, { "epoch": 0.93, "learning_rate": 3.4568380464388074e-05, "loss": 0.0016, "step": 47720 }, { "epoch": 0.93, "learning_rate": 3.456805698425946e-05, "loss": 0.013, "step": 47722 }, { "epoch": 0.93, "learning_rate": 3.456741002400223e-05, "loss": 0.0021, "step": 47724 }, { "epoch": 0.93, "learning_rate": 3.4566763063745e-05, "loss": 0.0001, "step": 47726 }, { "epoch": 0.93, "learning_rate": 3.4566116103487766e-05, "loss": 0.0028, "step": 47728 }, { "epoch": 0.93, "learning_rate": 3.4565469143230535e-05, "loss": 0.0, "step": 47730 }, { "epoch": 0.93, "learning_rate": 3.45648221829733e-05, "loss": 0.0001, "step": 47732 }, { "epoch": 0.93, "learning_rate": 3.4564175222716073e-05, "loss": 0.0, "step": 47734 }, { "epoch": 0.93, "learning_rate": 3.4563528262458836e-05, "loss": 0.0003, "step": 47736 }, { "epoch": 0.93, "learning_rate": 3.4562881302201605e-05, "loss": 0.0, "step": 47738 }, { "epoch": 0.93, "learning_rate": 3.4562234341944374e-05, "loss": 0.0, "step": 47740 }, { "epoch": 0.93, "learning_rate": 3.456158738168714e-05, "loss": 0.0003, "step": 47742 }, { "epoch": 0.93, "learning_rate": 3.456094042142992e-05, "loss": 0.0047, "step": 47744 }, { "epoch": 0.93, "learning_rate": 3.456029346117268e-05, "loss": 0.0079, "step": 47746 }, { "epoch": 0.93, "learning_rate": 3.455964650091545e-05, "loss": 0.0061, "step": 47748 }, { "epoch": 0.93, "learning_rate": 3.455899954065822e-05, "loss": 0.0002, "step": 47750 }, { "epoch": 0.93, "learning_rate": 3.455835258040099e-05, "loss": 0.0, "step": 47752 }, { "epoch": 0.93, "learning_rate": 3.455770562014376e-05, "loss": 0.0059, "step": 47754 }, { "epoch": 0.93, "learning_rate": 3.455705865988653e-05, "loss": 0.0076, "step": 47756 }, { "epoch": 0.93, "learning_rate": 3.4556411699629296e-05, "loss": 0.0008, "step": 47758 }, { "epoch": 0.93, "learning_rate": 3.455576473937206e-05, "loss": 0.0001, "step": 47760 }, { "epoch": 0.93, "learning_rate": 3.4555117779114834e-05, "loss": 0.0, "step": 47762 }, { "epoch": 0.93, "learning_rate": 3.4554470818857596e-05, "loss": 0.0003, "step": 47764 }, { "epoch": 0.93, "learning_rate": 3.455382385860037e-05, "loss": 0.0002, "step": 47766 }, { "epoch": 0.93, "learning_rate": 3.4553176898343134e-05, "loss": 0.0, "step": 47768 }, { "epoch": 0.93, "learning_rate": 3.4552529938085903e-05, "loss": 0.0, "step": 47770 }, { "epoch": 0.93, "learning_rate": 3.455188297782867e-05, "loss": 0.0, "step": 47772 }, { "epoch": 0.93, "learning_rate": 3.455123601757144e-05, "loss": 0.0001, "step": 47774 }, { "epoch": 0.93, "learning_rate": 3.455058905731421e-05, "loss": 0.0, "step": 47776 }, { "epoch": 0.93, "learning_rate": 3.454994209705698e-05, "loss": 0.0003, "step": 47778 }, { "epoch": 0.93, "learning_rate": 3.454929513679975e-05, "loss": 0.0, "step": 47780 }, { "epoch": 0.93, "learning_rate": 3.454864817654251e-05, "loss": 0.0, "step": 47782 }, { "epoch": 0.93, "learning_rate": 3.454800121628529e-05, "loss": 0.0, "step": 47784 }, { "epoch": 0.93, "learning_rate": 3.4547354256028056e-05, "loss": 0.0, "step": 47786 }, { "epoch": 0.93, "learning_rate": 3.4546707295770825e-05, "loss": 0.0, "step": 47788 }, { "epoch": 0.93, "learning_rate": 3.4546060335513594e-05, "loss": 0.0001, "step": 47790 }, { "epoch": 0.93, "learning_rate": 3.4545413375256357e-05, "loss": 0.0002, "step": 47792 }, { "epoch": 0.93, "learning_rate": 3.454476641499913e-05, "loss": 0.0001, "step": 47794 }, { "epoch": 0.93, "learning_rate": 3.4544119454741895e-05, "loss": 0.0, "step": 47796 }, { "epoch": 0.93, "learning_rate": 3.4543472494484664e-05, "loss": 0.0, "step": 47798 }, { "epoch": 0.93, "learning_rate": 3.454282553422743e-05, "loss": 0.0, "step": 47800 }, { "epoch": 0.93, "learning_rate": 3.45421785739702e-05, "loss": 0.0, "step": 47802 }, { "epoch": 0.93, "learning_rate": 3.454153161371297e-05, "loss": 0.0068, "step": 47804 }, { "epoch": 0.93, "learning_rate": 3.454088465345574e-05, "loss": 0.0, "step": 47806 }, { "epoch": 0.93, "learning_rate": 3.454023769319851e-05, "loss": 0.0, "step": 47808 }, { "epoch": 0.93, "learning_rate": 3.453959073294128e-05, "loss": 0.0, "step": 47810 }, { "epoch": 0.93, "learning_rate": 3.453894377268405e-05, "loss": 0.0, "step": 47812 }, { "epoch": 0.93, "learning_rate": 3.453829681242681e-05, "loss": 0.0001, "step": 47814 }, { "epoch": 0.93, "learning_rate": 3.4537649852169586e-05, "loss": 0.0074, "step": 47816 }, { "epoch": 0.93, "learning_rate": 3.453700289191235e-05, "loss": 0.0, "step": 47818 }, { "epoch": 0.93, "learning_rate": 3.453635593165512e-05, "loss": 0.0, "step": 47820 }, { "epoch": 0.93, "learning_rate": 3.453570897139789e-05, "loss": 0.0001, "step": 47822 }, { "epoch": 0.93, "learning_rate": 3.4535062011140655e-05, "loss": 0.0, "step": 47824 }, { "epoch": 0.93, "learning_rate": 3.453441505088343e-05, "loss": 0.0059, "step": 47826 }, { "epoch": 0.93, "learning_rate": 3.453376809062619e-05, "loss": 0.0, "step": 47828 }, { "epoch": 0.93, "learning_rate": 3.453312113036896e-05, "loss": 0.0002, "step": 47830 }, { "epoch": 0.93, "learning_rate": 3.453247417011173e-05, "loss": 0.0, "step": 47832 }, { "epoch": 0.93, "learning_rate": 3.45318272098545e-05, "loss": 0.0, "step": 47834 }, { "epoch": 0.93, "learning_rate": 3.453118024959727e-05, "loss": 0.0, "step": 47836 }, { "epoch": 0.93, "learning_rate": 3.453053328934004e-05, "loss": 0.0, "step": 47838 }, { "epoch": 0.93, "learning_rate": 3.452988632908281e-05, "loss": 0.0, "step": 47840 }, { "epoch": 0.93, "learning_rate": 3.452923936882557e-05, "loss": 0.0007, "step": 47842 }, { "epoch": 0.93, "learning_rate": 3.4528592408568346e-05, "loss": 0.0012, "step": 47844 }, { "epoch": 0.93, "learning_rate": 3.452794544831111e-05, "loss": 0.0, "step": 47846 }, { "epoch": 0.93, "learning_rate": 3.4527298488053884e-05, "loss": 0.0, "step": 47848 }, { "epoch": 0.93, "learning_rate": 3.4526651527796647e-05, "loss": 0.0, "step": 47850 }, { "epoch": 0.93, "learning_rate": 3.4526004567539416e-05, "loss": 0.0003, "step": 47852 }, { "epoch": 0.93, "learning_rate": 3.452535760728219e-05, "loss": 0.0, "step": 47854 }, { "epoch": 0.93, "learning_rate": 3.4524710647024954e-05, "loss": 0.0, "step": 47856 }, { "epoch": 0.93, "learning_rate": 3.452406368676772e-05, "loss": 0.0, "step": 47858 }, { "epoch": 0.93, "learning_rate": 3.452341672651049e-05, "loss": 0.0005, "step": 47860 }, { "epoch": 0.93, "learning_rate": 3.452276976625326e-05, "loss": 0.0055, "step": 47862 }, { "epoch": 0.93, "learning_rate": 3.452212280599602e-05, "loss": 0.0004, "step": 47864 }, { "epoch": 0.93, "learning_rate": 3.45214758457388e-05, "loss": 0.0066, "step": 47866 }, { "epoch": 0.93, "learning_rate": 3.452082888548157e-05, "loss": 0.0, "step": 47868 }, { "epoch": 0.93, "learning_rate": 3.452018192522434e-05, "loss": 0.0013, "step": 47870 }, { "epoch": 0.93, "learning_rate": 3.4519534964967106e-05, "loss": 0.001, "step": 47872 }, { "epoch": 0.93, "learning_rate": 3.451888800470987e-05, "loss": 0.0, "step": 47874 }, { "epoch": 0.93, "learning_rate": 3.4518241044452645e-05, "loss": 0.0, "step": 47876 }, { "epoch": 0.93, "learning_rate": 3.451759408419541e-05, "loss": 0.0, "step": 47878 }, { "epoch": 0.93, "learning_rate": 3.4516947123938176e-05, "loss": 0.0, "step": 47880 }, { "epoch": 0.93, "learning_rate": 3.4516300163680945e-05, "loss": 0.0, "step": 47882 }, { "epoch": 0.93, "learning_rate": 3.4515653203423714e-05, "loss": 0.0, "step": 47884 }, { "epoch": 0.93, "learning_rate": 3.451500624316649e-05, "loss": 0.0, "step": 47886 }, { "epoch": 0.93, "learning_rate": 3.451435928290925e-05, "loss": 0.0, "step": 47888 }, { "epoch": 0.93, "learning_rate": 3.451371232265202e-05, "loss": 0.0102, "step": 47890 }, { "epoch": 0.93, "learning_rate": 3.451306536239479e-05, "loss": 0.0, "step": 47892 }, { "epoch": 0.93, "learning_rate": 3.451241840213756e-05, "loss": 0.0002, "step": 47894 }, { "epoch": 0.93, "learning_rate": 3.451177144188032e-05, "loss": 0.0, "step": 47896 }, { "epoch": 0.93, "learning_rate": 3.45111244816231e-05, "loss": 0.0, "step": 47898 }, { "epoch": 0.93, "learning_rate": 3.451047752136587e-05, "loss": 0.0, "step": 47900 }, { "epoch": 0.93, "learning_rate": 3.450983056110863e-05, "loss": 0.0, "step": 47902 }, { "epoch": 0.93, "learning_rate": 3.4509183600851405e-05, "loss": 0.0001, "step": 47904 }, { "epoch": 0.93, "learning_rate": 3.450853664059417e-05, "loss": 0.0001, "step": 47906 }, { "epoch": 0.93, "learning_rate": 3.450788968033694e-05, "loss": 0.0, "step": 47908 }, { "epoch": 0.93, "learning_rate": 3.4507242720079705e-05, "loss": 0.0, "step": 47910 }, { "epoch": 0.93, "learning_rate": 3.4506595759822475e-05, "loss": 0.0003, "step": 47912 }, { "epoch": 0.93, "learning_rate": 3.4505948799565244e-05, "loss": 0.0001, "step": 47914 }, { "epoch": 0.93, "learning_rate": 3.450530183930801e-05, "loss": 0.0001, "step": 47916 }, { "epoch": 0.93, "learning_rate": 3.450465487905078e-05, "loss": 0.0, "step": 47918 }, { "epoch": 0.93, "learning_rate": 3.450400791879355e-05, "loss": 0.0, "step": 47920 }, { "epoch": 0.93, "learning_rate": 3.450336095853632e-05, "loss": 0.0, "step": 47922 }, { "epoch": 0.93, "learning_rate": 3.450271399827908e-05, "loss": 0.0001, "step": 47924 }, { "epoch": 0.93, "learning_rate": 3.450206703802186e-05, "loss": 0.0, "step": 47926 }, { "epoch": 0.93, "learning_rate": 3.450142007776462e-05, "loss": 0.0, "step": 47928 }, { "epoch": 0.93, "learning_rate": 3.4500773117507396e-05, "loss": 0.0158, "step": 47930 }, { "epoch": 0.93, "learning_rate": 3.4500126157250165e-05, "loss": 0.0, "step": 47932 }, { "epoch": 0.93, "learning_rate": 3.449947919699293e-05, "loss": 0.0038, "step": 47934 }, { "epoch": 0.93, "learning_rate": 3.4498832236735704e-05, "loss": 0.0, "step": 47936 }, { "epoch": 0.93, "learning_rate": 3.4498185276478466e-05, "loss": 0.0, "step": 47938 }, { "epoch": 0.93, "learning_rate": 3.4497538316221235e-05, "loss": 0.0368, "step": 47940 }, { "epoch": 0.93, "learning_rate": 3.4496891355964004e-05, "loss": 0.0001, "step": 47942 }, { "epoch": 0.93, "learning_rate": 3.449624439570677e-05, "loss": 0.0017, "step": 47944 }, { "epoch": 0.93, "learning_rate": 3.449559743544954e-05, "loss": 0.0, "step": 47946 }, { "epoch": 0.93, "learning_rate": 3.449495047519231e-05, "loss": 0.0, "step": 47948 }, { "epoch": 0.93, "learning_rate": 3.449430351493508e-05, "loss": 0.0, "step": 47950 }, { "epoch": 0.93, "learning_rate": 3.449365655467785e-05, "loss": 0.0, "step": 47952 }, { "epoch": 0.93, "learning_rate": 3.449300959442062e-05, "loss": 0.0017, "step": 47954 }, { "epoch": 0.93, "learning_rate": 3.449236263416338e-05, "loss": 0.0002, "step": 47956 }, { "epoch": 0.93, "learning_rate": 3.449171567390616e-05, "loss": 0.0, "step": 47958 }, { "epoch": 0.93, "learning_rate": 3.449106871364892e-05, "loss": 0.0024, "step": 47960 }, { "epoch": 0.93, "learning_rate": 3.449042175339169e-05, "loss": 0.0, "step": 47962 }, { "epoch": 0.93, "learning_rate": 3.448977479313446e-05, "loss": 0.0, "step": 47964 }, { "epoch": 0.93, "learning_rate": 3.4489127832877226e-05, "loss": 0.0, "step": 47966 }, { "epoch": 0.93, "learning_rate": 3.448848087262e-05, "loss": 0.0016, "step": 47968 }, { "epoch": 0.93, "learning_rate": 3.4487833912362764e-05, "loss": 0.0001, "step": 47970 }, { "epoch": 0.93, "learning_rate": 3.4487186952105534e-05, "loss": 0.0, "step": 47972 }, { "epoch": 0.93, "learning_rate": 3.44865399918483e-05, "loss": 0.0001, "step": 47974 }, { "epoch": 0.93, "learning_rate": 3.448589303159107e-05, "loss": 0.0018, "step": 47976 }, { "epoch": 0.93, "learning_rate": 3.448524607133384e-05, "loss": 0.0, "step": 47978 }, { "epoch": 0.93, "learning_rate": 3.448459911107661e-05, "loss": 0.0, "step": 47980 }, { "epoch": 0.93, "learning_rate": 3.448395215081938e-05, "loss": 0.0, "step": 47982 }, { "epoch": 0.93, "learning_rate": 3.448330519056214e-05, "loss": 0.0001, "step": 47984 }, { "epoch": 0.93, "learning_rate": 3.448265823030492e-05, "loss": 0.0, "step": 47986 }, { "epoch": 0.93, "learning_rate": 3.448201127004768e-05, "loss": 0.0, "step": 47988 }, { "epoch": 0.93, "learning_rate": 3.4481364309790455e-05, "loss": 0.0, "step": 47990 }, { "epoch": 0.93, "learning_rate": 3.448071734953322e-05, "loss": 0.0001, "step": 47992 }, { "epoch": 0.93, "learning_rate": 3.448007038927599e-05, "loss": 0.0009, "step": 47994 }, { "epoch": 0.93, "learning_rate": 3.4479423429018756e-05, "loss": 0.0, "step": 47996 }, { "epoch": 0.93, "learning_rate": 3.4478776468761525e-05, "loss": 0.0001, "step": 47998 }, { "epoch": 0.93, "learning_rate": 3.4478129508504294e-05, "loss": 0.0, "step": 48000 }, { "epoch": 0.93, "learning_rate": 3.447748254824706e-05, "loss": 0.0, "step": 48002 }, { "epoch": 0.93, "learning_rate": 3.447683558798983e-05, "loss": 0.0001, "step": 48004 }, { "epoch": 0.93, "learning_rate": 3.4476188627732594e-05, "loss": 0.0, "step": 48006 }, { "epoch": 0.93, "learning_rate": 3.447554166747537e-05, "loss": 0.0, "step": 48008 }, { "epoch": 0.93, "learning_rate": 3.447489470721814e-05, "loss": 0.0, "step": 48010 }, { "epoch": 0.93, "learning_rate": 3.447424774696091e-05, "loss": 0.0, "step": 48012 }, { "epoch": 0.93, "learning_rate": 3.447360078670368e-05, "loss": 0.0, "step": 48014 }, { "epoch": 0.93, "learning_rate": 3.447295382644644e-05, "loss": 0.0, "step": 48016 }, { "epoch": 0.93, "learning_rate": 3.4472306866189216e-05, "loss": 0.0, "step": 48018 }, { "epoch": 0.93, "learning_rate": 3.447165990593198e-05, "loss": 0.0, "step": 48020 }, { "epoch": 0.93, "learning_rate": 3.447101294567475e-05, "loss": 0.0, "step": 48022 }, { "epoch": 0.93, "learning_rate": 3.4470365985417516e-05, "loss": 0.0004, "step": 48024 }, { "epoch": 0.93, "learning_rate": 3.4469719025160285e-05, "loss": 0.0, "step": 48026 }, { "epoch": 0.93, "learning_rate": 3.4469072064903054e-05, "loss": 0.006, "step": 48028 }, { "epoch": 0.93, "learning_rate": 3.4468425104645823e-05, "loss": 0.0, "step": 48030 }, { "epoch": 0.93, "learning_rate": 3.446777814438859e-05, "loss": 0.0, "step": 48032 }, { "epoch": 0.93, "learning_rate": 3.446713118413136e-05, "loss": 0.0, "step": 48034 }, { "epoch": 0.93, "learning_rate": 3.446648422387413e-05, "loss": 0.0, "step": 48036 }, { "epoch": 0.93, "learning_rate": 3.446583726361689e-05, "loss": 0.0, "step": 48038 }, { "epoch": 0.93, "learning_rate": 3.446519030335967e-05, "loss": 0.0, "step": 48040 }, { "epoch": 0.93, "learning_rate": 3.446454334310243e-05, "loss": 0.0107, "step": 48042 }, { "epoch": 0.93, "learning_rate": 3.44638963828452e-05, "loss": 0.0, "step": 48044 }, { "epoch": 0.93, "learning_rate": 3.4463249422587976e-05, "loss": 0.0, "step": 48046 }, { "epoch": 0.93, "learning_rate": 3.446260246233074e-05, "loss": 0.0, "step": 48048 }, { "epoch": 0.93, "learning_rate": 3.4461955502073514e-05, "loss": 0.0, "step": 48050 }, { "epoch": 0.93, "learning_rate": 3.4461308541816277e-05, "loss": 0.0, "step": 48052 }, { "epoch": 0.93, "learning_rate": 3.4460661581559046e-05, "loss": 0.0294, "step": 48054 }, { "epoch": 0.93, "learning_rate": 3.4460014621301815e-05, "loss": 0.0001, "step": 48056 }, { "epoch": 0.93, "learning_rate": 3.4459367661044584e-05, "loss": 0.0001, "step": 48058 }, { "epoch": 0.93, "learning_rate": 3.445872070078735e-05, "loss": 0.0, "step": 48060 }, { "epoch": 0.93, "learning_rate": 3.445807374053012e-05, "loss": 0.0, "step": 48062 }, { "epoch": 0.93, "learning_rate": 3.445742678027289e-05, "loss": 0.0, "step": 48064 }, { "epoch": 0.93, "learning_rate": 3.4456779820015653e-05, "loss": 0.0, "step": 48066 }, { "epoch": 0.93, "learning_rate": 3.445613285975843e-05, "loss": 0.0, "step": 48068 }, { "epoch": 0.93, "learning_rate": 3.445548589950119e-05, "loss": 0.0001, "step": 48070 }, { "epoch": 0.93, "learning_rate": 3.445483893924397e-05, "loss": 0.0, "step": 48072 }, { "epoch": 0.93, "learning_rate": 3.445419197898673e-05, "loss": 0.0, "step": 48074 }, { "epoch": 0.93, "learning_rate": 3.44535450187295e-05, "loss": 0.0045, "step": 48076 }, { "epoch": 0.93, "learning_rate": 3.4452898058472275e-05, "loss": 0.0001, "step": 48078 }, { "epoch": 0.93, "learning_rate": 3.445225109821504e-05, "loss": 0.0018, "step": 48080 }, { "epoch": 0.93, "learning_rate": 3.4451604137957806e-05, "loss": 0.0, "step": 48082 }, { "epoch": 0.93, "learning_rate": 3.4450957177700575e-05, "loss": 0.0005, "step": 48084 }, { "epoch": 0.93, "learning_rate": 3.4450310217443344e-05, "loss": 0.0, "step": 48086 }, { "epoch": 0.93, "learning_rate": 3.444966325718611e-05, "loss": 0.0, "step": 48088 }, { "epoch": 0.93, "learning_rate": 3.444901629692888e-05, "loss": 0.0, "step": 48090 }, { "epoch": 0.93, "learning_rate": 3.444836933667165e-05, "loss": 0.0067, "step": 48092 }, { "epoch": 0.93, "learning_rate": 3.444772237641442e-05, "loss": 0.0, "step": 48094 }, { "epoch": 0.93, "learning_rate": 3.444707541615719e-05, "loss": 0.0003, "step": 48096 }, { "epoch": 0.93, "learning_rate": 3.444642845589995e-05, "loss": 0.0, "step": 48098 }, { "epoch": 0.93, "learning_rate": 3.444578149564273e-05, "loss": 0.0, "step": 48100 }, { "epoch": 0.93, "learning_rate": 3.444513453538549e-05, "loss": 0.0002, "step": 48102 }, { "epoch": 0.93, "learning_rate": 3.444448757512826e-05, "loss": 0.0003, "step": 48104 }, { "epoch": 0.93, "learning_rate": 3.444384061487103e-05, "loss": 0.0276, "step": 48106 }, { "epoch": 0.93, "learning_rate": 3.44431936546138e-05, "loss": 0.0115, "step": 48108 }, { "epoch": 0.93, "learning_rate": 3.444254669435657e-05, "loss": 0.0024, "step": 48110 }, { "epoch": 0.93, "learning_rate": 3.4441899734099336e-05, "loss": 0.0002, "step": 48112 }, { "epoch": 0.93, "learning_rate": 3.4441252773842105e-05, "loss": 0.0, "step": 48114 }, { "epoch": 0.93, "learning_rate": 3.4440605813584874e-05, "loss": 0.0, "step": 48116 }, { "epoch": 0.93, "learning_rate": 3.443995885332764e-05, "loss": 0.0, "step": 48118 }, { "epoch": 0.93, "learning_rate": 3.4439311893070405e-05, "loss": 0.0019, "step": 48120 }, { "epoch": 0.93, "learning_rate": 3.443866493281318e-05, "loss": 0.0, "step": 48122 }, { "epoch": 0.93, "learning_rate": 3.443801797255595e-05, "loss": 0.0, "step": 48124 }, { "epoch": 0.93, "learning_rate": 3.443737101229871e-05, "loss": 0.0001, "step": 48126 }, { "epoch": 0.93, "learning_rate": 3.443672405204149e-05, "loss": 0.0, "step": 48128 }, { "epoch": 0.93, "learning_rate": 3.443607709178425e-05, "loss": 0.0819, "step": 48130 }, { "epoch": 0.93, "learning_rate": 3.4435430131527026e-05, "loss": 0.0, "step": 48132 }, { "epoch": 0.93, "learning_rate": 3.443478317126979e-05, "loss": 0.0, "step": 48134 }, { "epoch": 0.93, "learning_rate": 3.443413621101256e-05, "loss": 0.0105, "step": 48136 }, { "epoch": 0.93, "learning_rate": 3.443348925075533e-05, "loss": 0.0, "step": 48138 }, { "epoch": 0.93, "learning_rate": 3.4432842290498096e-05, "loss": 0.006, "step": 48140 }, { "epoch": 0.93, "learning_rate": 3.4432195330240865e-05, "loss": 0.0, "step": 48142 }, { "epoch": 0.93, "learning_rate": 3.4431548369983634e-05, "loss": 0.0, "step": 48144 }, { "epoch": 0.93, "learning_rate": 3.44309014097264e-05, "loss": 0.0, "step": 48146 }, { "epoch": 0.93, "learning_rate": 3.4430254449469166e-05, "loss": 0.0001, "step": 48148 }, { "epoch": 0.93, "learning_rate": 3.442960748921194e-05, "loss": 0.0, "step": 48150 }, { "epoch": 0.93, "learning_rate": 3.4428960528954704e-05, "loss": 0.0001, "step": 48152 }, { "epoch": 0.93, "learning_rate": 3.442831356869748e-05, "loss": 0.0, "step": 48154 }, { "epoch": 0.93, "learning_rate": 3.442766660844025e-05, "loss": 0.0, "step": 48156 }, { "epoch": 0.93, "learning_rate": 3.442701964818301e-05, "loss": 0.0014, "step": 48158 }, { "epoch": 0.93, "learning_rate": 3.442637268792579e-05, "loss": 0.0, "step": 48160 }, { "epoch": 0.93, "learning_rate": 3.442572572766855e-05, "loss": 0.0, "step": 48162 }, { "epoch": 0.93, "learning_rate": 3.442507876741132e-05, "loss": 0.0, "step": 48164 }, { "epoch": 0.93, "learning_rate": 3.442443180715409e-05, "loss": 0.0035, "step": 48166 }, { "epoch": 0.93, "learning_rate": 3.4423784846896856e-05, "loss": 0.0, "step": 48168 }, { "epoch": 0.93, "learning_rate": 3.4423137886639625e-05, "loss": 0.0008, "step": 48170 }, { "epoch": 0.93, "learning_rate": 3.4422490926382395e-05, "loss": 0.0, "step": 48172 }, { "epoch": 0.93, "learning_rate": 3.4421843966125164e-05, "loss": 0.0003, "step": 48174 }, { "epoch": 0.94, "learning_rate": 3.442119700586793e-05, "loss": 0.0, "step": 48176 }, { "epoch": 0.94, "learning_rate": 3.44205500456107e-05, "loss": 0.0312, "step": 48178 }, { "epoch": 0.94, "learning_rate": 3.4419903085353464e-05, "loss": 0.0, "step": 48180 }, { "epoch": 0.94, "learning_rate": 3.441925612509624e-05, "loss": 0.0, "step": 48182 }, { "epoch": 0.94, "learning_rate": 3.4418609164839e-05, "loss": 0.0, "step": 48184 }, { "epoch": 0.94, "learning_rate": 3.441796220458177e-05, "loss": 0.0, "step": 48186 }, { "epoch": 0.94, "learning_rate": 3.441731524432455e-05, "loss": 0.0, "step": 48188 }, { "epoch": 0.94, "learning_rate": 3.441666828406731e-05, "loss": 0.0115, "step": 48190 }, { "epoch": 0.94, "learning_rate": 3.4416021323810085e-05, "loss": 0.0003, "step": 48192 }, { "epoch": 0.94, "learning_rate": 3.441537436355285e-05, "loss": 0.0, "step": 48194 }, { "epoch": 0.94, "learning_rate": 3.441472740329562e-05, "loss": 0.0022, "step": 48196 }, { "epoch": 0.94, "learning_rate": 3.4414080443038386e-05, "loss": 0.0, "step": 48198 }, { "epoch": 0.94, "learning_rate": 3.4413433482781155e-05, "loss": 0.0005, "step": 48200 }, { "epoch": 0.94, "learning_rate": 3.4412786522523924e-05, "loss": 0.0001, "step": 48202 }, { "epoch": 0.94, "learning_rate": 3.441213956226669e-05, "loss": 0.0003, "step": 48204 }, { "epoch": 0.94, "learning_rate": 3.441149260200946e-05, "loss": 0.0, "step": 48206 }, { "epoch": 0.94, "learning_rate": 3.4410845641752225e-05, "loss": 0.0, "step": 48208 }, { "epoch": 0.94, "learning_rate": 3.4410198681495e-05, "loss": 0.0001, "step": 48210 }, { "epoch": 0.94, "learning_rate": 3.440955172123776e-05, "loss": 0.0, "step": 48212 }, { "epoch": 0.94, "learning_rate": 3.440890476098054e-05, "loss": 0.0, "step": 48214 }, { "epoch": 0.94, "learning_rate": 3.44082578007233e-05, "loss": 0.0, "step": 48216 }, { "epoch": 0.94, "learning_rate": 3.440761084046607e-05, "loss": 0.0001, "step": 48218 }, { "epoch": 0.94, "learning_rate": 3.440696388020884e-05, "loss": 0.0277, "step": 48220 }, { "epoch": 0.94, "learning_rate": 3.440631691995161e-05, "loss": 0.0, "step": 48222 }, { "epoch": 0.94, "learning_rate": 3.440566995969438e-05, "loss": 0.0, "step": 48224 }, { "epoch": 0.94, "learning_rate": 3.4405022999437146e-05, "loss": 0.0001, "step": 48226 }, { "epoch": 0.94, "learning_rate": 3.4404376039179915e-05, "loss": 0.0038, "step": 48228 }, { "epoch": 0.94, "learning_rate": 3.440372907892268e-05, "loss": 0.0002, "step": 48230 }, { "epoch": 0.94, "learning_rate": 3.4403082118665454e-05, "loss": 0.0001, "step": 48232 }, { "epoch": 0.94, "learning_rate": 3.440243515840822e-05, "loss": 0.0007, "step": 48234 }, { "epoch": 0.94, "learning_rate": 3.440178819815099e-05, "loss": 0.0, "step": 48236 }, { "epoch": 0.94, "learning_rate": 3.440114123789376e-05, "loss": 0.0, "step": 48238 }, { "epoch": 0.94, "learning_rate": 3.440049427763652e-05, "loss": 0.0098, "step": 48240 }, { "epoch": 0.94, "learning_rate": 3.43998473173793e-05, "loss": 0.0, "step": 48242 }, { "epoch": 0.94, "learning_rate": 3.439920035712206e-05, "loss": 0.0012, "step": 48244 }, { "epoch": 0.94, "learning_rate": 3.439855339686483e-05, "loss": 0.0, "step": 48246 }, { "epoch": 0.94, "learning_rate": 3.43979064366076e-05, "loss": 0.0001, "step": 48248 }, { "epoch": 0.94, "learning_rate": 3.439725947635037e-05, "loss": 0.0001, "step": 48250 }, { "epoch": 0.94, "learning_rate": 3.439661251609314e-05, "loss": 0.0699, "step": 48252 }, { "epoch": 0.94, "learning_rate": 3.439596555583591e-05, "loss": 0.0, "step": 48254 }, { "epoch": 0.94, "learning_rate": 3.4395318595578676e-05, "loss": 0.0, "step": 48256 }, { "epoch": 0.94, "learning_rate": 3.4394671635321445e-05, "loss": 0.0317, "step": 48258 }, { "epoch": 0.94, "learning_rate": 3.4394024675064214e-05, "loss": 0.0, "step": 48260 }, { "epoch": 0.94, "learning_rate": 3.4393377714806976e-05, "loss": 0.0001, "step": 48262 }, { "epoch": 0.94, "learning_rate": 3.439273075454975e-05, "loss": 0.0002, "step": 48264 }, { "epoch": 0.94, "learning_rate": 3.439208379429252e-05, "loss": 0.0, "step": 48266 }, { "epoch": 0.94, "learning_rate": 3.4391436834035283e-05, "loss": 0.0, "step": 48268 }, { "epoch": 0.94, "learning_rate": 3.439078987377806e-05, "loss": 0.0006, "step": 48270 }, { "epoch": 0.94, "learning_rate": 3.439014291352082e-05, "loss": 0.0, "step": 48272 }, { "epoch": 0.94, "learning_rate": 3.43894959532636e-05, "loss": 0.0, "step": 48274 }, { "epoch": 0.94, "learning_rate": 3.438884899300636e-05, "loss": 0.0, "step": 48276 }, { "epoch": 0.94, "learning_rate": 3.438820203274913e-05, "loss": 0.0002, "step": 48278 }, { "epoch": 0.94, "learning_rate": 3.43875550724919e-05, "loss": 0.0, "step": 48280 }, { "epoch": 0.94, "learning_rate": 3.438690811223467e-05, "loss": 0.0, "step": 48282 }, { "epoch": 0.94, "learning_rate": 3.4386261151977436e-05, "loss": 0.0063, "step": 48284 }, { "epoch": 0.94, "learning_rate": 3.4385614191720205e-05, "loss": 0.0, "step": 48286 }, { "epoch": 0.94, "learning_rate": 3.4384967231462974e-05, "loss": 0.0018, "step": 48288 }, { "epoch": 0.94, "learning_rate": 3.438432027120574e-05, "loss": 0.0006, "step": 48290 }, { "epoch": 0.94, "learning_rate": 3.438367331094851e-05, "loss": 0.0003, "step": 48292 }, { "epoch": 0.94, "learning_rate": 3.4383026350691275e-05, "loss": 0.0, "step": 48294 }, { "epoch": 0.94, "learning_rate": 3.438237939043405e-05, "loss": 0.0, "step": 48296 }, { "epoch": 0.94, "learning_rate": 3.438173243017681e-05, "loss": 0.006, "step": 48298 }, { "epoch": 0.94, "learning_rate": 3.438108546991958e-05, "loss": 0.0018, "step": 48300 }, { "epoch": 0.94, "learning_rate": 3.438043850966236e-05, "loss": 0.0122, "step": 48302 }, { "epoch": 0.94, "learning_rate": 3.437979154940512e-05, "loss": 0.0, "step": 48304 }, { "epoch": 0.94, "learning_rate": 3.437914458914789e-05, "loss": 0.0135, "step": 48306 }, { "epoch": 0.94, "learning_rate": 3.437849762889066e-05, "loss": 0.0, "step": 48308 }, { "epoch": 0.94, "learning_rate": 3.437785066863343e-05, "loss": 0.0, "step": 48310 }, { "epoch": 0.94, "learning_rate": 3.4377203708376197e-05, "loss": 0.0, "step": 48312 }, { "epoch": 0.94, "learning_rate": 3.4376556748118966e-05, "loss": 0.0008, "step": 48314 }, { "epoch": 0.94, "learning_rate": 3.4375909787861735e-05, "loss": 0.0, "step": 48316 }, { "epoch": 0.94, "learning_rate": 3.4375262827604504e-05, "loss": 0.0, "step": 48318 }, { "epoch": 0.94, "learning_rate": 3.437461586734727e-05, "loss": 0.004, "step": 48320 }, { "epoch": 0.94, "learning_rate": 3.4373968907090035e-05, "loss": 0.0, "step": 48322 }, { "epoch": 0.94, "learning_rate": 3.437332194683281e-05, "loss": 0.0029, "step": 48324 }, { "epoch": 0.94, "learning_rate": 3.437267498657557e-05, "loss": 0.0214, "step": 48326 }, { "epoch": 0.94, "learning_rate": 3.437202802631834e-05, "loss": 0.0002, "step": 48328 }, { "epoch": 0.94, "learning_rate": 3.437138106606111e-05, "loss": 0.0, "step": 48330 }, { "epoch": 0.94, "learning_rate": 3.437073410580388e-05, "loss": 0.0, "step": 48332 }, { "epoch": 0.94, "learning_rate": 3.4370087145546656e-05, "loss": 0.0, "step": 48334 }, { "epoch": 0.94, "learning_rate": 3.436944018528942e-05, "loss": 0.0001, "step": 48336 }, { "epoch": 0.94, "learning_rate": 3.436879322503219e-05, "loss": 0.0, "step": 48338 }, { "epoch": 0.94, "learning_rate": 3.436814626477496e-05, "loss": 0.0, "step": 48340 }, { "epoch": 0.94, "learning_rate": 3.4367499304517726e-05, "loss": 0.0, "step": 48342 }, { "epoch": 0.94, "learning_rate": 3.436685234426049e-05, "loss": 0.0002, "step": 48344 }, { "epoch": 0.94, "learning_rate": 3.4366205384003264e-05, "loss": 0.0, "step": 48346 }, { "epoch": 0.94, "learning_rate": 3.436555842374603e-05, "loss": 0.0, "step": 48348 }, { "epoch": 0.94, "learning_rate": 3.4364911463488796e-05, "loss": 0.0, "step": 48350 }, { "epoch": 0.94, "learning_rate": 3.436426450323157e-05, "loss": 0.0, "step": 48352 }, { "epoch": 0.94, "learning_rate": 3.4363617542974334e-05, "loss": 0.0696, "step": 48354 }, { "epoch": 0.94, "learning_rate": 3.436297058271711e-05, "loss": 0.0, "step": 48356 }, { "epoch": 0.94, "learning_rate": 3.436232362245987e-05, "loss": 0.0002, "step": 48358 }, { "epoch": 0.94, "learning_rate": 3.436167666220264e-05, "loss": 0.0, "step": 48360 }, { "epoch": 0.94, "learning_rate": 3.436102970194541e-05, "loss": 0.0002, "step": 48362 }, { "epoch": 0.94, "learning_rate": 3.436038274168818e-05, "loss": 0.0, "step": 48364 }, { "epoch": 0.94, "learning_rate": 3.435973578143095e-05, "loss": 0.0013, "step": 48366 }, { "epoch": 0.94, "learning_rate": 3.435908882117372e-05, "loss": 0.0452, "step": 48368 }, { "epoch": 0.94, "learning_rate": 3.4358441860916486e-05, "loss": 0.0, "step": 48370 }, { "epoch": 0.94, "learning_rate": 3.435779490065925e-05, "loss": 0.0069, "step": 48372 }, { "epoch": 0.94, "learning_rate": 3.4357147940402025e-05, "loss": 0.0, "step": 48374 }, { "epoch": 0.94, "learning_rate": 3.435650098014479e-05, "loss": 0.0001, "step": 48376 }, { "epoch": 0.94, "learning_rate": 3.435585401988756e-05, "loss": 0.0018, "step": 48378 }, { "epoch": 0.94, "learning_rate": 3.435520705963033e-05, "loss": 0.0002, "step": 48380 }, { "epoch": 0.94, "learning_rate": 3.4354560099373094e-05, "loss": 0.0, "step": 48382 }, { "epoch": 0.94, "learning_rate": 3.435391313911587e-05, "loss": 0.0002, "step": 48384 }, { "epoch": 0.94, "learning_rate": 3.435326617885863e-05, "loss": 0.0102, "step": 48386 }, { "epoch": 0.94, "learning_rate": 3.43526192186014e-05, "loss": 0.0534, "step": 48388 }, { "epoch": 0.94, "learning_rate": 3.435197225834417e-05, "loss": 0.0004, "step": 48390 }, { "epoch": 0.94, "learning_rate": 3.435132529808694e-05, "loss": 0.0148, "step": 48392 }, { "epoch": 0.94, "learning_rate": 3.435067833782971e-05, "loss": 0.0, "step": 48394 }, { "epoch": 0.94, "learning_rate": 3.435003137757248e-05, "loss": 0.0006, "step": 48396 }, { "epoch": 0.94, "learning_rate": 3.434938441731525e-05, "loss": 0.0198, "step": 48398 }, { "epoch": 0.94, "learning_rate": 3.4348737457058016e-05, "loss": 0.0, "step": 48400 }, { "epoch": 0.94, "learning_rate": 3.4348090496800785e-05, "loss": 0.0, "step": 48402 }, { "epoch": 0.94, "learning_rate": 3.434744353654355e-05, "loss": 0.0, "step": 48404 }, { "epoch": 0.94, "learning_rate": 3.434679657628632e-05, "loss": 0.0, "step": 48406 }, { "epoch": 0.94, "learning_rate": 3.4346149616029086e-05, "loss": 0.0, "step": 48408 }, { "epoch": 0.94, "learning_rate": 3.4345502655771855e-05, "loss": 0.0, "step": 48410 }, { "epoch": 0.94, "learning_rate": 3.434485569551463e-05, "loss": 0.0, "step": 48412 }, { "epoch": 0.94, "learning_rate": 3.434420873525739e-05, "loss": 0.0, "step": 48414 }, { "epoch": 0.94, "learning_rate": 3.434356177500017e-05, "loss": 0.0002, "step": 48416 }, { "epoch": 0.94, "learning_rate": 3.434291481474293e-05, "loss": 0.0001, "step": 48418 }, { "epoch": 0.94, "learning_rate": 3.43422678544857e-05, "loss": 0.0, "step": 48420 }, { "epoch": 0.94, "learning_rate": 3.434162089422847e-05, "loss": 0.0275, "step": 48422 }, { "epoch": 0.94, "learning_rate": 3.434097393397124e-05, "loss": 0.0004, "step": 48424 }, { "epoch": 0.94, "learning_rate": 3.434032697371401e-05, "loss": 0.0, "step": 48426 }, { "epoch": 0.94, "learning_rate": 3.4339680013456776e-05, "loss": 0.0, "step": 48428 }, { "epoch": 0.94, "learning_rate": 3.4339033053199545e-05, "loss": 0.0, "step": 48430 }, { "epoch": 0.94, "learning_rate": 3.433838609294231e-05, "loss": 0.0069, "step": 48432 }, { "epoch": 0.94, "learning_rate": 3.4337739132685084e-05, "loss": 0.0, "step": 48434 }, { "epoch": 0.94, "learning_rate": 3.4337092172427846e-05, "loss": 0.0, "step": 48436 }, { "epoch": 0.94, "learning_rate": 3.433644521217062e-05, "loss": 0.0002, "step": 48438 }, { "epoch": 0.94, "learning_rate": 3.4335798251913384e-05, "loss": 0.0, "step": 48440 }, { "epoch": 0.94, "learning_rate": 3.433515129165615e-05, "loss": 0.0024, "step": 48442 }, { "epoch": 0.94, "learning_rate": 3.433450433139892e-05, "loss": 0.0198, "step": 48444 }, { "epoch": 0.94, "learning_rate": 3.433385737114169e-05, "loss": 0.0, "step": 48446 }, { "epoch": 0.94, "learning_rate": 3.433321041088446e-05, "loss": 0.0, "step": 48448 }, { "epoch": 0.94, "learning_rate": 3.433256345062723e-05, "loss": 0.0001, "step": 48450 }, { "epoch": 0.94, "learning_rate": 3.433191649037e-05, "loss": 0.0, "step": 48452 }, { "epoch": 0.94, "learning_rate": 3.433126953011277e-05, "loss": 0.0, "step": 48454 }, { "epoch": 0.94, "learning_rate": 3.433062256985554e-05, "loss": 0.0, "step": 48456 }, { "epoch": 0.94, "learning_rate": 3.4329975609598306e-05, "loss": 0.0, "step": 48458 }, { "epoch": 0.94, "learning_rate": 3.4329328649341075e-05, "loss": 0.0001, "step": 48460 }, { "epoch": 0.94, "learning_rate": 3.4328681689083844e-05, "loss": 0.0, "step": 48462 }, { "epoch": 0.94, "learning_rate": 3.4328034728826606e-05, "loss": 0.0001, "step": 48464 }, { "epoch": 0.94, "learning_rate": 3.432738776856938e-05, "loss": 0.0002, "step": 48466 }, { "epoch": 0.94, "learning_rate": 3.4326740808312144e-05, "loss": 0.0021, "step": 48468 }, { "epoch": 0.94, "learning_rate": 3.4326093848054914e-05, "loss": 0.015, "step": 48470 }, { "epoch": 0.94, "learning_rate": 3.432544688779768e-05, "loss": 0.0, "step": 48472 }, { "epoch": 0.94, "learning_rate": 3.432479992754045e-05, "loss": 0.0, "step": 48474 }, { "epoch": 0.94, "learning_rate": 3.432415296728322e-05, "loss": 0.0002, "step": 48476 }, { "epoch": 0.94, "learning_rate": 3.432350600702599e-05, "loss": 0.0, "step": 48478 }, { "epoch": 0.94, "learning_rate": 3.432285904676876e-05, "loss": 0.0, "step": 48480 }, { "epoch": 0.94, "learning_rate": 3.432221208651153e-05, "loss": 0.0073, "step": 48482 }, { "epoch": 0.94, "learning_rate": 3.43215651262543e-05, "loss": 0.0017, "step": 48484 }, { "epoch": 0.94, "learning_rate": 3.432091816599706e-05, "loss": 0.0, "step": 48486 }, { "epoch": 0.94, "learning_rate": 3.4320271205739835e-05, "loss": 0.0003, "step": 48488 }, { "epoch": 0.94, "learning_rate": 3.4319624245482604e-05, "loss": 0.0001, "step": 48490 }, { "epoch": 0.94, "learning_rate": 3.431897728522537e-05, "loss": 0.0, "step": 48492 }, { "epoch": 0.94, "learning_rate": 3.431833032496814e-05, "loss": 0.0123, "step": 48494 }, { "epoch": 0.94, "learning_rate": 3.4317683364710905e-05, "loss": 0.0002, "step": 48496 }, { "epoch": 0.94, "learning_rate": 3.431703640445368e-05, "loss": 0.0, "step": 48498 }, { "epoch": 0.94, "learning_rate": 3.431638944419644e-05, "loss": 0.0001, "step": 48500 }, { "epoch": 0.94, "learning_rate": 3.431574248393921e-05, "loss": 0.0, "step": 48502 }, { "epoch": 0.94, "learning_rate": 3.431509552368198e-05, "loss": 0.0, "step": 48504 }, { "epoch": 0.94, "learning_rate": 3.431444856342475e-05, "loss": 0.0, "step": 48506 }, { "epoch": 0.94, "learning_rate": 3.431380160316752e-05, "loss": 0.001, "step": 48508 }, { "epoch": 0.94, "learning_rate": 3.431315464291029e-05, "loss": 0.0042, "step": 48510 }, { "epoch": 0.94, "learning_rate": 3.431250768265306e-05, "loss": 0.0, "step": 48512 }, { "epoch": 0.94, "learning_rate": 3.431186072239582e-05, "loss": 0.0001, "step": 48514 }, { "epoch": 0.94, "learning_rate": 3.4311213762138596e-05, "loss": 0.0, "step": 48516 }, { "epoch": 0.94, "learning_rate": 3.431056680188136e-05, "loss": 0.0, "step": 48518 }, { "epoch": 0.94, "learning_rate": 3.4309919841624134e-05, "loss": 0.0, "step": 48520 }, { "epoch": 0.94, "learning_rate": 3.4309272881366896e-05, "loss": 0.0066, "step": 48522 }, { "epoch": 0.94, "learning_rate": 3.4308625921109665e-05, "loss": 0.0004, "step": 48524 }, { "epoch": 0.94, "learning_rate": 3.430797896085244e-05, "loss": 0.0004, "step": 48526 }, { "epoch": 0.94, "learning_rate": 3.430765548072382e-05, "loss": 0.0225, "step": 48528 }, { "epoch": 0.94, "learning_rate": 3.430700852046659e-05, "loss": 0.0001, "step": 48530 }, { "epoch": 0.94, "learning_rate": 3.430636156020936e-05, "loss": 0.0, "step": 48532 }, { "epoch": 0.94, "learning_rate": 3.4305714599952126e-05, "loss": 0.0, "step": 48534 }, { "epoch": 0.94, "learning_rate": 3.4305067639694895e-05, "loss": 0.0001, "step": 48536 }, { "epoch": 0.94, "learning_rate": 3.4304420679437664e-05, "loss": 0.0001, "step": 48538 }, { "epoch": 0.94, "learning_rate": 3.4303773719180433e-05, "loss": 0.0002, "step": 48540 }, { "epoch": 0.94, "learning_rate": 3.4303126758923196e-05, "loss": 0.0001, "step": 48542 }, { "epoch": 0.94, "learning_rate": 3.430247979866597e-05, "loss": 0.0, "step": 48544 }, { "epoch": 0.94, "learning_rate": 3.4301832838408734e-05, "loss": 0.001, "step": 48546 }, { "epoch": 0.94, "learning_rate": 3.430118587815151e-05, "loss": 0.0023, "step": 48548 }, { "epoch": 0.94, "learning_rate": 3.430053891789428e-05, "loss": 0.0021, "step": 48550 }, { "epoch": 0.94, "learning_rate": 3.429989195763704e-05, "loss": 0.0, "step": 48552 }, { "epoch": 0.94, "learning_rate": 3.429924499737982e-05, "loss": 0.0001, "step": 48554 }, { "epoch": 0.94, "learning_rate": 3.429859803712258e-05, "loss": 0.0, "step": 48556 }, { "epoch": 0.94, "learning_rate": 3.429795107686535e-05, "loss": 0.0005, "step": 48558 }, { "epoch": 0.94, "learning_rate": 3.429730411660812e-05, "loss": 0.0003, "step": 48560 }, { "epoch": 0.94, "learning_rate": 3.4296657156350887e-05, "loss": 0.0001, "step": 48562 }, { "epoch": 0.94, "learning_rate": 3.4296010196093656e-05, "loss": 0.0003, "step": 48564 }, { "epoch": 0.94, "learning_rate": 3.4295363235836425e-05, "loss": 0.0014, "step": 48566 }, { "epoch": 0.94, "learning_rate": 3.4294716275579194e-05, "loss": 0.0059, "step": 48568 }, { "epoch": 0.94, "learning_rate": 3.429406931532196e-05, "loss": 0.0008, "step": 48570 }, { "epoch": 0.94, "learning_rate": 3.429342235506473e-05, "loss": 0.0033, "step": 48572 }, { "epoch": 0.94, "learning_rate": 3.4292775394807494e-05, "loss": 0.0263, "step": 48574 }, { "epoch": 0.94, "learning_rate": 3.429212843455027e-05, "loss": 0.0, "step": 48576 }, { "epoch": 0.94, "learning_rate": 3.429148147429303e-05, "loss": 0.0001, "step": 48578 }, { "epoch": 0.94, "learning_rate": 3.42908345140358e-05, "loss": 0.0, "step": 48580 }, { "epoch": 0.94, "learning_rate": 3.429018755377858e-05, "loss": 0.0001, "step": 48582 }, { "epoch": 0.94, "learning_rate": 3.428954059352134e-05, "loss": 0.0091, "step": 48584 }, { "epoch": 0.94, "learning_rate": 3.428889363326411e-05, "loss": 0.0002, "step": 48586 }, { "epoch": 0.94, "learning_rate": 3.428824667300688e-05, "loss": 0.0001, "step": 48588 }, { "epoch": 0.94, "learning_rate": 3.428759971274965e-05, "loss": 0.0, "step": 48590 }, { "epoch": 0.94, "learning_rate": 3.4286952752492416e-05, "loss": 0.0003, "step": 48592 }, { "epoch": 0.94, "learning_rate": 3.4286305792235185e-05, "loss": 0.0009, "step": 48594 }, { "epoch": 0.94, "learning_rate": 3.4285658831977954e-05, "loss": 0.0001, "step": 48596 }, { "epoch": 0.94, "learning_rate": 3.428501187172072e-05, "loss": 0.0, "step": 48598 }, { "epoch": 0.94, "learning_rate": 3.428436491146349e-05, "loss": 0.0, "step": 48600 }, { "epoch": 0.94, "learning_rate": 3.4283717951206255e-05, "loss": 0.0006, "step": 48602 }, { "epoch": 0.94, "learning_rate": 3.428307099094903e-05, "loss": 0.0, "step": 48604 }, { "epoch": 0.94, "learning_rate": 3.428242403069179e-05, "loss": 0.0, "step": 48606 }, { "epoch": 0.94, "learning_rate": 3.428177707043457e-05, "loss": 0.0, "step": 48608 }, { "epoch": 0.94, "learning_rate": 3.428113011017733e-05, "loss": 0.0, "step": 48610 }, { "epoch": 0.94, "learning_rate": 3.42804831499201e-05, "loss": 0.0, "step": 48612 }, { "epoch": 0.94, "learning_rate": 3.4279836189662876e-05, "loss": 0.0033, "step": 48614 }, { "epoch": 0.94, "learning_rate": 3.427918922940564e-05, "loss": 0.0152, "step": 48616 }, { "epoch": 0.94, "learning_rate": 3.427854226914841e-05, "loss": 0.0, "step": 48618 }, { "epoch": 0.94, "learning_rate": 3.4277895308891176e-05, "loss": 0.0002, "step": 48620 }, { "epoch": 0.94, "learning_rate": 3.4277248348633946e-05, "loss": 0.0014, "step": 48622 }, { "epoch": 0.94, "learning_rate": 3.427660138837671e-05, "loss": 0.0, "step": 48624 }, { "epoch": 0.94, "learning_rate": 3.4275954428119484e-05, "loss": 0.0, "step": 48626 }, { "epoch": 0.94, "learning_rate": 3.427530746786225e-05, "loss": 0.0, "step": 48628 }, { "epoch": 0.94, "learning_rate": 3.427466050760502e-05, "loss": 0.002, "step": 48630 }, { "epoch": 0.94, "learning_rate": 3.427401354734779e-05, "loss": 0.0, "step": 48632 }, { "epoch": 0.94, "learning_rate": 3.427336658709055e-05, "loss": 0.0, "step": 48634 }, { "epoch": 0.94, "learning_rate": 3.427271962683333e-05, "loss": 0.0099, "step": 48636 }, { "epoch": 0.94, "learning_rate": 3.427207266657609e-05, "loss": 0.0, "step": 48638 }, { "epoch": 0.94, "learning_rate": 3.427142570631886e-05, "loss": 0.0, "step": 48640 }, { "epoch": 0.94, "learning_rate": 3.427077874606163e-05, "loss": 0.0064, "step": 48642 }, { "epoch": 0.94, "learning_rate": 3.42701317858044e-05, "loss": 0.0019, "step": 48644 }, { "epoch": 0.94, "learning_rate": 3.426948482554717e-05, "loss": 0.0029, "step": 48646 }, { "epoch": 0.94, "learning_rate": 3.426883786528994e-05, "loss": 0.0, "step": 48648 }, { "epoch": 0.94, "learning_rate": 3.4268190905032706e-05, "loss": 0.0, "step": 48650 }, { "epoch": 0.94, "learning_rate": 3.4267543944775475e-05, "loss": 0.0002, "step": 48652 }, { "epoch": 0.94, "learning_rate": 3.4266896984518244e-05, "loss": 0.0015, "step": 48654 }, { "epoch": 0.94, "learning_rate": 3.4266250024261006e-05, "loss": 0.0, "step": 48656 }, { "epoch": 0.94, "learning_rate": 3.426560306400378e-05, "loss": 0.0, "step": 48658 }, { "epoch": 0.94, "learning_rate": 3.426495610374655e-05, "loss": 0.0001, "step": 48660 }, { "epoch": 0.94, "learning_rate": 3.4264309143489314e-05, "loss": 0.0, "step": 48662 }, { "epoch": 0.94, "learning_rate": 3.426366218323209e-05, "loss": 0.0, "step": 48664 }, { "epoch": 0.94, "learning_rate": 3.426301522297485e-05, "loss": 0.0014, "step": 48666 }, { "epoch": 0.94, "learning_rate": 3.426236826271762e-05, "loss": 0.0, "step": 48668 }, { "epoch": 0.94, "learning_rate": 3.426172130246039e-05, "loss": 0.0, "step": 48670 }, { "epoch": 0.94, "learning_rate": 3.426107434220316e-05, "loss": 0.0001, "step": 48672 }, { "epoch": 0.94, "learning_rate": 3.426042738194593e-05, "loss": 0.0, "step": 48674 }, { "epoch": 0.94, "learning_rate": 3.42597804216887e-05, "loss": 0.0001, "step": 48676 }, { "epoch": 0.94, "learning_rate": 3.4259133461431466e-05, "loss": 0.001, "step": 48678 }, { "epoch": 0.94, "learning_rate": 3.4258486501174235e-05, "loss": 0.0009, "step": 48680 }, { "epoch": 0.94, "learning_rate": 3.4257839540917005e-05, "loss": 0.0, "step": 48682 }, { "epoch": 0.94, "learning_rate": 3.425719258065977e-05, "loss": 0.0, "step": 48684 }, { "epoch": 0.94, "learning_rate": 3.425654562040254e-05, "loss": 0.0, "step": 48686 }, { "epoch": 0.94, "learning_rate": 3.4255898660145305e-05, "loss": 0.0001, "step": 48688 }, { "epoch": 0.95, "learning_rate": 3.425525169988808e-05, "loss": 0.0094, "step": 48690 }, { "epoch": 0.95, "learning_rate": 3.425460473963085e-05, "loss": 0.0, "step": 48692 }, { "epoch": 0.95, "learning_rate": 3.425395777937361e-05, "loss": 0.0, "step": 48694 }, { "epoch": 0.95, "learning_rate": 3.425331081911639e-05, "loss": 0.0002, "step": 48696 }, { "epoch": 0.95, "learning_rate": 3.425266385885915e-05, "loss": 0.0, "step": 48698 }, { "epoch": 0.95, "learning_rate": 3.425201689860192e-05, "loss": 0.0, "step": 48700 }, { "epoch": 0.95, "learning_rate": 3.425136993834469e-05, "loss": 0.0, "step": 48702 }, { "epoch": 0.95, "learning_rate": 3.425072297808746e-05, "loss": 0.0, "step": 48704 }, { "epoch": 0.95, "learning_rate": 3.425007601783023e-05, "loss": 0.0, "step": 48706 }, { "epoch": 0.95, "learning_rate": 3.4249429057572996e-05, "loss": 0.0, "step": 48708 }, { "epoch": 0.95, "learning_rate": 3.4248782097315765e-05, "loss": 0.0045, "step": 48710 }, { "epoch": 0.95, "learning_rate": 3.4248135137058534e-05, "loss": 0.0, "step": 48712 }, { "epoch": 0.95, "learning_rate": 3.42474881768013e-05, "loss": 0.0, "step": 48714 }, { "epoch": 0.95, "learning_rate": 3.4246841216544065e-05, "loss": 0.0002, "step": 48716 }, { "epoch": 0.95, "learning_rate": 3.424619425628684e-05, "loss": 0.0, "step": 48718 }, { "epoch": 0.95, "learning_rate": 3.4245547296029604e-05, "loss": 0.0, "step": 48720 }, { "epoch": 0.95, "learning_rate": 3.424490033577237e-05, "loss": 0.0003, "step": 48722 }, { "epoch": 0.95, "learning_rate": 3.424425337551514e-05, "loss": 0.0114, "step": 48724 }, { "epoch": 0.95, "learning_rate": 3.424360641525791e-05, "loss": 0.0001, "step": 48726 }, { "epoch": 0.95, "learning_rate": 3.424295945500068e-05, "loss": 0.0072, "step": 48728 }, { "epoch": 0.95, "learning_rate": 3.424231249474345e-05, "loss": 0.0046, "step": 48730 }, { "epoch": 0.95, "learning_rate": 3.424166553448622e-05, "loss": 0.0001, "step": 48732 }, { "epoch": 0.95, "learning_rate": 3.424101857422899e-05, "loss": 0.0, "step": 48734 }, { "epoch": 0.95, "learning_rate": 3.4240371613971756e-05, "loss": 0.0002, "step": 48736 }, { "epoch": 0.95, "learning_rate": 3.4239724653714525e-05, "loss": 0.003, "step": 48738 }, { "epoch": 0.95, "learning_rate": 3.4239077693457294e-05, "loss": 0.0, "step": 48740 }, { "epoch": 0.95, "learning_rate": 3.4238430733200064e-05, "loss": 0.0, "step": 48742 }, { "epoch": 0.95, "learning_rate": 3.4237783772942826e-05, "loss": 0.0, "step": 48744 }, { "epoch": 0.95, "learning_rate": 3.42371368126856e-05, "loss": 0.0014, "step": 48746 }, { "epoch": 0.95, "learning_rate": 3.4236489852428364e-05, "loss": 0.0, "step": 48748 }, { "epoch": 0.95, "learning_rate": 3.423584289217114e-05, "loss": 0.0, "step": 48750 }, { "epoch": 0.95, "learning_rate": 3.42351959319139e-05, "loss": 0.0, "step": 48752 }, { "epoch": 0.95, "learning_rate": 3.423454897165667e-05, "loss": 0.0016, "step": 48754 }, { "epoch": 0.95, "learning_rate": 3.423390201139944e-05, "loss": 0.0, "step": 48756 }, { "epoch": 0.95, "learning_rate": 3.423325505114221e-05, "loss": 0.0, "step": 48758 }, { "epoch": 0.95, "learning_rate": 3.423260809088498e-05, "loss": 0.0175, "step": 48760 }, { "epoch": 0.95, "learning_rate": 3.423196113062775e-05, "loss": 0.0006, "step": 48762 }, { "epoch": 0.95, "learning_rate": 3.423131417037052e-05, "loss": 0.0, "step": 48764 }, { "epoch": 0.95, "learning_rate": 3.423066721011328e-05, "loss": 0.0022, "step": 48766 }, { "epoch": 0.95, "learning_rate": 3.4230020249856055e-05, "loss": 0.0, "step": 48768 }, { "epoch": 0.95, "learning_rate": 3.4229373289598824e-05, "loss": 0.0, "step": 48770 }, { "epoch": 0.95, "learning_rate": 3.422872632934159e-05, "loss": 0.0001, "step": 48772 }, { "epoch": 0.95, "learning_rate": 3.422807936908436e-05, "loss": 0.0188, "step": 48774 }, { "epoch": 0.95, "learning_rate": 3.4227432408827124e-05, "loss": 0.0, "step": 48776 }, { "epoch": 0.95, "learning_rate": 3.42267854485699e-05, "loss": 0.0067, "step": 48778 }, { "epoch": 0.95, "learning_rate": 3.422613848831266e-05, "loss": 0.0047, "step": 48780 }, { "epoch": 0.95, "learning_rate": 3.422549152805543e-05, "loss": 0.0002, "step": 48782 }, { "epoch": 0.95, "learning_rate": 3.42248445677982e-05, "loss": 0.0, "step": 48784 }, { "epoch": 0.95, "learning_rate": 3.422419760754097e-05, "loss": 0.0015, "step": 48786 }, { "epoch": 0.95, "learning_rate": 3.422355064728374e-05, "loss": 0.0008, "step": 48788 }, { "epoch": 0.95, "learning_rate": 3.422290368702651e-05, "loss": 0.0, "step": 48790 }, { "epoch": 0.95, "learning_rate": 3.422225672676928e-05, "loss": 0.0002, "step": 48792 }, { "epoch": 0.95, "learning_rate": 3.4221609766512046e-05, "loss": 0.0, "step": 48794 }, { "epoch": 0.95, "learning_rate": 3.4220962806254815e-05, "loss": 0.0, "step": 48796 }, { "epoch": 0.95, "learning_rate": 3.422031584599758e-05, "loss": 0.0021, "step": 48798 }, { "epoch": 0.95, "learning_rate": 3.4219668885740353e-05, "loss": 0.0014, "step": 48800 }, { "epoch": 0.95, "learning_rate": 3.4219021925483116e-05, "loss": 0.0001, "step": 48802 }, { "epoch": 0.95, "learning_rate": 3.4218374965225885e-05, "loss": 0.0, "step": 48804 }, { "epoch": 0.95, "learning_rate": 3.421772800496866e-05, "loss": 0.0, "step": 48806 }, { "epoch": 0.95, "learning_rate": 3.421708104471142e-05, "loss": 0.0065, "step": 48808 }, { "epoch": 0.95, "learning_rate": 3.421643408445419e-05, "loss": 0.0, "step": 48810 }, { "epoch": 0.95, "learning_rate": 3.421578712419696e-05, "loss": 0.0003, "step": 48812 }, { "epoch": 0.95, "learning_rate": 3.421514016393973e-05, "loss": 0.0, "step": 48814 }, { "epoch": 0.95, "learning_rate": 3.42144932036825e-05, "loss": 0.0, "step": 48816 }, { "epoch": 0.95, "learning_rate": 3.421384624342527e-05, "loss": 0.0, "step": 48818 }, { "epoch": 0.95, "learning_rate": 3.421319928316804e-05, "loss": 0.0, "step": 48820 }, { "epoch": 0.95, "learning_rate": 3.4212552322910807e-05, "loss": 0.0009, "step": 48822 }, { "epoch": 0.95, "learning_rate": 3.4211905362653576e-05, "loss": 0.0001, "step": 48824 }, { "epoch": 0.95, "learning_rate": 3.421125840239634e-05, "loss": 0.0, "step": 48826 }, { "epoch": 0.95, "learning_rate": 3.4210611442139114e-05, "loss": 0.0001, "step": 48828 }, { "epoch": 0.95, "learning_rate": 3.4209964481881876e-05, "loss": 0.0, "step": 48830 }, { "epoch": 0.95, "learning_rate": 3.420931752162465e-05, "loss": 0.0001, "step": 48832 }, { "epoch": 0.95, "learning_rate": 3.4208670561367414e-05, "loss": 0.0, "step": 48834 }, { "epoch": 0.95, "learning_rate": 3.420802360111018e-05, "loss": 0.0, "step": 48836 }, { "epoch": 0.95, "learning_rate": 3.420737664085296e-05, "loss": 0.0015, "step": 48838 }, { "epoch": 0.95, "learning_rate": 3.420672968059572e-05, "loss": 0.0016, "step": 48840 }, { "epoch": 0.95, "learning_rate": 3.420608272033849e-05, "loss": 0.003, "step": 48842 }, { "epoch": 0.95, "learning_rate": 3.420543576008126e-05, "loss": 0.0, "step": 48844 }, { "epoch": 0.95, "learning_rate": 3.420478879982403e-05, "loss": 0.0002, "step": 48846 }, { "epoch": 0.95, "learning_rate": 3.42041418395668e-05, "loss": 0.0193, "step": 48848 }, { "epoch": 0.95, "learning_rate": 3.420349487930957e-05, "loss": 0.0, "step": 48850 }, { "epoch": 0.95, "learning_rate": 3.4202847919052336e-05, "loss": 0.0, "step": 48852 }, { "epoch": 0.95, "learning_rate": 3.4202200958795105e-05, "loss": 0.0001, "step": 48854 }, { "epoch": 0.95, "learning_rate": 3.4201553998537874e-05, "loss": 0.0, "step": 48856 }, { "epoch": 0.95, "learning_rate": 3.4200907038280637e-05, "loss": 0.0, "step": 48858 }, { "epoch": 0.95, "learning_rate": 3.420026007802341e-05, "loss": 0.0001, "step": 48860 }, { "epoch": 0.95, "learning_rate": 3.4199613117766175e-05, "loss": 0.0002, "step": 48862 }, { "epoch": 0.95, "learning_rate": 3.4198966157508944e-05, "loss": 0.0002, "step": 48864 }, { "epoch": 0.95, "learning_rate": 3.419831919725171e-05, "loss": 0.0, "step": 48866 }, { "epoch": 0.95, "learning_rate": 3.419767223699448e-05, "loss": 0.0001, "step": 48868 }, { "epoch": 0.95, "learning_rate": 3.419702527673725e-05, "loss": 0.0, "step": 48870 }, { "epoch": 0.95, "learning_rate": 3.419637831648002e-05, "loss": 0.0, "step": 48872 }, { "epoch": 0.95, "learning_rate": 3.419573135622279e-05, "loss": 0.0, "step": 48874 }, { "epoch": 0.95, "learning_rate": 3.419508439596556e-05, "loss": 0.0051, "step": 48876 }, { "epoch": 0.95, "learning_rate": 3.419443743570833e-05, "loss": 0.0, "step": 48878 }, { "epoch": 0.95, "learning_rate": 3.419379047545109e-05, "loss": 0.0, "step": 48880 }, { "epoch": 0.95, "learning_rate": 3.4193143515193866e-05, "loss": 0.006, "step": 48882 }, { "epoch": 0.95, "learning_rate": 3.4192496554936635e-05, "loss": 0.0, "step": 48884 }, { "epoch": 0.95, "learning_rate": 3.41918495946794e-05, "loss": 0.0, "step": 48886 }, { "epoch": 0.95, "learning_rate": 3.419120263442217e-05, "loss": 0.0, "step": 48888 }, { "epoch": 0.95, "learning_rate": 3.4190555674164935e-05, "loss": 0.0004, "step": 48890 }, { "epoch": 0.95, "learning_rate": 3.418990871390771e-05, "loss": 0.0, "step": 48892 }, { "epoch": 0.95, "learning_rate": 3.418926175365047e-05, "loss": 0.0, "step": 48894 }, { "epoch": 0.95, "learning_rate": 3.418861479339324e-05, "loss": 0.0003, "step": 48896 }, { "epoch": 0.95, "learning_rate": 3.418796783313601e-05, "loss": 0.0061, "step": 48898 }, { "epoch": 0.95, "learning_rate": 3.418732087287878e-05, "loss": 0.0045, "step": 48900 }, { "epoch": 0.95, "learning_rate": 3.418667391262155e-05, "loss": 0.0001, "step": 48902 }, { "epoch": 0.95, "learning_rate": 3.418602695236432e-05, "loss": 0.0, "step": 48904 }, { "epoch": 0.95, "learning_rate": 3.418537999210709e-05, "loss": 0.0, "step": 48906 }, { "epoch": 0.95, "learning_rate": 3.418473303184985e-05, "loss": 0.0001, "step": 48908 }, { "epoch": 0.95, "learning_rate": 3.4184086071592626e-05, "loss": 0.0, "step": 48910 }, { "epoch": 0.95, "learning_rate": 3.418343911133539e-05, "loss": 0.0013, "step": 48912 }, { "epoch": 0.95, "learning_rate": 3.4182792151078164e-05, "loss": 0.0008, "step": 48914 }, { "epoch": 0.95, "learning_rate": 3.418214519082093e-05, "loss": 0.0, "step": 48916 }, { "epoch": 0.95, "learning_rate": 3.4181498230563696e-05, "loss": 0.0, "step": 48918 }, { "epoch": 0.95, "learning_rate": 3.418085127030647e-05, "loss": 0.0, "step": 48920 }, { "epoch": 0.95, "learning_rate": 3.4180204310049234e-05, "loss": 0.0001, "step": 48922 }, { "epoch": 0.95, "learning_rate": 3.4179557349792e-05, "loss": 0.0004, "step": 48924 }, { "epoch": 0.95, "learning_rate": 3.417891038953477e-05, "loss": 0.0049, "step": 48926 }, { "epoch": 0.95, "learning_rate": 3.417826342927754e-05, "loss": 0.0, "step": 48928 }, { "epoch": 0.95, "learning_rate": 3.417761646902031e-05, "loss": 0.0, "step": 48930 }, { "epoch": 0.95, "learning_rate": 3.417696950876308e-05, "loss": 0.0002, "step": 48932 }, { "epoch": 0.95, "learning_rate": 3.417632254850585e-05, "loss": 0.0, "step": 48934 }, { "epoch": 0.95, "learning_rate": 3.417567558824862e-05, "loss": 0.0119, "step": 48936 }, { "epoch": 0.95, "learning_rate": 3.4175028627991386e-05, "loss": 0.0, "step": 48938 }, { "epoch": 0.95, "learning_rate": 3.417438166773415e-05, "loss": 0.0, "step": 48940 }, { "epoch": 0.95, "learning_rate": 3.4173734707476925e-05, "loss": 0.0, "step": 48942 }, { "epoch": 0.95, "learning_rate": 3.417308774721969e-05, "loss": 0.0, "step": 48944 }, { "epoch": 0.95, "learning_rate": 3.4172440786962456e-05, "loss": 0.0, "step": 48946 }, { "epoch": 0.95, "learning_rate": 3.417179382670523e-05, "loss": 0.0001, "step": 48948 }, { "epoch": 0.95, "learning_rate": 3.4171146866447994e-05, "loss": 0.0, "step": 48950 }, { "epoch": 0.95, "learning_rate": 3.417049990619076e-05, "loss": 0.0, "step": 48952 }, { "epoch": 0.95, "learning_rate": 3.416985294593353e-05, "loss": 0.0, "step": 48954 }, { "epoch": 0.95, "learning_rate": 3.41692059856763e-05, "loss": 0.0001, "step": 48956 }, { "epoch": 0.95, "learning_rate": 3.416855902541907e-05, "loss": 0.0, "step": 48958 }, { "epoch": 0.95, "learning_rate": 3.416791206516184e-05, "loss": 0.0016, "step": 48960 }, { "epoch": 0.95, "learning_rate": 3.416726510490461e-05, "loss": 0.0, "step": 48962 }, { "epoch": 0.95, "learning_rate": 3.416661814464738e-05, "loss": 0.0005, "step": 48964 }, { "epoch": 0.95, "learning_rate": 3.416597118439015e-05, "loss": 0.0057, "step": 48966 }, { "epoch": 0.95, "learning_rate": 3.416532422413291e-05, "loss": 0.0003, "step": 48968 }, { "epoch": 0.95, "learning_rate": 3.4164677263875685e-05, "loss": 0.0037, "step": 48970 }, { "epoch": 0.95, "learning_rate": 3.416403030361845e-05, "loss": 0.0, "step": 48972 }, { "epoch": 0.95, "learning_rate": 3.416338334336122e-05, "loss": 0.0038, "step": 48974 }, { "epoch": 0.95, "learning_rate": 3.4162736383103985e-05, "loss": 0.0, "step": 48976 }, { "epoch": 0.95, "learning_rate": 3.4162089422846754e-05, "loss": 0.0, "step": 48978 }, { "epoch": 0.95, "learning_rate": 3.4161442462589524e-05, "loss": 0.0, "step": 48980 }, { "epoch": 0.95, "learning_rate": 3.416079550233229e-05, "loss": 0.0001, "step": 48982 }, { "epoch": 0.95, "learning_rate": 3.416014854207506e-05, "loss": 0.0, "step": 48984 }, { "epoch": 0.95, "learning_rate": 3.415950158181783e-05, "loss": 0.0, "step": 48986 }, { "epoch": 0.95, "learning_rate": 3.41588546215606e-05, "loss": 0.0076, "step": 48988 }, { "epoch": 0.95, "learning_rate": 3.415820766130336e-05, "loss": 0.0, "step": 48990 }, { "epoch": 0.95, "learning_rate": 3.415756070104614e-05, "loss": 0.0, "step": 48992 }, { "epoch": 0.95, "learning_rate": 3.415691374078891e-05, "loss": 0.0055, "step": 48994 }, { "epoch": 0.95, "learning_rate": 3.4156266780531676e-05, "loss": 0.0, "step": 48996 }, { "epoch": 0.95, "learning_rate": 3.4155619820274445e-05, "loss": 0.0, "step": 48998 }, { "epoch": 0.95, "learning_rate": 3.415497286001721e-05, "loss": 0.0001, "step": 49000 }, { "epoch": 0.95, "learning_rate": 3.4154325899759984e-05, "loss": 0.0001, "step": 49002 }, { "epoch": 0.95, "learning_rate": 3.4153678939502746e-05, "loss": 0.0, "step": 49004 }, { "epoch": 0.95, "learning_rate": 3.4153031979245515e-05, "loss": 0.0, "step": 49006 }, { "epoch": 0.95, "learning_rate": 3.4152385018988284e-05, "loss": 0.0, "step": 49008 }, { "epoch": 0.95, "learning_rate": 3.415173805873105e-05, "loss": 0.0001, "step": 49010 }, { "epoch": 0.95, "learning_rate": 3.415109109847382e-05, "loss": 0.0002, "step": 49012 }, { "epoch": 0.95, "learning_rate": 3.415044413821659e-05, "loss": 0.004, "step": 49014 }, { "epoch": 0.95, "learning_rate": 3.414979717795936e-05, "loss": 0.0, "step": 49016 }, { "epoch": 0.95, "learning_rate": 3.414915021770213e-05, "loss": 0.0001, "step": 49018 }, { "epoch": 0.95, "learning_rate": 3.41485032574449e-05, "loss": 0.0, "step": 49020 }, { "epoch": 0.95, "learning_rate": 3.414785629718766e-05, "loss": 0.0, "step": 49022 }, { "epoch": 0.95, "learning_rate": 3.414720933693044e-05, "loss": 0.0, "step": 49024 }, { "epoch": 0.95, "learning_rate": 3.41465623766732e-05, "loss": 0.0, "step": 49026 }, { "epoch": 0.95, "learning_rate": 3.414591541641597e-05, "loss": 0.0054, "step": 49028 }, { "epoch": 0.95, "learning_rate": 3.4145268456158744e-05, "loss": 0.0004, "step": 49030 }, { "epoch": 0.95, "learning_rate": 3.4144621495901506e-05, "loss": 0.0, "step": 49032 }, { "epoch": 0.95, "learning_rate": 3.414397453564428e-05, "loss": 0.0107, "step": 49034 }, { "epoch": 0.95, "learning_rate": 3.4143327575387044e-05, "loss": 0.0, "step": 49036 }, { "epoch": 0.95, "learning_rate": 3.4142680615129813e-05, "loss": 0.0017, "step": 49038 }, { "epoch": 0.95, "learning_rate": 3.414203365487258e-05, "loss": 0.0004, "step": 49040 }, { "epoch": 0.95, "learning_rate": 3.414138669461535e-05, "loss": 0.0, "step": 49042 }, { "epoch": 0.95, "learning_rate": 3.414073973435812e-05, "loss": 0.0001, "step": 49044 }, { "epoch": 0.95, "learning_rate": 3.414009277410089e-05, "loss": 0.0155, "step": 49046 }, { "epoch": 0.95, "learning_rate": 3.413944581384366e-05, "loss": 0.0047, "step": 49048 }, { "epoch": 0.95, "learning_rate": 3.413879885358642e-05, "loss": 0.0, "step": 49050 }, { "epoch": 0.95, "learning_rate": 3.41381518933292e-05, "loss": 0.0, "step": 49052 }, { "epoch": 0.95, "learning_rate": 3.413750493307196e-05, "loss": 0.0007, "step": 49054 }, { "epoch": 0.95, "learning_rate": 3.4136857972814735e-05, "loss": 0.0, "step": 49056 }, { "epoch": 0.95, "learning_rate": 3.41362110125575e-05, "loss": 0.0001, "step": 49058 }, { "epoch": 0.95, "learning_rate": 3.413556405230027e-05, "loss": 0.0001, "step": 49060 }, { "epoch": 0.95, "learning_rate": 3.413491709204304e-05, "loss": 0.0, "step": 49062 }, { "epoch": 0.95, "learning_rate": 3.4134270131785805e-05, "loss": 0.0, "step": 49064 }, { "epoch": 0.95, "learning_rate": 3.4133623171528574e-05, "loss": 0.0114, "step": 49066 }, { "epoch": 0.95, "learning_rate": 3.413297621127134e-05, "loss": 0.0, "step": 49068 }, { "epoch": 0.95, "learning_rate": 3.413232925101411e-05, "loss": 0.0, "step": 49070 }, { "epoch": 0.95, "learning_rate": 3.413168229075688e-05, "loss": 0.0001, "step": 49072 }, { "epoch": 0.95, "learning_rate": 3.413103533049965e-05, "loss": 0.0002, "step": 49074 }, { "epoch": 0.95, "learning_rate": 3.413038837024242e-05, "loss": 0.0, "step": 49076 }, { "epoch": 0.95, "learning_rate": 3.412974140998519e-05, "loss": 0.0, "step": 49078 }, { "epoch": 0.95, "learning_rate": 3.412909444972796e-05, "loss": 0.0, "step": 49080 }, { "epoch": 0.95, "learning_rate": 3.412844748947072e-05, "loss": 0.0001, "step": 49082 }, { "epoch": 0.95, "learning_rate": 3.4127800529213496e-05, "loss": 0.0003, "step": 49084 }, { "epoch": 0.95, "learning_rate": 3.412715356895626e-05, "loss": 0.0, "step": 49086 }, { "epoch": 0.95, "learning_rate": 3.412650660869903e-05, "loss": 0.0001, "step": 49088 }, { "epoch": 0.95, "learning_rate": 3.4125859648441796e-05, "loss": 0.0, "step": 49090 }, { "epoch": 0.95, "learning_rate": 3.4125212688184565e-05, "loss": 0.0001, "step": 49092 }, { "epoch": 0.95, "learning_rate": 3.4124565727927334e-05, "loss": 0.0213, "step": 49094 }, { "epoch": 0.95, "learning_rate": 3.41239187676701e-05, "loss": 0.0004, "step": 49096 }, { "epoch": 0.95, "learning_rate": 3.412327180741287e-05, "loss": 0.0001, "step": 49098 }, { "epoch": 0.95, "learning_rate": 3.412262484715564e-05, "loss": 0.0118, "step": 49100 }, { "epoch": 0.95, "learning_rate": 3.412197788689841e-05, "loss": 0.0, "step": 49102 }, { "epoch": 0.95, "learning_rate": 3.412133092664117e-05, "loss": 0.0, "step": 49104 }, { "epoch": 0.95, "learning_rate": 3.412068396638395e-05, "loss": 0.0, "step": 49106 }, { "epoch": 0.95, "learning_rate": 3.412003700612672e-05, "loss": 0.0, "step": 49108 }, { "epoch": 0.95, "learning_rate": 3.411939004586948e-05, "loss": 0.0066, "step": 49110 }, { "epoch": 0.95, "learning_rate": 3.4118743085612256e-05, "loss": 0.0613, "step": 49112 }, { "epoch": 0.95, "learning_rate": 3.411809612535502e-05, "loss": 0.0001, "step": 49114 }, { "epoch": 0.95, "learning_rate": 3.4117449165097794e-05, "loss": 0.0, "step": 49116 }, { "epoch": 0.95, "learning_rate": 3.4116802204840557e-05, "loss": 0.0026, "step": 49118 }, { "epoch": 0.95, "learning_rate": 3.4116155244583326e-05, "loss": 0.0, "step": 49120 }, { "epoch": 0.95, "learning_rate": 3.4115508284326095e-05, "loss": 0.0064, "step": 49122 }, { "epoch": 0.95, "learning_rate": 3.4114861324068864e-05, "loss": 0.0, "step": 49124 }, { "epoch": 0.95, "learning_rate": 3.411421436381163e-05, "loss": 0.0008, "step": 49126 }, { "epoch": 0.95, "learning_rate": 3.41135674035544e-05, "loss": 0.0, "step": 49128 }, { "epoch": 0.95, "learning_rate": 3.411292044329717e-05, "loss": 0.0001, "step": 49130 }, { "epoch": 0.95, "learning_rate": 3.411227348303993e-05, "loss": 0.0, "step": 49132 }, { "epoch": 0.95, "learning_rate": 3.411162652278271e-05, "loss": 0.0, "step": 49134 }, { "epoch": 0.95, "learning_rate": 3.411097956252547e-05, "loss": 0.0, "step": 49136 }, { "epoch": 0.95, "learning_rate": 3.411033260226825e-05, "loss": 0.0, "step": 49138 }, { "epoch": 0.95, "learning_rate": 3.4109685642011016e-05, "loss": 0.0137, "step": 49140 }, { "epoch": 0.95, "learning_rate": 3.410903868175378e-05, "loss": 0.0, "step": 49142 }, { "epoch": 0.95, "learning_rate": 3.4108391721496555e-05, "loss": 0.0, "step": 49144 }, { "epoch": 0.95, "learning_rate": 3.410774476123932e-05, "loss": 0.0, "step": 49146 }, { "epoch": 0.95, "learning_rate": 3.4107097800982086e-05, "loss": 0.0, "step": 49148 }, { "epoch": 0.95, "learning_rate": 3.4106450840724855e-05, "loss": 0.0054, "step": 49150 }, { "epoch": 0.95, "learning_rate": 3.4105803880467624e-05, "loss": 0.0, "step": 49152 }, { "epoch": 0.95, "learning_rate": 3.410515692021039e-05, "loss": 0.0075, "step": 49154 }, { "epoch": 0.95, "learning_rate": 3.410450995995316e-05, "loss": 0.0, "step": 49156 }, { "epoch": 0.95, "learning_rate": 3.410386299969593e-05, "loss": 0.0, "step": 49158 }, { "epoch": 0.95, "learning_rate": 3.41032160394387e-05, "loss": 0.0, "step": 49160 }, { "epoch": 0.95, "learning_rate": 3.410256907918147e-05, "loss": 0.0005, "step": 49162 }, { "epoch": 0.95, "learning_rate": 3.410192211892423e-05, "loss": 0.0001, "step": 49164 }, { "epoch": 0.95, "learning_rate": 3.410127515866701e-05, "loss": 0.0, "step": 49166 }, { "epoch": 0.95, "learning_rate": 3.410062819840977e-05, "loss": 0.0013, "step": 49168 }, { "epoch": 0.95, "learning_rate": 3.409998123815254e-05, "loss": 0.0002, "step": 49170 }, { "epoch": 0.95, "learning_rate": 3.4099334277895315e-05, "loss": 0.0, "step": 49172 }, { "epoch": 0.95, "learning_rate": 3.409868731763808e-05, "loss": 0.0002, "step": 49174 }, { "epoch": 0.95, "learning_rate": 3.409804035738085e-05, "loss": 0.0001, "step": 49176 }, { "epoch": 0.95, "learning_rate": 3.4097393397123615e-05, "loss": 0.0, "step": 49178 }, { "epoch": 0.95, "learning_rate": 3.4096746436866385e-05, "loss": 0.0001, "step": 49180 }, { "epoch": 0.95, "learning_rate": 3.4096099476609154e-05, "loss": 0.0, "step": 49182 }, { "epoch": 0.95, "learning_rate": 3.409545251635192e-05, "loss": 0.0, "step": 49184 }, { "epoch": 0.95, "learning_rate": 3.409480555609469e-05, "loss": 0.0001, "step": 49186 }, { "epoch": 0.95, "learning_rate": 3.409415859583746e-05, "loss": 0.0006, "step": 49188 }, { "epoch": 0.95, "learning_rate": 3.409351163558023e-05, "loss": 0.0, "step": 49190 }, { "epoch": 0.95, "learning_rate": 3.409286467532299e-05, "loss": 0.0, "step": 49192 }, { "epoch": 0.95, "learning_rate": 3.409221771506577e-05, "loss": 0.0, "step": 49194 }, { "epoch": 0.95, "learning_rate": 3.409157075480853e-05, "loss": 0.0001, "step": 49196 }, { "epoch": 0.95, "learning_rate": 3.4090923794551306e-05, "loss": 0.0, "step": 49198 }, { "epoch": 0.95, "learning_rate": 3.409027683429407e-05, "loss": 0.0, "step": 49200 }, { "epoch": 0.95, "learning_rate": 3.408962987403684e-05, "loss": 0.0001, "step": 49202 }, { "epoch": 0.95, "learning_rate": 3.408898291377961e-05, "loss": 0.0264, "step": 49204 }, { "epoch": 0.96, "learning_rate": 3.4088335953522376e-05, "loss": 0.0, "step": 49206 }, { "epoch": 0.96, "learning_rate": 3.4087688993265145e-05, "loss": 0.0165, "step": 49208 }, { "epoch": 0.96, "learning_rate": 3.4087042033007914e-05, "loss": 0.0001, "step": 49210 }, { "epoch": 0.96, "learning_rate": 3.408639507275068e-05, "loss": 0.0, "step": 49212 }, { "epoch": 0.96, "learning_rate": 3.4085748112493445e-05, "loss": 0.0012, "step": 49214 }, { "epoch": 0.96, "learning_rate": 3.408510115223622e-05, "loss": 0.0, "step": 49216 }, { "epoch": 0.96, "learning_rate": 3.408445419197899e-05, "loss": 0.0, "step": 49218 }, { "epoch": 0.96, "learning_rate": 3.408380723172176e-05, "loss": 0.0001, "step": 49220 }, { "epoch": 0.96, "learning_rate": 3.408316027146453e-05, "loss": 0.0, "step": 49222 }, { "epoch": 0.96, "learning_rate": 3.408251331120729e-05, "loss": 0.0, "step": 49224 }, { "epoch": 0.96, "learning_rate": 3.408186635095007e-05, "loss": 0.0, "step": 49226 }, { "epoch": 0.96, "learning_rate": 3.408121939069283e-05, "loss": 0.0008, "step": 49228 }, { "epoch": 0.96, "learning_rate": 3.40805724304356e-05, "loss": 0.0, "step": 49230 }, { "epoch": 0.96, "learning_rate": 3.407992547017837e-05, "loss": 0.0038, "step": 49232 }, { "epoch": 0.96, "learning_rate": 3.4079278509921136e-05, "loss": 0.0004, "step": 49234 }, { "epoch": 0.96, "learning_rate": 3.4078631549663905e-05, "loss": 0.0013, "step": 49236 }, { "epoch": 0.96, "learning_rate": 3.4077984589406674e-05, "loss": 0.0, "step": 49238 }, { "epoch": 0.96, "learning_rate": 3.4077337629149444e-05, "loss": 0.0, "step": 49240 }, { "epoch": 0.96, "learning_rate": 3.407669066889221e-05, "loss": 0.0, "step": 49242 }, { "epoch": 0.96, "learning_rate": 3.407604370863498e-05, "loss": 0.002, "step": 49244 }, { "epoch": 0.96, "learning_rate": 3.4075396748377744e-05, "loss": 0.0, "step": 49246 }, { "epoch": 0.96, "learning_rate": 3.407474978812052e-05, "loss": 0.0001, "step": 49248 }, { "epoch": 0.96, "learning_rate": 3.407410282786329e-05, "loss": 0.0, "step": 49250 }, { "epoch": 0.96, "learning_rate": 3.407345586760605e-05, "loss": 0.0, "step": 49252 }, { "epoch": 0.96, "learning_rate": 3.407280890734883e-05, "loss": 0.0, "step": 49254 }, { "epoch": 0.96, "learning_rate": 3.407216194709159e-05, "loss": 0.0, "step": 49256 }, { "epoch": 0.96, "learning_rate": 3.4071514986834365e-05, "loss": 0.0, "step": 49258 }, { "epoch": 0.96, "learning_rate": 3.407086802657713e-05, "loss": 0.0006, "step": 49260 }, { "epoch": 0.96, "learning_rate": 3.40702210663199e-05, "loss": 0.0, "step": 49262 }, { "epoch": 0.96, "learning_rate": 3.4069574106062666e-05, "loss": 0.0009, "step": 49264 }, { "epoch": 0.96, "learning_rate": 3.4068927145805435e-05, "loss": 0.0001, "step": 49266 }, { "epoch": 0.96, "learning_rate": 3.4068280185548204e-05, "loss": 0.0, "step": 49268 }, { "epoch": 0.96, "learning_rate": 3.406763322529097e-05, "loss": 0.0002, "step": 49270 }, { "epoch": 0.96, "learning_rate": 3.406698626503374e-05, "loss": 0.0, "step": 49272 }, { "epoch": 0.96, "learning_rate": 3.4066339304776504e-05, "loss": 0.006, "step": 49274 }, { "epoch": 0.96, "learning_rate": 3.406569234451928e-05, "loss": 0.0, "step": 49276 }, { "epoch": 0.96, "learning_rate": 3.406504538426204e-05, "loss": 0.0046, "step": 49278 }, { "epoch": 0.96, "learning_rate": 3.406439842400482e-05, "loss": 0.0, "step": 49280 }, { "epoch": 0.96, "learning_rate": 3.406375146374758e-05, "loss": 0.0, "step": 49282 }, { "epoch": 0.96, "learning_rate": 3.406310450349035e-05, "loss": 0.0006, "step": 49284 }, { "epoch": 0.96, "learning_rate": 3.4062457543233126e-05, "loss": 0.0, "step": 49286 }, { "epoch": 0.96, "learning_rate": 3.406181058297589e-05, "loss": 0.0, "step": 49288 }, { "epoch": 0.96, "learning_rate": 3.406116362271866e-05, "loss": 0.0, "step": 49290 }, { "epoch": 0.96, "learning_rate": 3.4060516662461426e-05, "loss": 0.0014, "step": 49292 }, { "epoch": 0.96, "learning_rate": 3.4059869702204195e-05, "loss": 0.0, "step": 49294 }, { "epoch": 0.96, "learning_rate": 3.4059222741946964e-05, "loss": 0.0, "step": 49296 }, { "epoch": 0.96, "learning_rate": 3.4058575781689733e-05, "loss": 0.0, "step": 49298 }, { "epoch": 0.96, "learning_rate": 3.40579288214325e-05, "loss": 0.0, "step": 49300 }, { "epoch": 0.96, "learning_rate": 3.405728186117527e-05, "loss": 0.0, "step": 49302 }, { "epoch": 0.96, "learning_rate": 3.405663490091804e-05, "loss": 0.0, "step": 49304 }, { "epoch": 0.96, "learning_rate": 3.40559879406608e-05, "loss": 0.0, "step": 49306 }, { "epoch": 0.96, "learning_rate": 3.405534098040358e-05, "loss": 0.0, "step": 49308 }, { "epoch": 0.96, "learning_rate": 3.405469402014634e-05, "loss": 0.0, "step": 49310 }, { "epoch": 0.96, "learning_rate": 3.405404705988911e-05, "loss": 0.0032, "step": 49312 }, { "epoch": 0.96, "learning_rate": 3.405340009963188e-05, "loss": 0.004, "step": 49314 }, { "epoch": 0.96, "learning_rate": 3.405275313937465e-05, "loss": 0.0, "step": 49316 }, { "epoch": 0.96, "learning_rate": 3.4052106179117424e-05, "loss": 0.0039, "step": 49318 }, { "epoch": 0.96, "learning_rate": 3.4051459218860187e-05, "loss": 0.02, "step": 49320 }, { "epoch": 0.96, "learning_rate": 3.4050812258602956e-05, "loss": 0.0, "step": 49322 }, { "epoch": 0.96, "learning_rate": 3.4050165298345725e-05, "loss": 0.0, "step": 49324 }, { "epoch": 0.96, "learning_rate": 3.4049518338088494e-05, "loss": 0.001, "step": 49326 }, { "epoch": 0.96, "learning_rate": 3.404887137783126e-05, "loss": 0.0, "step": 49328 }, { "epoch": 0.96, "learning_rate": 3.404822441757403e-05, "loss": 0.0001, "step": 49330 }, { "epoch": 0.96, "learning_rate": 3.40475774573168e-05, "loss": 0.0, "step": 49332 }, { "epoch": 0.96, "learning_rate": 3.4046930497059563e-05, "loss": 0.0, "step": 49334 }, { "epoch": 0.96, "learning_rate": 3.404628353680234e-05, "loss": 0.0, "step": 49336 }, { "epoch": 0.96, "learning_rate": 3.40456365765451e-05, "loss": 0.0006, "step": 49338 }, { "epoch": 0.96, "learning_rate": 3.404498961628788e-05, "loss": 0.0, "step": 49340 }, { "epoch": 0.96, "learning_rate": 3.404434265603064e-05, "loss": 0.0, "step": 49342 }, { "epoch": 0.96, "learning_rate": 3.404369569577341e-05, "loss": 0.0003, "step": 49344 }, { "epoch": 0.96, "learning_rate": 3.404304873551618e-05, "loss": 0.0, "step": 49346 }, { "epoch": 0.96, "learning_rate": 3.404240177525895e-05, "loss": 0.0038, "step": 49348 }, { "epoch": 0.96, "learning_rate": 3.4041754815001716e-05, "loss": 0.0001, "step": 49350 }, { "epoch": 0.96, "learning_rate": 3.4041107854744485e-05, "loss": 0.0009, "step": 49352 }, { "epoch": 0.96, "learning_rate": 3.4040460894487254e-05, "loss": 0.0001, "step": 49354 }, { "epoch": 0.96, "learning_rate": 3.4039813934230017e-05, "loss": 0.0, "step": 49356 }, { "epoch": 0.96, "learning_rate": 3.403916697397279e-05, "loss": 0.0, "step": 49358 }, { "epoch": 0.96, "learning_rate": 3.4038520013715555e-05, "loss": 0.0, "step": 49360 }, { "epoch": 0.96, "learning_rate": 3.403787305345833e-05, "loss": 0.0001, "step": 49362 }, { "epoch": 0.96, "learning_rate": 3.40372260932011e-05, "loss": 0.0, "step": 49364 }, { "epoch": 0.96, "learning_rate": 3.403657913294386e-05, "loss": 0.0001, "step": 49366 }, { "epoch": 0.96, "learning_rate": 3.403593217268664e-05, "loss": 0.0, "step": 49368 }, { "epoch": 0.96, "learning_rate": 3.40352852124294e-05, "loss": 0.0, "step": 49370 }, { "epoch": 0.96, "learning_rate": 3.403463825217217e-05, "loss": 0.0001, "step": 49372 }, { "epoch": 0.96, "learning_rate": 3.403399129191494e-05, "loss": 0.0017, "step": 49374 }, { "epoch": 0.96, "learning_rate": 3.403334433165771e-05, "loss": 0.0, "step": 49376 }, { "epoch": 0.96, "learning_rate": 3.4032697371400476e-05, "loss": 0.0, "step": 49378 }, { "epoch": 0.96, "learning_rate": 3.4032050411143246e-05, "loss": 0.0005, "step": 49380 }, { "epoch": 0.96, "learning_rate": 3.4031403450886015e-05, "loss": 0.0001, "step": 49382 }, { "epoch": 0.96, "learning_rate": 3.4030756490628784e-05, "loss": 0.0044, "step": 49384 }, { "epoch": 0.96, "learning_rate": 3.403010953037155e-05, "loss": 0.0026, "step": 49386 }, { "epoch": 0.96, "learning_rate": 3.4029462570114315e-05, "loss": 0.0005, "step": 49388 }, { "epoch": 0.96, "learning_rate": 3.402881560985709e-05, "loss": 0.0, "step": 49390 }, { "epoch": 0.96, "learning_rate": 3.402816864959985e-05, "loss": 0.0001, "step": 49392 }, { "epoch": 0.96, "learning_rate": 3.402752168934262e-05, "loss": 0.0, "step": 49394 }, { "epoch": 0.96, "learning_rate": 3.40268747290854e-05, "loss": 0.0008, "step": 49396 }, { "epoch": 0.96, "learning_rate": 3.402622776882816e-05, "loss": 0.0, "step": 49398 }, { "epoch": 0.96, "learning_rate": 3.4025580808570936e-05, "loss": 0.0, "step": 49400 }, { "epoch": 0.96, "learning_rate": 3.40249338483137e-05, "loss": 0.0, "step": 49402 }, { "epoch": 0.96, "learning_rate": 3.402428688805647e-05, "loss": 0.0, "step": 49404 }, { "epoch": 0.96, "learning_rate": 3.402363992779924e-05, "loss": 0.0, "step": 49406 }, { "epoch": 0.96, "learning_rate": 3.4022992967542006e-05, "loss": 0.0008, "step": 49408 }, { "epoch": 0.96, "learning_rate": 3.4022346007284775e-05, "loss": 0.0027, "step": 49410 }, { "epoch": 0.96, "learning_rate": 3.4021699047027544e-05, "loss": 0.0, "step": 49412 }, { "epoch": 0.96, "learning_rate": 3.402105208677031e-05, "loss": 0.0001, "step": 49414 }, { "epoch": 0.96, "learning_rate": 3.4020405126513076e-05, "loss": 0.002, "step": 49416 }, { "epoch": 0.96, "learning_rate": 3.401975816625585e-05, "loss": 0.0001, "step": 49418 }, { "epoch": 0.96, "learning_rate": 3.4019111205998614e-05, "loss": 0.0036, "step": 49420 }, { "epoch": 0.96, "learning_rate": 3.401846424574139e-05, "loss": 0.0, "step": 49422 }, { "epoch": 0.96, "learning_rate": 3.401781728548415e-05, "loss": 0.0, "step": 49424 }, { "epoch": 0.96, "learning_rate": 3.401717032522692e-05, "loss": 0.0062, "step": 49426 }, { "epoch": 0.96, "learning_rate": 3.40165233649697e-05, "loss": 0.0, "step": 49428 }, { "epoch": 0.96, "learning_rate": 3.401587640471246e-05, "loss": 0.0, "step": 49430 }, { "epoch": 0.96, "learning_rate": 3.401522944445523e-05, "loss": 0.0, "step": 49432 }, { "epoch": 0.96, "learning_rate": 3.4014582484198e-05, "loss": 0.0001, "step": 49434 }, { "epoch": 0.96, "learning_rate": 3.4013935523940766e-05, "loss": 0.001, "step": 49436 }, { "epoch": 0.96, "learning_rate": 3.401328856368353e-05, "loss": 0.0001, "step": 49438 }, { "epoch": 0.96, "learning_rate": 3.4012641603426305e-05, "loss": 0.0018, "step": 49440 }, { "epoch": 0.96, "learning_rate": 3.4011994643169074e-05, "loss": 0.0001, "step": 49442 }, { "epoch": 0.96, "learning_rate": 3.401134768291184e-05, "loss": 0.0039, "step": 49444 }, { "epoch": 0.96, "learning_rate": 3.401070072265461e-05, "loss": 0.0, "step": 49446 }, { "epoch": 0.96, "learning_rate": 3.4010053762397374e-05, "loss": 0.0, "step": 49448 }, { "epoch": 0.96, "learning_rate": 3.400940680214015e-05, "loss": 0.0, "step": 49450 }, { "epoch": 0.96, "learning_rate": 3.400875984188291e-05, "loss": 0.0, "step": 49452 }, { "epoch": 0.96, "learning_rate": 3.400811288162568e-05, "loss": 0.0037, "step": 49454 }, { "epoch": 0.96, "learning_rate": 3.400746592136845e-05, "loss": 0.0, "step": 49456 }, { "epoch": 0.96, "learning_rate": 3.400681896111122e-05, "loss": 0.0, "step": 49458 }, { "epoch": 0.96, "learning_rate": 3.400617200085399e-05, "loss": 0.0, "step": 49460 }, { "epoch": 0.96, "learning_rate": 3.400552504059676e-05, "loss": 0.0002, "step": 49462 }, { "epoch": 0.96, "learning_rate": 3.400487808033953e-05, "loss": 0.0083, "step": 49464 }, { "epoch": 0.96, "learning_rate": 3.4004231120082296e-05, "loss": 0.0, "step": 49466 }, { "epoch": 0.96, "learning_rate": 3.4003584159825065e-05, "loss": 0.0, "step": 49468 }, { "epoch": 0.96, "learning_rate": 3.400293719956783e-05, "loss": 0.0, "step": 49470 }, { "epoch": 0.96, "learning_rate": 3.40022902393106e-05, "loss": 0.0, "step": 49472 }, { "epoch": 0.96, "learning_rate": 3.400164327905337e-05, "loss": 0.0, "step": 49474 }, { "epoch": 0.96, "learning_rate": 3.4000996318796135e-05, "loss": 0.0, "step": 49476 }, { "epoch": 0.96, "learning_rate": 3.400034935853891e-05, "loss": 0.0, "step": 49478 }, { "epoch": 0.96, "learning_rate": 3.399970239828167e-05, "loss": 0.0, "step": 49480 }, { "epoch": 0.96, "learning_rate": 3.399905543802445e-05, "loss": 0.0002, "step": 49482 }, { "epoch": 0.96, "learning_rate": 3.399840847776721e-05, "loss": 0.0, "step": 49484 }, { "epoch": 0.96, "learning_rate": 3.399776151750998e-05, "loss": 0.0, "step": 49486 }, { "epoch": 0.96, "learning_rate": 3.399711455725275e-05, "loss": 0.0, "step": 49488 }, { "epoch": 0.96, "learning_rate": 3.399646759699552e-05, "loss": 0.0, "step": 49490 }, { "epoch": 0.96, "learning_rate": 3.399582063673829e-05, "loss": 0.0104, "step": 49492 }, { "epoch": 0.96, "learning_rate": 3.3995173676481056e-05, "loss": 0.0, "step": 49494 }, { "epoch": 0.96, "learning_rate": 3.3994526716223825e-05, "loss": 0.0, "step": 49496 }, { "epoch": 0.96, "learning_rate": 3.399387975596659e-05, "loss": 0.0, "step": 49498 }, { "epoch": 0.96, "learning_rate": 3.3993232795709364e-05, "loss": 0.0, "step": 49500 }, { "epoch": 0.96, "learning_rate": 3.3992585835452126e-05, "loss": 0.0001, "step": 49502 }, { "epoch": 0.96, "learning_rate": 3.39919388751949e-05, "loss": 0.0001, "step": 49504 }, { "epoch": 0.96, "learning_rate": 3.3991291914937664e-05, "loss": 0.0001, "step": 49506 }, { "epoch": 0.96, "learning_rate": 3.399064495468043e-05, "loss": 0.0, "step": 49508 }, { "epoch": 0.96, "learning_rate": 3.398999799442321e-05, "loss": 0.0, "step": 49510 }, { "epoch": 0.96, "learning_rate": 3.398935103416597e-05, "loss": 0.0008, "step": 49512 }, { "epoch": 0.96, "learning_rate": 3.398870407390874e-05, "loss": 0.0, "step": 49514 }, { "epoch": 0.96, "learning_rate": 3.398805711365151e-05, "loss": 0.0, "step": 49516 }, { "epoch": 0.96, "learning_rate": 3.398741015339428e-05, "loss": 0.0, "step": 49518 }, { "epoch": 0.96, "learning_rate": 3.398676319313705e-05, "loss": 0.0002, "step": 49520 }, { "epoch": 0.96, "learning_rate": 3.398611623287982e-05, "loss": 0.0, "step": 49522 }, { "epoch": 0.96, "learning_rate": 3.3985469272622586e-05, "loss": 0.0, "step": 49524 }, { "epoch": 0.96, "learning_rate": 3.3984822312365355e-05, "loss": 0.014, "step": 49526 }, { "epoch": 0.96, "learning_rate": 3.3984175352108124e-05, "loss": 0.0, "step": 49528 }, { "epoch": 0.96, "learning_rate": 3.3983528391850886e-05, "loss": 0.0, "step": 49530 }, { "epoch": 0.96, "learning_rate": 3.398288143159366e-05, "loss": 0.0, "step": 49532 }, { "epoch": 0.96, "learning_rate": 3.3982234471336424e-05, "loss": 0.0005, "step": 49534 }, { "epoch": 0.96, "learning_rate": 3.3981587511079194e-05, "loss": 0.0072, "step": 49536 }, { "epoch": 0.96, "learning_rate": 3.398094055082196e-05, "loss": 0.0, "step": 49538 }, { "epoch": 0.96, "learning_rate": 3.398029359056473e-05, "loss": 0.0001, "step": 49540 }, { "epoch": 0.96, "learning_rate": 3.397964663030751e-05, "loss": 0.0143, "step": 49542 }, { "epoch": 0.96, "learning_rate": 3.397899967005027e-05, "loss": 0.0053, "step": 49544 }, { "epoch": 0.96, "learning_rate": 3.397835270979304e-05, "loss": 0.0171, "step": 49546 }, { "epoch": 0.96, "learning_rate": 3.397770574953581e-05, "loss": 0.0083, "step": 49548 }, { "epoch": 0.96, "learning_rate": 3.397705878927858e-05, "loss": 0.0, "step": 49550 }, { "epoch": 0.96, "learning_rate": 3.3976411829021346e-05, "loss": 0.0, "step": 49552 }, { "epoch": 0.96, "learning_rate": 3.3975764868764115e-05, "loss": 0.0, "step": 49554 }, { "epoch": 0.96, "learning_rate": 3.3975117908506884e-05, "loss": 0.0, "step": 49556 }, { "epoch": 0.96, "learning_rate": 3.397447094824965e-05, "loss": 0.0013, "step": 49558 }, { "epoch": 0.96, "learning_rate": 3.397382398799242e-05, "loss": 0.0, "step": 49560 }, { "epoch": 0.96, "learning_rate": 3.3973177027735185e-05, "loss": 0.0056, "step": 49562 }, { "epoch": 0.96, "learning_rate": 3.397253006747796e-05, "loss": 0.0, "step": 49564 }, { "epoch": 0.96, "learning_rate": 3.397188310722072e-05, "loss": 0.0, "step": 49566 }, { "epoch": 0.96, "learning_rate": 3.397123614696349e-05, "loss": 0.0001, "step": 49568 }, { "epoch": 0.96, "learning_rate": 3.397058918670626e-05, "loss": 0.0001, "step": 49570 }, { "epoch": 0.96, "learning_rate": 3.396994222644903e-05, "loss": 0.0001, "step": 49572 }, { "epoch": 0.96, "learning_rate": 3.39692952661918e-05, "loss": 0.0139, "step": 49574 }, { "epoch": 0.96, "learning_rate": 3.396864830593457e-05, "loss": 0.0001, "step": 49576 }, { "epoch": 0.96, "learning_rate": 3.396800134567734e-05, "loss": 0.0, "step": 49578 }, { "epoch": 0.96, "learning_rate": 3.39673543854201e-05, "loss": 0.0, "step": 49580 }, { "epoch": 0.96, "learning_rate": 3.3966707425162876e-05, "loss": 0.0, "step": 49582 }, { "epoch": 0.96, "learning_rate": 3.396606046490564e-05, "loss": 0.0, "step": 49584 }, { "epoch": 0.96, "learning_rate": 3.3965413504648414e-05, "loss": 0.0, "step": 49586 }, { "epoch": 0.96, "learning_rate": 3.396476654439118e-05, "loss": 0.0, "step": 49588 }, { "epoch": 0.96, "learning_rate": 3.3964119584133945e-05, "loss": 0.0, "step": 49590 }, { "epoch": 0.96, "learning_rate": 3.396347262387672e-05, "loss": 0.0, "step": 49592 }, { "epoch": 0.96, "learning_rate": 3.3962825663619483e-05, "loss": 0.0001, "step": 49594 }, { "epoch": 0.96, "learning_rate": 3.396217870336225e-05, "loss": 0.0017, "step": 49596 }, { "epoch": 0.96, "learning_rate": 3.396153174310502e-05, "loss": 0.0045, "step": 49598 }, { "epoch": 0.96, "learning_rate": 3.396088478284779e-05, "loss": 0.0191, "step": 49600 }, { "epoch": 0.96, "learning_rate": 3.396023782259056e-05, "loss": 0.014, "step": 49602 }, { "epoch": 0.96, "learning_rate": 3.395959086233333e-05, "loss": 0.0, "step": 49604 }, { "epoch": 0.96, "learning_rate": 3.39589439020761e-05, "loss": 0.0, "step": 49606 }, { "epoch": 0.96, "learning_rate": 3.395829694181887e-05, "loss": 0.0, "step": 49608 }, { "epoch": 0.96, "learning_rate": 3.3957649981561636e-05, "loss": 0.0001, "step": 49610 }, { "epoch": 0.96, "learning_rate": 3.39570030213044e-05, "loss": 0.0, "step": 49612 }, { "epoch": 0.96, "learning_rate": 3.3956356061047174e-05, "loss": 0.0, "step": 49614 }, { "epoch": 0.96, "learning_rate": 3.3955709100789937e-05, "loss": 0.0004, "step": 49616 }, { "epoch": 0.96, "learning_rate": 3.3955062140532706e-05, "loss": 0.0001, "step": 49618 }, { "epoch": 0.96, "learning_rate": 3.395441518027548e-05, "loss": 0.0063, "step": 49620 }, { "epoch": 0.96, "learning_rate": 3.3953768220018244e-05, "loss": 0.0, "step": 49622 }, { "epoch": 0.96, "learning_rate": 3.395312125976102e-05, "loss": 0.0, "step": 49624 }, { "epoch": 0.96, "learning_rate": 3.395247429950378e-05, "loss": 0.0, "step": 49626 }, { "epoch": 0.96, "learning_rate": 3.395182733924655e-05, "loss": 0.0, "step": 49628 }, { "epoch": 0.96, "learning_rate": 3.395118037898932e-05, "loss": 0.0001, "step": 49630 }, { "epoch": 0.96, "learning_rate": 3.395053341873209e-05, "loss": 0.0, "step": 49632 }, { "epoch": 0.96, "learning_rate": 3.394988645847486e-05, "loss": 0.0001, "step": 49634 }, { "epoch": 0.96, "learning_rate": 3.394923949821763e-05, "loss": 0.0, "step": 49636 }, { "epoch": 0.96, "learning_rate": 3.3948592537960396e-05, "loss": 0.0, "step": 49638 }, { "epoch": 0.96, "learning_rate": 3.394794557770316e-05, "loss": 0.0004, "step": 49640 }, { "epoch": 0.96, "learning_rate": 3.3947298617445935e-05, "loss": 0.0, "step": 49642 }, { "epoch": 0.96, "learning_rate": 3.39466516571887e-05, "loss": 0.0, "step": 49644 }, { "epoch": 0.96, "learning_rate": 3.394600469693147e-05, "loss": 0.0002, "step": 49646 }, { "epoch": 0.96, "learning_rate": 3.3945357736674235e-05, "loss": 0.0, "step": 49648 }, { "epoch": 0.96, "learning_rate": 3.3944710776417004e-05, "loss": 0.0031, "step": 49650 }, { "epoch": 0.96, "learning_rate": 3.394406381615978e-05, "loss": 0.0, "step": 49652 }, { "epoch": 0.96, "learning_rate": 3.394341685590254e-05, "loss": 0.0018, "step": 49654 }, { "epoch": 0.96, "learning_rate": 3.394276989564531e-05, "loss": 0.0, "step": 49656 }, { "epoch": 0.96, "learning_rate": 3.394212293538808e-05, "loss": 0.0076, "step": 49658 }, { "epoch": 0.96, "learning_rate": 3.394147597513085e-05, "loss": 0.0, "step": 49660 }, { "epoch": 0.96, "learning_rate": 3.394082901487361e-05, "loss": 0.0299, "step": 49662 }, { "epoch": 0.96, "learning_rate": 3.394018205461639e-05, "loss": 0.0017, "step": 49664 }, { "epoch": 0.96, "learning_rate": 3.393953509435916e-05, "loss": 0.0, "step": 49666 }, { "epoch": 0.96, "learning_rate": 3.3938888134101926e-05, "loss": 0.0, "step": 49668 }, { "epoch": 0.96, "learning_rate": 3.3938241173844695e-05, "loss": 0.0001, "step": 49670 }, { "epoch": 0.96, "learning_rate": 3.393759421358746e-05, "loss": 0.0, "step": 49672 }, { "epoch": 0.96, "learning_rate": 3.393694725333023e-05, "loss": 0.0007, "step": 49674 }, { "epoch": 0.96, "learning_rate": 3.3936300293072996e-05, "loss": 0.0002, "step": 49676 }, { "epoch": 0.96, "learning_rate": 3.3935653332815765e-05, "loss": 0.0, "step": 49678 }, { "epoch": 0.96, "learning_rate": 3.3935006372558534e-05, "loss": 0.0, "step": 49680 }, { "epoch": 0.96, "learning_rate": 3.39343594123013e-05, "loss": 0.0002, "step": 49682 }, { "epoch": 0.96, "learning_rate": 3.393371245204407e-05, "loss": 0.0003, "step": 49684 }, { "epoch": 0.96, "learning_rate": 3.393306549178684e-05, "loss": 0.0128, "step": 49686 }, { "epoch": 0.96, "learning_rate": 3.393241853152961e-05, "loss": 0.0, "step": 49688 }, { "epoch": 0.96, "learning_rate": 3.393177157127238e-05, "loss": 0.0004, "step": 49690 }, { "epoch": 0.96, "learning_rate": 3.393112461101515e-05, "loss": 0.0, "step": 49692 }, { "epoch": 0.96, "learning_rate": 3.393047765075791e-05, "loss": 0.0084, "step": 49694 }, { "epoch": 0.96, "learning_rate": 3.3929830690500686e-05, "loss": 0.0001, "step": 49696 }, { "epoch": 0.96, "learning_rate": 3.3929183730243455e-05, "loss": 0.0, "step": 49698 }, { "epoch": 0.96, "learning_rate": 3.392853676998622e-05, "loss": 0.0, "step": 49700 }, { "epoch": 0.96, "learning_rate": 3.3927889809728994e-05, "loss": 0.0, "step": 49702 }, { "epoch": 0.96, "learning_rate": 3.3927242849471756e-05, "loss": 0.0, "step": 49704 }, { "epoch": 0.96, "learning_rate": 3.392659588921453e-05, "loss": 0.0, "step": 49706 }, { "epoch": 0.96, "learning_rate": 3.3925948928957294e-05, "loss": 0.0, "step": 49708 }, { "epoch": 0.96, "learning_rate": 3.392530196870006e-05, "loss": 0.0022, "step": 49710 }, { "epoch": 0.96, "learning_rate": 3.392465500844283e-05, "loss": 0.0, "step": 49712 }, { "epoch": 0.96, "learning_rate": 3.39240080481856e-05, "loss": 0.0, "step": 49714 }, { "epoch": 0.96, "learning_rate": 3.392336108792837e-05, "loss": 0.0001, "step": 49716 }, { "epoch": 0.96, "learning_rate": 3.392271412767114e-05, "loss": 0.0, "step": 49718 }, { "epoch": 0.97, "learning_rate": 3.392206716741391e-05, "loss": 0.0, "step": 49720 }, { "epoch": 0.97, "learning_rate": 3.392142020715667e-05, "loss": 0.0002, "step": 49722 }, { "epoch": 0.97, "learning_rate": 3.392077324689945e-05, "loss": 0.0, "step": 49724 }, { "epoch": 0.97, "learning_rate": 3.392012628664221e-05, "loss": 0.0, "step": 49726 }, { "epoch": 0.97, "learning_rate": 3.3919479326384985e-05, "loss": 0.0011, "step": 49728 }, { "epoch": 0.97, "learning_rate": 3.3918832366127754e-05, "loss": 0.0002, "step": 49730 }, { "epoch": 0.97, "learning_rate": 3.3918185405870516e-05, "loss": 0.0, "step": 49732 }, { "epoch": 0.97, "learning_rate": 3.391753844561329e-05, "loss": 0.0041, "step": 49734 }, { "epoch": 0.97, "learning_rate": 3.3916891485356055e-05, "loss": 0.0, "step": 49736 }, { "epoch": 0.97, "learning_rate": 3.3916244525098824e-05, "loss": 0.0003, "step": 49738 }, { "epoch": 0.97, "learning_rate": 3.391559756484159e-05, "loss": 0.0, "step": 49740 }, { "epoch": 0.97, "learning_rate": 3.391495060458436e-05, "loss": 0.0062, "step": 49742 }, { "epoch": 0.97, "learning_rate": 3.391430364432713e-05, "loss": 0.0, "step": 49744 }, { "epoch": 0.97, "learning_rate": 3.39136566840699e-05, "loss": 0.0, "step": 49746 }, { "epoch": 0.97, "learning_rate": 3.391300972381267e-05, "loss": 0.0021, "step": 49748 }, { "epoch": 0.97, "learning_rate": 3.391236276355544e-05, "loss": 0.0, "step": 49750 }, { "epoch": 0.97, "learning_rate": 3.391171580329821e-05, "loss": 0.0001, "step": 49752 }, { "epoch": 0.97, "learning_rate": 3.391106884304097e-05, "loss": 0.0, "step": 49754 }, { "epoch": 0.97, "learning_rate": 3.3910421882783745e-05, "loss": 0.0001, "step": 49756 }, { "epoch": 0.97, "learning_rate": 3.390977492252651e-05, "loss": 0.0, "step": 49758 }, { "epoch": 0.97, "learning_rate": 3.390912796226928e-05, "loss": 0.0, "step": 49760 }, { "epoch": 0.97, "learning_rate": 3.3908481002012046e-05, "loss": 0.0441, "step": 49762 }, { "epoch": 0.97, "learning_rate": 3.3907834041754815e-05, "loss": 0.0, "step": 49764 }, { "epoch": 0.97, "learning_rate": 3.390718708149759e-05, "loss": 0.0002, "step": 49766 }, { "epoch": 0.97, "learning_rate": 3.390654012124035e-05, "loss": 0.0, "step": 49768 }, { "epoch": 0.97, "learning_rate": 3.390589316098312e-05, "loss": 0.0, "step": 49770 }, { "epoch": 0.97, "learning_rate": 3.390524620072589e-05, "loss": 0.0, "step": 49772 }, { "epoch": 0.97, "learning_rate": 3.390459924046866e-05, "loss": 0.0, "step": 49774 }, { "epoch": 0.97, "learning_rate": 3.390395228021143e-05, "loss": 0.0, "step": 49776 }, { "epoch": 0.97, "learning_rate": 3.39033053199542e-05, "loss": 0.0002, "step": 49778 }, { "epoch": 0.97, "learning_rate": 3.390265835969697e-05, "loss": 0.0029, "step": 49780 }, { "epoch": 0.97, "learning_rate": 3.390201139943973e-05, "loss": 0.0038, "step": 49782 }, { "epoch": 0.97, "learning_rate": 3.3901364439182506e-05, "loss": 0.0, "step": 49784 }, { "epoch": 0.97, "learning_rate": 3.390071747892527e-05, "loss": 0.0001, "step": 49786 }, { "epoch": 0.97, "learning_rate": 3.3900070518668044e-05, "loss": 0.0001, "step": 49788 }, { "epoch": 0.97, "learning_rate": 3.3899423558410806e-05, "loss": 0.0, "step": 49790 }, { "epoch": 0.97, "learning_rate": 3.3898776598153575e-05, "loss": 0.0, "step": 49792 }, { "epoch": 0.97, "learning_rate": 3.3898129637896344e-05, "loss": 0.0001, "step": 49794 }, { "epoch": 0.97, "learning_rate": 3.3897482677639113e-05, "loss": 0.0003, "step": 49796 }, { "epoch": 0.97, "learning_rate": 3.389683571738188e-05, "loss": 0.0, "step": 49798 }, { "epoch": 0.97, "learning_rate": 3.389618875712465e-05, "loss": 0.0, "step": 49800 }, { "epoch": 0.97, "learning_rate": 3.389554179686742e-05, "loss": 0.0011, "step": 49802 }, { "epoch": 0.97, "learning_rate": 3.389489483661018e-05, "loss": 0.0001, "step": 49804 }, { "epoch": 0.97, "learning_rate": 3.389424787635296e-05, "loss": 0.0, "step": 49806 }, { "epoch": 0.97, "learning_rate": 3.389360091609573e-05, "loss": 0.0001, "step": 49808 }, { "epoch": 0.97, "learning_rate": 3.38929539558385e-05, "loss": 0.0004, "step": 49810 }, { "epoch": 0.97, "learning_rate": 3.3892306995581266e-05, "loss": 0.0, "step": 49812 }, { "epoch": 0.97, "learning_rate": 3.389166003532403e-05, "loss": 0.0, "step": 49814 }, { "epoch": 0.97, "learning_rate": 3.3891013075066804e-05, "loss": 0.0, "step": 49816 }, { "epoch": 0.97, "learning_rate": 3.389036611480957e-05, "loss": 0.0, "step": 49818 }, { "epoch": 0.97, "learning_rate": 3.3889719154552336e-05, "loss": 0.0001, "step": 49820 }, { "epoch": 0.97, "learning_rate": 3.3889072194295105e-05, "loss": 0.0, "step": 49822 }, { "epoch": 0.97, "learning_rate": 3.3888425234037874e-05, "loss": 0.0002, "step": 49824 }, { "epoch": 0.97, "learning_rate": 3.388777827378064e-05, "loss": 0.0001, "step": 49826 }, { "epoch": 0.97, "learning_rate": 3.388713131352341e-05, "loss": 0.0, "step": 49828 }, { "epoch": 0.97, "learning_rate": 3.388648435326618e-05, "loss": 0.0, "step": 49830 }, { "epoch": 0.97, "learning_rate": 3.388583739300895e-05, "loss": 0.0011, "step": 49832 }, { "epoch": 0.97, "learning_rate": 3.388519043275172e-05, "loss": 0.0, "step": 49834 }, { "epoch": 0.97, "learning_rate": 3.388454347249448e-05, "loss": 0.0, "step": 49836 }, { "epoch": 0.97, "learning_rate": 3.388389651223726e-05, "loss": 0.0002, "step": 49838 }, { "epoch": 0.97, "learning_rate": 3.388324955198002e-05, "loss": 0.001, "step": 49840 }, { "epoch": 0.97, "learning_rate": 3.388260259172279e-05, "loss": 0.0, "step": 49842 }, { "epoch": 0.97, "learning_rate": 3.3881955631465565e-05, "loss": 0.0, "step": 49844 }, { "epoch": 0.97, "learning_rate": 3.388130867120833e-05, "loss": 0.0002, "step": 49846 }, { "epoch": 0.97, "learning_rate": 3.38806617109511e-05, "loss": 0.0, "step": 49848 }, { "epoch": 0.97, "learning_rate": 3.3880014750693865e-05, "loss": 0.0003, "step": 49850 }, { "epoch": 0.97, "learning_rate": 3.3879367790436634e-05, "loss": 0.0071, "step": 49852 }, { "epoch": 0.97, "learning_rate": 3.38787208301794e-05, "loss": 0.0, "step": 49854 }, { "epoch": 0.97, "learning_rate": 3.387807386992217e-05, "loss": 0.0, "step": 49856 }, { "epoch": 0.97, "learning_rate": 3.387742690966494e-05, "loss": 0.0, "step": 49858 }, { "epoch": 0.97, "learning_rate": 3.387677994940771e-05, "loss": 0.0, "step": 49860 }, { "epoch": 0.97, "learning_rate": 3.387613298915048e-05, "loss": 0.0032, "step": 49862 }, { "epoch": 0.97, "learning_rate": 3.387548602889324e-05, "loss": 0.0009, "step": 49864 }, { "epoch": 0.97, "learning_rate": 3.387483906863602e-05, "loss": 0.0, "step": 49866 }, { "epoch": 0.97, "learning_rate": 3.387419210837878e-05, "loss": 0.0025, "step": 49868 }, { "epoch": 0.97, "learning_rate": 3.3873545148121556e-05, "loss": 0.0, "step": 49870 }, { "epoch": 0.97, "learning_rate": 3.387289818786432e-05, "loss": 0.0009, "step": 49872 }, { "epoch": 0.97, "learning_rate": 3.387225122760709e-05, "loss": 0.0, "step": 49874 }, { "epoch": 0.97, "learning_rate": 3.387160426734986e-05, "loss": 0.0, "step": 49876 }, { "epoch": 0.97, "learning_rate": 3.3870957307092626e-05, "loss": 0.0, "step": 49878 }, { "epoch": 0.97, "learning_rate": 3.3870310346835395e-05, "loss": 0.0247, "step": 49880 }, { "epoch": 0.97, "learning_rate": 3.3869663386578164e-05, "loss": 0.0017, "step": 49882 }, { "epoch": 0.97, "learning_rate": 3.386901642632093e-05, "loss": 0.0, "step": 49884 }, { "epoch": 0.97, "learning_rate": 3.38683694660637e-05, "loss": 0.0043, "step": 49886 }, { "epoch": 0.97, "learning_rate": 3.386772250580647e-05, "loss": 0.0007, "step": 49888 }, { "epoch": 0.97, "learning_rate": 3.386707554554924e-05, "loss": 0.0, "step": 49890 }, { "epoch": 0.97, "learning_rate": 3.386642858529201e-05, "loss": 0.0, "step": 49892 }, { "epoch": 0.97, "learning_rate": 3.386578162503478e-05, "loss": 0.0, "step": 49894 }, { "epoch": 0.97, "learning_rate": 3.386513466477754e-05, "loss": 0.0, "step": 49896 }, { "epoch": 0.97, "learning_rate": 3.3864487704520316e-05, "loss": 0.0001, "step": 49898 }, { "epoch": 0.97, "learning_rate": 3.386384074426308e-05, "loss": 0.0001, "step": 49900 }, { "epoch": 0.97, "learning_rate": 3.386319378400585e-05, "loss": 0.0002, "step": 49902 }, { "epoch": 0.97, "learning_rate": 3.386254682374862e-05, "loss": 0.0, "step": 49904 }, { "epoch": 0.97, "learning_rate": 3.3861899863491386e-05, "loss": 0.0024, "step": 49906 }, { "epoch": 0.97, "learning_rate": 3.386125290323416e-05, "loss": 0.0, "step": 49908 }, { "epoch": 0.97, "learning_rate": 3.3860605942976924e-05, "loss": 0.0009, "step": 49910 }, { "epoch": 0.97, "learning_rate": 3.385995898271969e-05, "loss": 0.0, "step": 49912 }, { "epoch": 0.97, "learning_rate": 3.385931202246246e-05, "loss": 0.0033, "step": 49914 }, { "epoch": 0.97, "learning_rate": 3.385866506220523e-05, "loss": 0.0075, "step": 49916 }, { "epoch": 0.97, "learning_rate": 3.3858018101947994e-05, "loss": 0.001, "step": 49918 }, { "epoch": 0.97, "learning_rate": 3.385737114169077e-05, "loss": 0.0, "step": 49920 }, { "epoch": 0.97, "learning_rate": 3.385672418143354e-05, "loss": 0.0, "step": 49922 }, { "epoch": 0.97, "learning_rate": 3.38560772211763e-05, "loss": 0.0, "step": 49924 }, { "epoch": 0.97, "learning_rate": 3.385543026091908e-05, "loss": 0.0, "step": 49926 }, { "epoch": 0.97, "learning_rate": 3.385478330066184e-05, "loss": 0.0, "step": 49928 }, { "epoch": 0.97, "learning_rate": 3.3854136340404615e-05, "loss": 0.0005, "step": 49930 }, { "epoch": 0.97, "learning_rate": 3.385348938014738e-05, "loss": 0.0, "step": 49932 }, { "epoch": 0.97, "learning_rate": 3.3852842419890146e-05, "loss": 0.0002, "step": 49934 }, { "epoch": 0.97, "learning_rate": 3.3852195459632916e-05, "loss": 0.0002, "step": 49936 }, { "epoch": 0.97, "learning_rate": 3.3851548499375685e-05, "loss": 0.0, "step": 49938 }, { "epoch": 0.97, "learning_rate": 3.3850901539118454e-05, "loss": 0.0004, "step": 49940 }, { "epoch": 0.97, "learning_rate": 3.385025457886122e-05, "loss": 0.0, "step": 49942 }, { "epoch": 0.97, "learning_rate": 3.384960761860399e-05, "loss": 0.0034, "step": 49944 }, { "epoch": 0.97, "learning_rate": 3.3848960658346754e-05, "loss": 0.0076, "step": 49946 }, { "epoch": 0.97, "learning_rate": 3.384831369808953e-05, "loss": 0.0001, "step": 49948 }, { "epoch": 0.97, "learning_rate": 3.384766673783229e-05, "loss": 0.0, "step": 49950 }, { "epoch": 0.97, "learning_rate": 3.384701977757507e-05, "loss": 0.0, "step": 49952 }, { "epoch": 0.97, "learning_rate": 3.384637281731784e-05, "loss": 0.0056, "step": 49954 }, { "epoch": 0.97, "learning_rate": 3.38457258570606e-05, "loss": 0.0, "step": 49956 }, { "epoch": 0.97, "learning_rate": 3.3845078896803375e-05, "loss": 0.0, "step": 49958 }, { "epoch": 0.97, "learning_rate": 3.384443193654614e-05, "loss": 0.0, "step": 49960 }, { "epoch": 0.97, "learning_rate": 3.384378497628891e-05, "loss": 0.0237, "step": 49962 }, { "epoch": 0.97, "learning_rate": 3.3843138016031676e-05, "loss": 0.0, "step": 49964 }, { "epoch": 0.97, "learning_rate": 3.3842491055774445e-05, "loss": 0.0028, "step": 49966 }, { "epoch": 0.97, "learning_rate": 3.3841844095517214e-05, "loss": 0.0008, "step": 49968 }, { "epoch": 0.97, "learning_rate": 3.384119713525998e-05, "loss": 0.0001, "step": 49970 }, { "epoch": 0.97, "learning_rate": 3.384055017500275e-05, "loss": 0.0, "step": 49972 }, { "epoch": 0.97, "learning_rate": 3.383990321474552e-05, "loss": 0.0, "step": 49974 }, { "epoch": 0.97, "learning_rate": 3.383925625448829e-05, "loss": 0.0094, "step": 49976 }, { "epoch": 0.97, "learning_rate": 3.383860929423105e-05, "loss": 0.0, "step": 49978 }, { "epoch": 0.97, "learning_rate": 3.383796233397383e-05, "loss": 0.0001, "step": 49980 }, { "epoch": 0.97, "learning_rate": 3.383731537371659e-05, "loss": 0.0, "step": 49982 }, { "epoch": 0.97, "learning_rate": 3.383666841345936e-05, "loss": 0.0, "step": 49984 }, { "epoch": 0.97, "learning_rate": 3.3836021453202136e-05, "loss": 0.0, "step": 49986 }, { "epoch": 0.97, "learning_rate": 3.38353744929449e-05, "loss": 0.0, "step": 49988 }, { "epoch": 0.97, "learning_rate": 3.3834727532687674e-05, "loss": 0.0001, "step": 49990 }, { "epoch": 0.97, "learning_rate": 3.3834080572430436e-05, "loss": 0.0, "step": 49992 }, { "epoch": 0.97, "learning_rate": 3.3833433612173205e-05, "loss": 0.0093, "step": 49994 }, { "epoch": 0.97, "learning_rate": 3.3832786651915974e-05, "loss": 0.0, "step": 49996 }, { "epoch": 0.97, "learning_rate": 3.3832139691658744e-05, "loss": 0.0001, "step": 49998 }, { "epoch": 0.97, "learning_rate": 3.383149273140151e-05, "loss": 0.0, "step": 50000 }, { "epoch": 0.97, "learning_rate": 3.383084577114428e-05, "loss": 0.0, "step": 50002 }, { "epoch": 0.97, "learning_rate": 3.383019881088705e-05, "loss": 0.0001, "step": 50004 }, { "epoch": 0.97, "learning_rate": 3.382955185062981e-05, "loss": 0.0007, "step": 50006 }, { "epoch": 0.97, "learning_rate": 3.382890489037259e-05, "loss": 0.0, "step": 50008 }, { "epoch": 0.97, "learning_rate": 3.382825793011535e-05, "loss": 0.0, "step": 50010 }, { "epoch": 0.97, "learning_rate": 3.382761096985813e-05, "loss": 0.0001, "step": 50012 }, { "epoch": 0.97, "learning_rate": 3.382696400960089e-05, "loss": 0.0, "step": 50014 }, { "epoch": 0.97, "learning_rate": 3.382631704934366e-05, "loss": 0.0, "step": 50016 }, { "epoch": 0.97, "learning_rate": 3.382567008908643e-05, "loss": 0.0, "step": 50018 }, { "epoch": 0.97, "learning_rate": 3.38250231288292e-05, "loss": 0.0, "step": 50020 }, { "epoch": 0.97, "learning_rate": 3.3824376168571966e-05, "loss": 0.0, "step": 50022 }, { "epoch": 0.97, "learning_rate": 3.3823729208314735e-05, "loss": 0.0, "step": 50024 }, { "epoch": 0.97, "learning_rate": 3.3823082248057504e-05, "loss": 0.0, "step": 50026 }, { "epoch": 0.97, "learning_rate": 3.382243528780027e-05, "loss": 0.0091, "step": 50028 }, { "epoch": 0.97, "learning_rate": 3.382178832754304e-05, "loss": 0.0, "step": 50030 }, { "epoch": 0.97, "learning_rate": 3.382114136728581e-05, "loss": 0.0, "step": 50032 }, { "epoch": 0.97, "learning_rate": 3.382049440702858e-05, "loss": 0.0001, "step": 50034 }, { "epoch": 0.97, "learning_rate": 3.381984744677135e-05, "loss": 0.0, "step": 50036 }, { "epoch": 0.97, "learning_rate": 3.381920048651411e-05, "loss": 0.0, "step": 50038 }, { "epoch": 0.97, "learning_rate": 3.381855352625689e-05, "loss": 0.0001, "step": 50040 }, { "epoch": 0.97, "learning_rate": 3.381790656599965e-05, "loss": 0.0, "step": 50042 }, { "epoch": 0.97, "learning_rate": 3.381725960574242e-05, "loss": 0.0, "step": 50044 }, { "epoch": 0.97, "learning_rate": 3.381661264548519e-05, "loss": 0.0, "step": 50046 }, { "epoch": 0.97, "learning_rate": 3.381596568522796e-05, "loss": 0.0005, "step": 50048 }, { "epoch": 0.97, "learning_rate": 3.3815318724970726e-05, "loss": 0.0, "step": 50050 }, { "epoch": 0.97, "learning_rate": 3.3814671764713495e-05, "loss": 0.0, "step": 50052 }, { "epoch": 0.97, "learning_rate": 3.3814024804456264e-05, "loss": 0.0, "step": 50054 }, { "epoch": 0.97, "learning_rate": 3.3813377844199033e-05, "loss": 0.0, "step": 50056 }, { "epoch": 0.97, "learning_rate": 3.38127308839418e-05, "loss": 0.0, "step": 50058 }, { "epoch": 0.97, "learning_rate": 3.3812083923684565e-05, "loss": 0.0, "step": 50060 }, { "epoch": 0.97, "learning_rate": 3.381143696342734e-05, "loss": 0.0, "step": 50062 }, { "epoch": 0.97, "learning_rate": 3.38107900031701e-05, "loss": 0.0012, "step": 50064 }, { "epoch": 0.97, "learning_rate": 3.381014304291287e-05, "loss": 0.0, "step": 50066 }, { "epoch": 0.97, "learning_rate": 3.380949608265565e-05, "loss": 0.0037, "step": 50068 }, { "epoch": 0.97, "learning_rate": 3.380884912239841e-05, "loss": 0.0001, "step": 50070 }, { "epoch": 0.97, "learning_rate": 3.3808202162141186e-05, "loss": 0.0, "step": 50072 }, { "epoch": 0.97, "learning_rate": 3.380755520188395e-05, "loss": 0.0002, "step": 50074 }, { "epoch": 0.97, "learning_rate": 3.380690824162672e-05, "loss": 0.0258, "step": 50076 }, { "epoch": 0.97, "learning_rate": 3.380626128136949e-05, "loss": 0.0, "step": 50078 }, { "epoch": 0.97, "learning_rate": 3.3805614321112256e-05, "loss": 0.0038, "step": 50080 }, { "epoch": 0.97, "learning_rate": 3.3804967360855025e-05, "loss": 0.0, "step": 50082 }, { "epoch": 0.97, "learning_rate": 3.3804320400597794e-05, "loss": 0.011, "step": 50084 }, { "epoch": 0.97, "learning_rate": 3.380367344034056e-05, "loss": 0.0, "step": 50086 }, { "epoch": 0.97, "learning_rate": 3.3803026480083325e-05, "loss": 0.0013, "step": 50088 }, { "epoch": 0.97, "learning_rate": 3.38023795198261e-05, "loss": 0.0115, "step": 50090 }, { "epoch": 0.97, "learning_rate": 3.3801732559568863e-05, "loss": 0.0, "step": 50092 }, { "epoch": 0.97, "learning_rate": 3.380108559931164e-05, "loss": 0.0047, "step": 50094 }, { "epoch": 0.97, "learning_rate": 3.38004386390544e-05, "loss": 0.0, "step": 50096 }, { "epoch": 0.97, "learning_rate": 3.379979167879717e-05, "loss": 0.0004, "step": 50098 }, { "epoch": 0.97, "learning_rate": 3.3799144718539947e-05, "loss": 0.0001, "step": 50100 }, { "epoch": 0.97, "learning_rate": 3.379849775828271e-05, "loss": 0.0, "step": 50102 }, { "epoch": 0.97, "learning_rate": 3.379785079802548e-05, "loss": 0.0, "step": 50104 }, { "epoch": 0.97, "learning_rate": 3.379720383776825e-05, "loss": 0.0, "step": 50106 }, { "epoch": 0.97, "learning_rate": 3.3796556877511016e-05, "loss": 0.0, "step": 50108 }, { "epoch": 0.97, "learning_rate": 3.3795909917253785e-05, "loss": 0.0, "step": 50110 }, { "epoch": 0.97, "learning_rate": 3.3795262956996554e-05, "loss": 0.0, "step": 50112 }, { "epoch": 0.97, "learning_rate": 3.379461599673932e-05, "loss": 0.0, "step": 50114 }, { "epoch": 0.97, "learning_rate": 3.379396903648209e-05, "loss": 0.0201, "step": 50116 }, { "epoch": 0.97, "learning_rate": 3.379332207622486e-05, "loss": 0.0001, "step": 50118 }, { "epoch": 0.97, "learning_rate": 3.3792675115967624e-05, "loss": 0.0, "step": 50120 }, { "epoch": 0.97, "learning_rate": 3.37920281557104e-05, "loss": 0.0005, "step": 50122 }, { "epoch": 0.97, "learning_rate": 3.379138119545316e-05, "loss": 0.0, "step": 50124 }, { "epoch": 0.97, "learning_rate": 3.379073423519593e-05, "loss": 0.0, "step": 50126 }, { "epoch": 0.97, "learning_rate": 3.37900872749387e-05, "loss": 0.0, "step": 50128 }, { "epoch": 0.97, "learning_rate": 3.378944031468147e-05, "loss": 0.0001, "step": 50130 }, { "epoch": 0.97, "learning_rate": 3.3788793354424245e-05, "loss": 0.0, "step": 50132 }, { "epoch": 0.97, "learning_rate": 3.378814639416701e-05, "loss": 0.0008, "step": 50134 }, { "epoch": 0.97, "learning_rate": 3.3787499433909777e-05, "loss": 0.0243, "step": 50136 }, { "epoch": 0.97, "learning_rate": 3.3786852473652546e-05, "loss": 0.0001, "step": 50138 }, { "epoch": 0.97, "learning_rate": 3.3786205513395315e-05, "loss": 0.0, "step": 50140 }, { "epoch": 0.97, "learning_rate": 3.378555855313808e-05, "loss": 0.0129, "step": 50142 }, { "epoch": 0.97, "learning_rate": 3.378491159288085e-05, "loss": 0.0, "step": 50144 }, { "epoch": 0.97, "learning_rate": 3.378426463262362e-05, "loss": 0.0, "step": 50146 }, { "epoch": 0.97, "learning_rate": 3.3783617672366384e-05, "loss": 0.0011, "step": 50148 }, { "epoch": 0.97, "learning_rate": 3.378297071210916e-05, "loss": 0.0267, "step": 50150 }, { "epoch": 0.97, "learning_rate": 3.378232375185192e-05, "loss": 0.0, "step": 50152 }, { "epoch": 0.97, "learning_rate": 3.37816767915947e-05, "loss": 0.0, "step": 50154 }, { "epoch": 0.97, "learning_rate": 3.378102983133746e-05, "loss": 0.0, "step": 50156 }, { "epoch": 0.97, "learning_rate": 3.378038287108023e-05, "loss": 0.0, "step": 50158 }, { "epoch": 0.97, "learning_rate": 3.3779735910823e-05, "loss": 0.0, "step": 50160 }, { "epoch": 0.97, "learning_rate": 3.377908895056577e-05, "loss": 0.0005, "step": 50162 }, { "epoch": 0.97, "learning_rate": 3.377844199030854e-05, "loss": 0.0, "step": 50164 }, { "epoch": 0.97, "learning_rate": 3.3777795030051306e-05, "loss": 0.0, "step": 50166 }, { "epoch": 0.97, "learning_rate": 3.3777148069794075e-05, "loss": 0.0, "step": 50168 }, { "epoch": 0.97, "learning_rate": 3.3776501109536844e-05, "loss": 0.0, "step": 50170 }, { "epoch": 0.97, "learning_rate": 3.377585414927961e-05, "loss": 0.0, "step": 50172 }, { "epoch": 0.97, "learning_rate": 3.3775207189022376e-05, "loss": 0.0001, "step": 50174 }, { "epoch": 0.97, "learning_rate": 3.377456022876515e-05, "loss": 0.0187, "step": 50176 }, { "epoch": 0.97, "learning_rate": 3.377391326850792e-05, "loss": 0.0, "step": 50178 }, { "epoch": 0.97, "learning_rate": 3.377326630825068e-05, "loss": 0.0002, "step": 50180 }, { "epoch": 0.97, "learning_rate": 3.377261934799346e-05, "loss": 0.0, "step": 50182 }, { "epoch": 0.97, "learning_rate": 3.377197238773622e-05, "loss": 0.0002, "step": 50184 }, { "epoch": 0.97, "learning_rate": 3.377132542747899e-05, "loss": 0.0201, "step": 50186 }, { "epoch": 0.97, "learning_rate": 3.377067846722176e-05, "loss": 0.0116, "step": 50188 }, { "epoch": 0.97, "learning_rate": 3.377003150696453e-05, "loss": 0.0004, "step": 50190 }, { "epoch": 0.97, "learning_rate": 3.37693845467073e-05, "loss": 0.0, "step": 50192 }, { "epoch": 0.97, "learning_rate": 3.3768737586450066e-05, "loss": 0.0, "step": 50194 }, { "epoch": 0.97, "learning_rate": 3.3768090626192835e-05, "loss": 0.0, "step": 50196 }, { "epoch": 0.97, "learning_rate": 3.3767443665935605e-05, "loss": 0.0029, "step": 50198 }, { "epoch": 0.97, "learning_rate": 3.3766796705678374e-05, "loss": 0.0003, "step": 50200 }, { "epoch": 0.97, "learning_rate": 3.3766149745421136e-05, "loss": 0.0001, "step": 50202 }, { "epoch": 0.97, "learning_rate": 3.376550278516391e-05, "loss": 0.0, "step": 50204 }, { "epoch": 0.97, "learning_rate": 3.3764855824906674e-05, "loss": 0.0, "step": 50206 }, { "epoch": 0.97, "learning_rate": 3.376420886464944e-05, "loss": 0.0, "step": 50208 }, { "epoch": 0.97, "learning_rate": 3.376356190439222e-05, "loss": 0.0, "step": 50210 }, { "epoch": 0.97, "learning_rate": 3.376291494413498e-05, "loss": 0.0054, "step": 50212 }, { "epoch": 0.97, "learning_rate": 3.376226798387776e-05, "loss": 0.0045, "step": 50214 }, { "epoch": 0.97, "learning_rate": 3.376162102362052e-05, "loss": 0.0, "step": 50216 }, { "epoch": 0.97, "learning_rate": 3.376097406336329e-05, "loss": 0.0023, "step": 50218 }, { "epoch": 0.97, "learning_rate": 3.376032710310606e-05, "loss": 0.0, "step": 50220 }, { "epoch": 0.97, "learning_rate": 3.375968014284883e-05, "loss": 0.0, "step": 50222 }, { "epoch": 0.97, "learning_rate": 3.3759033182591596e-05, "loss": 0.0, "step": 50224 }, { "epoch": 0.97, "learning_rate": 3.3758386222334365e-05, "loss": 0.0, "step": 50226 }, { "epoch": 0.97, "learning_rate": 3.3757739262077134e-05, "loss": 0.0, "step": 50228 }, { "epoch": 0.97, "learning_rate": 3.3757092301819896e-05, "loss": 0.0, "step": 50230 }, { "epoch": 0.97, "learning_rate": 3.375644534156267e-05, "loss": 0.0, "step": 50232 }, { "epoch": 0.97, "learning_rate": 3.3755798381305435e-05, "loss": 0.0, "step": 50234 }, { "epoch": 0.98, "learning_rate": 3.375515142104821e-05, "loss": 0.0001, "step": 50236 }, { "epoch": 0.98, "learning_rate": 3.375450446079097e-05, "loss": 0.0, "step": 50238 }, { "epoch": 0.98, "learning_rate": 3.375385750053374e-05, "loss": 0.0, "step": 50240 }, { "epoch": 0.98, "learning_rate": 3.375321054027651e-05, "loss": 0.0001, "step": 50242 }, { "epoch": 0.98, "learning_rate": 3.375256358001928e-05, "loss": 0.0, "step": 50244 }, { "epoch": 0.98, "learning_rate": 3.375191661976205e-05, "loss": 0.0, "step": 50246 }, { "epoch": 0.98, "learning_rate": 3.375126965950482e-05, "loss": 0.0, "step": 50248 }, { "epoch": 0.98, "learning_rate": 3.375062269924759e-05, "loss": 0.0, "step": 50250 }, { "epoch": 0.98, "learning_rate": 3.3749975738990356e-05, "loss": 0.0001, "step": 50252 }, { "epoch": 0.98, "learning_rate": 3.3749328778733125e-05, "loss": 0.0, "step": 50254 }, { "epoch": 0.98, "learning_rate": 3.3748681818475894e-05, "loss": 0.0, "step": 50256 }, { "epoch": 0.98, "learning_rate": 3.3748034858218664e-05, "loss": 0.0007, "step": 50258 }, { "epoch": 0.98, "learning_rate": 3.374738789796143e-05, "loss": 0.0109, "step": 50260 }, { "epoch": 0.98, "learning_rate": 3.3746740937704195e-05, "loss": 0.0, "step": 50262 }, { "epoch": 0.98, "learning_rate": 3.374609397744697e-05, "loss": 0.0016, "step": 50264 }, { "epoch": 0.98, "learning_rate": 3.374544701718973e-05, "loss": 0.0, "step": 50266 }, { "epoch": 0.98, "learning_rate": 3.37448000569325e-05, "loss": 0.0109, "step": 50268 }, { "epoch": 0.98, "learning_rate": 3.374415309667527e-05, "loss": 0.0, "step": 50270 }, { "epoch": 0.98, "learning_rate": 3.374350613641804e-05, "loss": 0.0003, "step": 50272 }, { "epoch": 0.98, "learning_rate": 3.374285917616081e-05, "loss": 0.0019, "step": 50274 }, { "epoch": 0.98, "learning_rate": 3.374221221590358e-05, "loss": 0.0, "step": 50276 }, { "epoch": 0.98, "learning_rate": 3.374156525564635e-05, "loss": 0.0001, "step": 50278 }, { "epoch": 0.98, "learning_rate": 3.374091829538912e-05, "loss": 0.0, "step": 50280 }, { "epoch": 0.98, "learning_rate": 3.3740271335131886e-05, "loss": 0.0, "step": 50282 }, { "epoch": 0.98, "learning_rate": 3.373962437487465e-05, "loss": 0.0, "step": 50284 }, { "epoch": 0.98, "learning_rate": 3.3738977414617424e-05, "loss": 0.0, "step": 50286 }, { "epoch": 0.98, "learning_rate": 3.373833045436019e-05, "loss": 0.0, "step": 50288 }, { "epoch": 0.98, "learning_rate": 3.3737683494102955e-05, "loss": 0.0001, "step": 50290 }, { "epoch": 0.98, "learning_rate": 3.373703653384573e-05, "loss": 0.0001, "step": 50292 }, { "epoch": 0.98, "learning_rate": 3.3736389573588494e-05, "loss": 0.0, "step": 50294 }, { "epoch": 0.98, "learning_rate": 3.373574261333127e-05, "loss": 0.0, "step": 50296 }, { "epoch": 0.98, "learning_rate": 3.373509565307403e-05, "loss": 0.0011, "step": 50298 }, { "epoch": 0.98, "learning_rate": 3.37344486928168e-05, "loss": 0.0001, "step": 50300 }, { "epoch": 0.98, "learning_rate": 3.373380173255957e-05, "loss": 0.0001, "step": 50302 }, { "epoch": 0.98, "learning_rate": 3.373315477230234e-05, "loss": 0.0, "step": 50304 }, { "epoch": 0.98, "learning_rate": 3.373250781204511e-05, "loss": 0.0, "step": 50306 }, { "epoch": 0.98, "learning_rate": 3.373186085178788e-05, "loss": 0.0, "step": 50308 }, { "epoch": 0.98, "learning_rate": 3.3731213891530646e-05, "loss": 0.0, "step": 50310 }, { "epoch": 0.98, "learning_rate": 3.373056693127341e-05, "loss": 0.0, "step": 50312 }, { "epoch": 0.98, "learning_rate": 3.3729919971016184e-05, "loss": 0.0, "step": 50314 }, { "epoch": 0.98, "learning_rate": 3.372927301075895e-05, "loss": 0.0007, "step": 50316 }, { "epoch": 0.98, "learning_rate": 3.372862605050172e-05, "loss": 0.0001, "step": 50318 }, { "epoch": 0.98, "learning_rate": 3.3727979090244485e-05, "loss": 0.0, "step": 50320 }, { "epoch": 0.98, "learning_rate": 3.3727332129987254e-05, "loss": 0.0, "step": 50322 }, { "epoch": 0.98, "learning_rate": 3.372668516973003e-05, "loss": 0.0, "step": 50324 }, { "epoch": 0.98, "learning_rate": 3.372603820947279e-05, "loss": 0.0, "step": 50326 }, { "epoch": 0.98, "learning_rate": 3.372539124921556e-05, "loss": 0.0258, "step": 50328 }, { "epoch": 0.98, "learning_rate": 3.372474428895833e-05, "loss": 0.0005, "step": 50330 }, { "epoch": 0.98, "learning_rate": 3.37240973287011e-05, "loss": 0.0, "step": 50332 }, { "epoch": 0.98, "learning_rate": 3.372345036844387e-05, "loss": 0.0017, "step": 50334 }, { "epoch": 0.98, "learning_rate": 3.372280340818664e-05, "loss": 0.0152, "step": 50336 }, { "epoch": 0.98, "learning_rate": 3.3722156447929407e-05, "loss": 0.0, "step": 50338 }, { "epoch": 0.98, "learning_rate": 3.3721509487672176e-05, "loss": 0.0, "step": 50340 }, { "epoch": 0.98, "learning_rate": 3.3720862527414945e-05, "loss": 0.0, "step": 50342 }, { "epoch": 0.98, "learning_rate": 3.372021556715771e-05, "loss": 0.0015, "step": 50344 }, { "epoch": 0.98, "learning_rate": 3.371956860690048e-05, "loss": 0.0001, "step": 50346 }, { "epoch": 0.98, "learning_rate": 3.3718921646643245e-05, "loss": 0.0022, "step": 50348 }, { "epoch": 0.98, "learning_rate": 3.3718274686386014e-05, "loss": 0.0064, "step": 50350 }, { "epoch": 0.98, "learning_rate": 3.3717627726128783e-05, "loss": 0.0, "step": 50352 }, { "epoch": 0.98, "learning_rate": 3.371698076587155e-05, "loss": 0.0018, "step": 50354 }, { "epoch": 0.98, "learning_rate": 3.371633380561433e-05, "loss": 0.0, "step": 50356 }, { "epoch": 0.98, "learning_rate": 3.371568684535709e-05, "loss": 0.0, "step": 50358 }, { "epoch": 0.98, "learning_rate": 3.371503988509986e-05, "loss": 0.0, "step": 50360 }, { "epoch": 0.98, "learning_rate": 3.371439292484263e-05, "loss": 0.0046, "step": 50362 }, { "epoch": 0.98, "learning_rate": 3.37137459645854e-05, "loss": 0.0, "step": 50364 }, { "epoch": 0.98, "learning_rate": 3.371309900432816e-05, "loss": 0.0022, "step": 50366 }, { "epoch": 0.98, "learning_rate": 3.3712452044070936e-05, "loss": 0.0, "step": 50368 }, { "epoch": 0.98, "learning_rate": 3.3711805083813705e-05, "loss": 0.0019, "step": 50370 }, { "epoch": 0.98, "learning_rate": 3.371115812355647e-05, "loss": 0.0, "step": 50372 }, { "epoch": 0.98, "learning_rate": 3.371051116329924e-05, "loss": 0.0, "step": 50374 }, { "epoch": 0.98, "learning_rate": 3.3709864203042006e-05, "loss": 0.0, "step": 50376 }, { "epoch": 0.98, "learning_rate": 3.370921724278478e-05, "loss": 0.0, "step": 50378 }, { "epoch": 0.98, "learning_rate": 3.3708570282527544e-05, "loss": 0.0, "step": 50380 }, { "epoch": 0.98, "learning_rate": 3.370792332227031e-05, "loss": 0.0, "step": 50382 }, { "epoch": 0.98, "learning_rate": 3.370727636201308e-05, "loss": 0.0, "step": 50384 }, { "epoch": 0.98, "learning_rate": 3.370662940175585e-05, "loss": 0.0001, "step": 50386 }, { "epoch": 0.98, "learning_rate": 3.370598244149862e-05, "loss": 0.0, "step": 50388 }, { "epoch": 0.98, "learning_rate": 3.370533548124139e-05, "loss": 0.0, "step": 50390 }, { "epoch": 0.98, "learning_rate": 3.370468852098416e-05, "loss": 0.0001, "step": 50392 }, { "epoch": 0.98, "learning_rate": 3.370404156072693e-05, "loss": 0.0, "step": 50394 }, { "epoch": 0.98, "learning_rate": 3.3703394600469696e-05, "loss": 0.0, "step": 50396 }, { "epoch": 0.98, "learning_rate": 3.370274764021246e-05, "loss": 0.0206, "step": 50398 }, { "epoch": 0.98, "learning_rate": 3.3702100679955235e-05, "loss": 0.0, "step": 50400 }, { "epoch": 0.98, "learning_rate": 3.3701453719698004e-05, "loss": 0.0, "step": 50402 }, { "epoch": 0.98, "learning_rate": 3.3700806759440766e-05, "loss": 0.0, "step": 50404 }, { "epoch": 0.98, "learning_rate": 3.370015979918354e-05, "loss": 0.0, "step": 50406 }, { "epoch": 0.98, "learning_rate": 3.3699512838926304e-05, "loss": 0.0, "step": 50408 }, { "epoch": 0.98, "learning_rate": 3.369886587866907e-05, "loss": 0.0001, "step": 50410 }, { "epoch": 0.98, "learning_rate": 3.369821891841184e-05, "loss": 0.0001, "step": 50412 }, { "epoch": 0.98, "learning_rate": 3.369757195815461e-05, "loss": 0.0026, "step": 50414 }, { "epoch": 0.98, "learning_rate": 3.369692499789738e-05, "loss": 0.0019, "step": 50416 }, { "epoch": 0.98, "learning_rate": 3.369627803764015e-05, "loss": 0.0152, "step": 50418 }, { "epoch": 0.98, "learning_rate": 3.369563107738292e-05, "loss": 0.0, "step": 50420 }, { "epoch": 0.98, "learning_rate": 3.369498411712569e-05, "loss": 0.0, "step": 50422 }, { "epoch": 0.98, "learning_rate": 3.369433715686846e-05, "loss": 0.0021, "step": 50424 }, { "epoch": 0.98, "learning_rate": 3.369369019661122e-05, "loss": 0.0036, "step": 50426 }, { "epoch": 0.98, "learning_rate": 3.3693043236353995e-05, "loss": 0.0036, "step": 50428 }, { "epoch": 0.98, "learning_rate": 3.369239627609676e-05, "loss": 0.0, "step": 50430 }, { "epoch": 0.98, "learning_rate": 3.3691749315839526e-05, "loss": 0.0, "step": 50432 }, { "epoch": 0.98, "learning_rate": 3.36911023555823e-05, "loss": 0.0, "step": 50434 }, { "epoch": 0.98, "learning_rate": 3.3690455395325065e-05, "loss": 0.0, "step": 50436 }, { "epoch": 0.98, "learning_rate": 3.368980843506784e-05, "loss": 0.0, "step": 50438 }, { "epoch": 0.98, "learning_rate": 3.36891614748106e-05, "loss": 0.0, "step": 50440 }, { "epoch": 0.98, "learning_rate": 3.368851451455337e-05, "loss": 0.0, "step": 50442 }, { "epoch": 0.98, "learning_rate": 3.368786755429614e-05, "loss": 0.0001, "step": 50444 }, { "epoch": 0.98, "learning_rate": 3.368722059403891e-05, "loss": 0.0052, "step": 50446 }, { "epoch": 0.98, "learning_rate": 3.368657363378168e-05, "loss": 0.0002, "step": 50448 }, { "epoch": 0.98, "learning_rate": 3.368592667352445e-05, "loss": 0.0072, "step": 50450 }, { "epoch": 0.98, "learning_rate": 3.368527971326722e-05, "loss": 0.0001, "step": 50452 }, { "epoch": 0.98, "learning_rate": 3.368463275300998e-05, "loss": 0.0, "step": 50454 }, { "epoch": 0.98, "learning_rate": 3.3683985792752755e-05, "loss": 0.0, "step": 50456 }, { "epoch": 0.98, "learning_rate": 3.368333883249552e-05, "loss": 0.0, "step": 50458 }, { "epoch": 0.98, "learning_rate": 3.3682691872238294e-05, "loss": 0.0011, "step": 50460 }, { "epoch": 0.98, "learning_rate": 3.3682044911981056e-05, "loss": 0.0007, "step": 50462 }, { "epoch": 0.98, "learning_rate": 3.3681397951723825e-05, "loss": 0.0, "step": 50464 }, { "epoch": 0.98, "learning_rate": 3.36807509914666e-05, "loss": 0.0, "step": 50466 }, { "epoch": 0.98, "learning_rate": 3.368010403120936e-05, "loss": 0.0, "step": 50468 }, { "epoch": 0.98, "learning_rate": 3.367945707095213e-05, "loss": 0.0, "step": 50470 }, { "epoch": 0.98, "learning_rate": 3.36788101106949e-05, "loss": 0.0, "step": 50472 }, { "epoch": 0.98, "learning_rate": 3.367816315043767e-05, "loss": 0.0006, "step": 50474 }, { "epoch": 0.98, "learning_rate": 3.367751619018044e-05, "loss": 0.0021, "step": 50476 }, { "epoch": 0.98, "learning_rate": 3.367686922992321e-05, "loss": 0.0292, "step": 50478 }, { "epoch": 0.98, "learning_rate": 3.367622226966598e-05, "loss": 0.0, "step": 50480 }, { "epoch": 0.98, "learning_rate": 3.367557530940875e-05, "loss": 0.0002, "step": 50482 }, { "epoch": 0.98, "learning_rate": 3.3674928349151516e-05, "loss": 0.0006, "step": 50484 }, { "epoch": 0.98, "learning_rate": 3.367428138889428e-05, "loss": 0.0, "step": 50486 }, { "epoch": 0.98, "learning_rate": 3.3673634428637054e-05, "loss": 0.0132, "step": 50488 }, { "epoch": 0.98, "learning_rate": 3.3672987468379816e-05, "loss": 0.0001, "step": 50490 }, { "epoch": 0.98, "learning_rate": 3.3672340508122585e-05, "loss": 0.0042, "step": 50492 }, { "epoch": 0.98, "learning_rate": 3.3671693547865355e-05, "loss": 0.0, "step": 50494 }, { "epoch": 0.98, "learning_rate": 3.3671046587608124e-05, "loss": 0.0, "step": 50496 }, { "epoch": 0.98, "learning_rate": 3.367039962735089e-05, "loss": 0.0001, "step": 50498 }, { "epoch": 0.98, "learning_rate": 3.366975266709366e-05, "loss": 0.0, "step": 50500 }, { "epoch": 0.98, "learning_rate": 3.366910570683643e-05, "loss": 0.0, "step": 50502 }, { "epoch": 0.98, "learning_rate": 3.36684587465792e-05, "loss": 0.0, "step": 50504 }, { "epoch": 0.98, "learning_rate": 3.366781178632197e-05, "loss": 0.0, "step": 50506 }, { "epoch": 0.98, "learning_rate": 3.366716482606473e-05, "loss": 0.0, "step": 50508 }, { "epoch": 0.98, "learning_rate": 3.366651786580751e-05, "loss": 0.0, "step": 50510 }, { "epoch": 0.98, "learning_rate": 3.3665870905550276e-05, "loss": 0.0, "step": 50512 }, { "epoch": 0.98, "learning_rate": 3.366522394529304e-05, "loss": 0.0, "step": 50514 }, { "epoch": 0.98, "learning_rate": 3.3664576985035814e-05, "loss": 0.0, "step": 50516 }, { "epoch": 0.98, "learning_rate": 3.366393002477858e-05, "loss": 0.0063, "step": 50518 }, { "epoch": 0.98, "learning_rate": 3.366328306452135e-05, "loss": 0.0076, "step": 50520 }, { "epoch": 0.98, "learning_rate": 3.3662636104264115e-05, "loss": 0.0072, "step": 50522 }, { "epoch": 0.98, "learning_rate": 3.3661989144006884e-05, "loss": 0.0, "step": 50524 }, { "epoch": 0.98, "learning_rate": 3.366134218374965e-05, "loss": 0.0, "step": 50526 }, { "epoch": 0.98, "learning_rate": 3.366069522349242e-05, "loss": 0.0, "step": 50528 }, { "epoch": 0.98, "learning_rate": 3.366004826323519e-05, "loss": 0.0002, "step": 50530 }, { "epoch": 0.98, "learning_rate": 3.365940130297796e-05, "loss": 0.0, "step": 50532 }, { "epoch": 0.98, "learning_rate": 3.365875434272073e-05, "loss": 0.0, "step": 50534 }, { "epoch": 0.98, "learning_rate": 3.36581073824635e-05, "loss": 0.0007, "step": 50536 }, { "epoch": 0.98, "learning_rate": 3.365746042220627e-05, "loss": 0.0003, "step": 50538 }, { "epoch": 0.98, "learning_rate": 3.365681346194903e-05, "loss": 0.0001, "step": 50540 }, { "epoch": 0.98, "learning_rate": 3.3656166501691806e-05, "loss": 0.0, "step": 50542 }, { "epoch": 0.98, "learning_rate": 3.365551954143457e-05, "loss": 0.0, "step": 50544 }, { "epoch": 0.98, "learning_rate": 3.365487258117734e-05, "loss": 0.0, "step": 50546 }, { "epoch": 0.98, "learning_rate": 3.365422562092011e-05, "loss": 0.0003, "step": 50548 }, { "epoch": 0.98, "learning_rate": 3.3653578660662875e-05, "loss": 0.0003, "step": 50550 }, { "epoch": 0.98, "learning_rate": 3.3652931700405644e-05, "loss": 0.0366, "step": 50552 }, { "epoch": 0.98, "learning_rate": 3.3652284740148414e-05, "loss": 0.0, "step": 50554 }, { "epoch": 0.98, "learning_rate": 3.365163777989118e-05, "loss": 0.0, "step": 50556 }, { "epoch": 0.98, "learning_rate": 3.365099081963395e-05, "loss": 0.0, "step": 50558 }, { "epoch": 0.98, "learning_rate": 3.365034385937672e-05, "loss": 0.0, "step": 50560 }, { "epoch": 0.98, "learning_rate": 3.364969689911949e-05, "loss": 0.0107, "step": 50562 }, { "epoch": 0.98, "learning_rate": 3.364904993886226e-05, "loss": 0.0026, "step": 50564 }, { "epoch": 0.98, "learning_rate": 3.364840297860503e-05, "loss": 0.0, "step": 50566 }, { "epoch": 0.98, "learning_rate": 3.364775601834779e-05, "loss": 0.0001, "step": 50568 }, { "epoch": 0.98, "learning_rate": 3.3647109058090566e-05, "loss": 0.0, "step": 50570 }, { "epoch": 0.98, "learning_rate": 3.364646209783333e-05, "loss": 0.0, "step": 50572 }, { "epoch": 0.98, "learning_rate": 3.36458151375761e-05, "loss": 0.0009, "step": 50574 }, { "epoch": 0.98, "learning_rate": 3.364516817731887e-05, "loss": 0.0, "step": 50576 }, { "epoch": 0.98, "learning_rate": 3.3644521217061636e-05, "loss": 0.0, "step": 50578 }, { "epoch": 0.98, "learning_rate": 3.364387425680441e-05, "loss": 0.0, "step": 50580 }, { "epoch": 0.98, "learning_rate": 3.3643227296547174e-05, "loss": 0.0, "step": 50582 }, { "epoch": 0.98, "learning_rate": 3.364258033628994e-05, "loss": 0.0, "step": 50584 }, { "epoch": 0.98, "learning_rate": 3.364193337603271e-05, "loss": 0.002, "step": 50586 }, { "epoch": 0.98, "learning_rate": 3.364128641577548e-05, "loss": 0.0, "step": 50588 }, { "epoch": 0.98, "learning_rate": 3.364063945551825e-05, "loss": 0.001, "step": 50590 }, { "epoch": 0.98, "learning_rate": 3.363999249526102e-05, "loss": 0.0, "step": 50592 }, { "epoch": 0.98, "learning_rate": 3.363934553500379e-05, "loss": 0.0, "step": 50594 }, { "epoch": 0.98, "learning_rate": 3.363869857474655e-05, "loss": 0.0, "step": 50596 }, { "epoch": 0.98, "learning_rate": 3.3638051614489327e-05, "loss": 0.0001, "step": 50598 }, { "epoch": 0.98, "learning_rate": 3.363740465423209e-05, "loss": 0.0, "step": 50600 }, { "epoch": 0.98, "learning_rate": 3.3636757693974865e-05, "loss": 0.002, "step": 50602 }, { "epoch": 0.98, "learning_rate": 3.363611073371763e-05, "loss": 0.0, "step": 50604 }, { "epoch": 0.98, "learning_rate": 3.3635463773460396e-05, "loss": 0.0, "step": 50606 }, { "epoch": 0.98, "learning_rate": 3.3634816813203165e-05, "loss": 0.0115, "step": 50608 }, { "epoch": 0.98, "learning_rate": 3.3634169852945934e-05, "loss": 0.0, "step": 50610 }, { "epoch": 0.98, "learning_rate": 3.3633522892688703e-05, "loss": 0.0, "step": 50612 }, { "epoch": 0.98, "learning_rate": 3.363287593243147e-05, "loss": 0.0, "step": 50614 }, { "epoch": 0.98, "learning_rate": 3.363222897217424e-05, "loss": 0.0023, "step": 50616 }, { "epoch": 0.98, "learning_rate": 3.363158201191701e-05, "loss": 0.0001, "step": 50618 }, { "epoch": 0.98, "learning_rate": 3.363093505165978e-05, "loss": 0.0, "step": 50620 }, { "epoch": 0.98, "learning_rate": 3.363028809140254e-05, "loss": 0.0, "step": 50622 }, { "epoch": 0.98, "learning_rate": 3.362964113114532e-05, "loss": 0.0, "step": 50624 }, { "epoch": 0.98, "learning_rate": 3.362899417088809e-05, "loss": 0.0025, "step": 50626 }, { "epoch": 0.98, "learning_rate": 3.362834721063085e-05, "loss": 0.0, "step": 50628 }, { "epoch": 0.98, "learning_rate": 3.3627700250373625e-05, "loss": 0.0002, "step": 50630 }, { "epoch": 0.98, "learning_rate": 3.362705329011639e-05, "loss": 0.0, "step": 50632 }, { "epoch": 0.98, "learning_rate": 3.3626406329859157e-05, "loss": 0.0, "step": 50634 }, { "epoch": 0.98, "learning_rate": 3.3625759369601926e-05, "loss": 0.0, "step": 50636 }, { "epoch": 0.98, "learning_rate": 3.3625112409344695e-05, "loss": 0.0004, "step": 50638 }, { "epoch": 0.98, "learning_rate": 3.3624465449087464e-05, "loss": 0.0, "step": 50640 }, { "epoch": 0.98, "learning_rate": 3.362381848883023e-05, "loss": 0.0, "step": 50642 }, { "epoch": 0.98, "learning_rate": 3.3623171528573e-05, "loss": 0.0001, "step": 50644 }, { "epoch": 0.98, "learning_rate": 3.362252456831577e-05, "loss": 0.0046, "step": 50646 }, { "epoch": 0.98, "learning_rate": 3.362187760805854e-05, "loss": 0.0001, "step": 50648 }, { "epoch": 0.98, "learning_rate": 3.36212306478013e-05, "loss": 0.0, "step": 50650 }, { "epoch": 0.98, "learning_rate": 3.362058368754408e-05, "loss": 0.0, "step": 50652 }, { "epoch": 0.98, "learning_rate": 3.361993672728684e-05, "loss": 0.0, "step": 50654 }, { "epoch": 0.98, "learning_rate": 3.361928976702961e-05, "loss": 0.0, "step": 50656 }, { "epoch": 0.98, "learning_rate": 3.3618642806772386e-05, "loss": 0.0007, "step": 50658 }, { "epoch": 0.98, "learning_rate": 3.361799584651515e-05, "loss": 0.0195, "step": 50660 }, { "epoch": 0.98, "learning_rate": 3.3617348886257924e-05, "loss": 0.0003, "step": 50662 }, { "epoch": 0.98, "learning_rate": 3.3616701926000686e-05, "loss": 0.0, "step": 50664 }, { "epoch": 0.98, "learning_rate": 3.3616054965743455e-05, "loss": 0.0, "step": 50666 }, { "epoch": 0.98, "learning_rate": 3.3615408005486224e-05, "loss": 0.0001, "step": 50668 }, { "epoch": 0.98, "learning_rate": 3.361476104522899e-05, "loss": 0.0, "step": 50670 }, { "epoch": 0.98, "learning_rate": 3.361411408497176e-05, "loss": 0.0, "step": 50672 }, { "epoch": 0.98, "learning_rate": 3.361346712471453e-05, "loss": 0.0, "step": 50674 }, { "epoch": 0.98, "learning_rate": 3.36128201644573e-05, "loss": 0.0, "step": 50676 }, { "epoch": 0.98, "learning_rate": 3.361217320420007e-05, "loss": 0.0001, "step": 50678 }, { "epoch": 0.98, "learning_rate": 3.361152624394284e-05, "loss": 0.0079, "step": 50680 }, { "epoch": 0.98, "learning_rate": 3.36108792836856e-05, "loss": 0.0018, "step": 50682 }, { "epoch": 0.98, "learning_rate": 3.361023232342838e-05, "loss": 0.0003, "step": 50684 }, { "epoch": 0.98, "learning_rate": 3.360958536317114e-05, "loss": 0.0, "step": 50686 }, { "epoch": 0.98, "learning_rate": 3.360893840291391e-05, "loss": 0.0, "step": 50688 }, { "epoch": 0.98, "learning_rate": 3.3608291442656684e-05, "loss": 0.0, "step": 50690 }, { "epoch": 0.98, "learning_rate": 3.3607644482399446e-05, "loss": 0.0003, "step": 50692 }, { "epoch": 0.98, "learning_rate": 3.3606997522142216e-05, "loss": 0.0002, "step": 50694 }, { "epoch": 0.98, "learning_rate": 3.3606350561884985e-05, "loss": 0.0, "step": 50696 }, { "epoch": 0.98, "learning_rate": 3.3605703601627754e-05, "loss": 0.0, "step": 50698 }, { "epoch": 0.98, "learning_rate": 3.360505664137052e-05, "loss": 0.0002, "step": 50700 }, { "epoch": 0.98, "learning_rate": 3.360440968111329e-05, "loss": 0.0001, "step": 50702 }, { "epoch": 0.98, "learning_rate": 3.360376272085606e-05, "loss": 0.0, "step": 50704 }, { "epoch": 0.98, "learning_rate": 3.360311576059883e-05, "loss": 0.0, "step": 50706 }, { "epoch": 0.98, "learning_rate": 3.36024688003416e-05, "loss": 0.0015, "step": 50708 }, { "epoch": 0.98, "learning_rate": 3.360182184008436e-05, "loss": 0.0, "step": 50710 }, { "epoch": 0.98, "learning_rate": 3.360117487982714e-05, "loss": 0.0001, "step": 50712 }, { "epoch": 0.98, "learning_rate": 3.36005279195699e-05, "loss": 0.0051, "step": 50714 }, { "epoch": 0.98, "learning_rate": 3.359988095931267e-05, "loss": 0.0001, "step": 50716 }, { "epoch": 0.98, "learning_rate": 3.359923399905544e-05, "loss": 0.0, "step": 50718 }, { "epoch": 0.98, "learning_rate": 3.359858703879821e-05, "loss": 0.0, "step": 50720 }, { "epoch": 0.98, "learning_rate": 3.3597940078540976e-05, "loss": 0.0001, "step": 50722 }, { "epoch": 0.98, "learning_rate": 3.3597293118283745e-05, "loss": 0.0, "step": 50724 }, { "epoch": 0.98, "learning_rate": 3.3596646158026514e-05, "loss": 0.0004, "step": 50726 }, { "epoch": 0.98, "learning_rate": 3.359599919776928e-05, "loss": 0.0, "step": 50728 }, { "epoch": 0.98, "learning_rate": 3.359535223751205e-05, "loss": 0.0, "step": 50730 }, { "epoch": 0.98, "learning_rate": 3.3594705277254815e-05, "loss": 0.0, "step": 50732 }, { "epoch": 0.98, "learning_rate": 3.359405831699759e-05, "loss": 0.0, "step": 50734 }, { "epoch": 0.98, "learning_rate": 3.359341135674036e-05, "loss": 0.0, "step": 50736 }, { "epoch": 0.98, "learning_rate": 3.359276439648312e-05, "loss": 0.0, "step": 50738 }, { "epoch": 0.98, "learning_rate": 3.35921174362259e-05, "loss": 0.0031, "step": 50740 }, { "epoch": 0.98, "learning_rate": 3.359147047596866e-05, "loss": 0.0006, "step": 50742 }, { "epoch": 0.98, "learning_rate": 3.3590823515711436e-05, "loss": 0.0067, "step": 50744 }, { "epoch": 0.98, "learning_rate": 3.35901765554542e-05, "loss": 0.0, "step": 50746 }, { "epoch": 0.98, "learning_rate": 3.358952959519697e-05, "loss": 0.0044, "step": 50748 }, { "epoch": 0.98, "learning_rate": 3.3588882634939736e-05, "loss": 0.0001, "step": 50750 }, { "epoch": 0.99, "learning_rate": 3.3588235674682505e-05, "loss": 0.0004, "step": 50752 }, { "epoch": 0.99, "learning_rate": 3.3587588714425275e-05, "loss": 0.0002, "step": 50754 }, { "epoch": 0.99, "learning_rate": 3.3586941754168044e-05, "loss": 0.0067, "step": 50756 }, { "epoch": 0.99, "learning_rate": 3.358629479391081e-05, "loss": 0.0002, "step": 50758 }, { "epoch": 0.99, "learning_rate": 3.358564783365358e-05, "loss": 0.0, "step": 50760 }, { "epoch": 0.99, "learning_rate": 3.358500087339635e-05, "loss": 0.0004, "step": 50762 }, { "epoch": 0.99, "learning_rate": 3.358435391313911e-05, "loss": 0.0, "step": 50764 }, { "epoch": 0.99, "learning_rate": 3.358370695288189e-05, "loss": 0.0, "step": 50766 }, { "epoch": 0.99, "learning_rate": 3.358305999262466e-05, "loss": 0.0, "step": 50768 }, { "epoch": 0.99, "learning_rate": 3.358241303236742e-05, "loss": 0.0, "step": 50770 }, { "epoch": 0.99, "learning_rate": 3.3581766072110196e-05, "loss": 0.0, "step": 50772 }, { "epoch": 0.99, "learning_rate": 3.358111911185296e-05, "loss": 0.0, "step": 50774 }, { "epoch": 0.99, "learning_rate": 3.358047215159573e-05, "loss": 0.0, "step": 50776 }, { "epoch": 0.99, "learning_rate": 3.35798251913385e-05, "loss": 0.0, "step": 50778 }, { "epoch": 0.99, "learning_rate": 3.3579178231081266e-05, "loss": 0.0, "step": 50780 }, { "epoch": 0.99, "learning_rate": 3.3578531270824035e-05, "loss": 0.0001, "step": 50782 }, { "epoch": 0.99, "learning_rate": 3.3577884310566804e-05, "loss": 0.0, "step": 50784 }, { "epoch": 0.99, "learning_rate": 3.357723735030957e-05, "loss": 0.0001, "step": 50786 }, { "epoch": 0.99, "learning_rate": 3.357659039005234e-05, "loss": 0.0, "step": 50788 }, { "epoch": 0.99, "learning_rate": 3.357594342979511e-05, "loss": 0.0008, "step": 50790 }, { "epoch": 0.99, "learning_rate": 3.3575296469537874e-05, "loss": 0.0, "step": 50792 }, { "epoch": 0.99, "learning_rate": 3.357464950928065e-05, "loss": 0.0001, "step": 50794 }, { "epoch": 0.99, "learning_rate": 3.357400254902341e-05, "loss": 0.0, "step": 50796 }, { "epoch": 0.99, "learning_rate": 3.357335558876618e-05, "loss": 0.0056, "step": 50798 }, { "epoch": 0.99, "learning_rate": 3.357270862850895e-05, "loss": 0.0, "step": 50800 }, { "epoch": 0.99, "learning_rate": 3.357206166825172e-05, "loss": 0.0001, "step": 50802 }, { "epoch": 0.99, "learning_rate": 3.3571414707994495e-05, "loss": 0.015, "step": 50804 }, { "epoch": 0.99, "learning_rate": 3.357076774773726e-05, "loss": 0.0044, "step": 50806 }, { "epoch": 0.99, "learning_rate": 3.3570120787480026e-05, "loss": 0.0002, "step": 50808 }, { "epoch": 0.99, "learning_rate": 3.3569473827222795e-05, "loss": 0.0002, "step": 50810 }, { "epoch": 0.99, "learning_rate": 3.3568826866965564e-05, "loss": 0.0, "step": 50812 }, { "epoch": 0.99, "learning_rate": 3.3568179906708333e-05, "loss": 0.0, "step": 50814 }, { "epoch": 0.99, "learning_rate": 3.35675329464511e-05, "loss": 0.0, "step": 50816 }, { "epoch": 0.99, "learning_rate": 3.356688598619387e-05, "loss": 0.0001, "step": 50818 }, { "epoch": 0.99, "learning_rate": 3.356623902593664e-05, "loss": 0.0006, "step": 50820 }, { "epoch": 0.99, "learning_rate": 3.356559206567941e-05, "loss": 0.0, "step": 50822 }, { "epoch": 0.99, "learning_rate": 3.356494510542217e-05, "loss": 0.0001, "step": 50824 }, { "epoch": 0.99, "learning_rate": 3.356429814516495e-05, "loss": 0.0, "step": 50826 }, { "epoch": 0.99, "learning_rate": 3.356365118490771e-05, "loss": 0.0, "step": 50828 }, { "epoch": 0.99, "learning_rate": 3.356300422465048e-05, "loss": 0.0, "step": 50830 }, { "epoch": 0.99, "learning_rate": 3.356235726439325e-05, "loss": 0.0, "step": 50832 }, { "epoch": 0.99, "learning_rate": 3.356171030413602e-05, "loss": 0.0001, "step": 50834 }, { "epoch": 0.99, "learning_rate": 3.356106334387879e-05, "loss": 0.0, "step": 50836 }, { "epoch": 0.99, "learning_rate": 3.3560416383621556e-05, "loss": 0.0005, "step": 50838 }, { "epoch": 0.99, "learning_rate": 3.3559769423364325e-05, "loss": 0.0, "step": 50840 }, { "epoch": 0.99, "learning_rate": 3.3559122463107094e-05, "loss": 0.0004, "step": 50842 }, { "epoch": 0.99, "learning_rate": 3.355847550284986e-05, "loss": 0.0003, "step": 50844 }, { "epoch": 0.99, "learning_rate": 3.3557828542592625e-05, "loss": 0.0, "step": 50846 }, { "epoch": 0.99, "learning_rate": 3.35571815823354e-05, "loss": 0.0052, "step": 50848 }, { "epoch": 0.99, "learning_rate": 3.355653462207817e-05, "loss": 0.0024, "step": 50850 }, { "epoch": 0.99, "learning_rate": 3.355588766182093e-05, "loss": 0.0001, "step": 50852 }, { "epoch": 0.99, "learning_rate": 3.355524070156371e-05, "loss": 0.0074, "step": 50854 }, { "epoch": 0.99, "learning_rate": 3.355459374130647e-05, "loss": 0.0, "step": 50856 }, { "epoch": 0.99, "learning_rate": 3.355394678104924e-05, "loss": 0.0, "step": 50858 }, { "epoch": 0.99, "learning_rate": 3.355329982079201e-05, "loss": 0.0, "step": 50860 }, { "epoch": 0.99, "learning_rate": 3.355265286053478e-05, "loss": 0.0001, "step": 50862 }, { "epoch": 0.99, "learning_rate": 3.355200590027755e-05, "loss": 0.0, "step": 50864 }, { "epoch": 0.99, "learning_rate": 3.3551358940020316e-05, "loss": 0.0, "step": 50866 }, { "epoch": 0.99, "learning_rate": 3.3550711979763085e-05, "loss": 0.0, "step": 50868 }, { "epoch": 0.99, "learning_rate": 3.3550065019505854e-05, "loss": 0.0, "step": 50870 }, { "epoch": 0.99, "learning_rate": 3.354941805924862e-05, "loss": 0.0, "step": 50872 }, { "epoch": 0.99, "learning_rate": 3.3548771098991386e-05, "loss": 0.0038, "step": 50874 }, { "epoch": 0.99, "learning_rate": 3.354812413873416e-05, "loss": 0.0009, "step": 50876 }, { "epoch": 0.99, "learning_rate": 3.3547477178476924e-05, "loss": 0.0, "step": 50878 }, { "epoch": 0.99, "learning_rate": 3.354683021821969e-05, "loss": 0.0001, "step": 50880 }, { "epoch": 0.99, "learning_rate": 3.354618325796247e-05, "loss": 0.0, "step": 50882 }, { "epoch": 0.99, "learning_rate": 3.354553629770523e-05, "loss": 0.0, "step": 50884 }, { "epoch": 0.99, "learning_rate": 3.354488933744801e-05, "loss": 0.0037, "step": 50886 }, { "epoch": 0.99, "learning_rate": 3.354424237719077e-05, "loss": 0.0, "step": 50888 }, { "epoch": 0.99, "learning_rate": 3.354359541693354e-05, "loss": 0.0, "step": 50890 }, { "epoch": 0.99, "learning_rate": 3.354294845667631e-05, "loss": 0.0, "step": 50892 }, { "epoch": 0.99, "learning_rate": 3.3542301496419077e-05, "loss": 0.0, "step": 50894 }, { "epoch": 0.99, "learning_rate": 3.3541654536161846e-05, "loss": 0.0, "step": 50896 }, { "epoch": 0.99, "learning_rate": 3.3541007575904615e-05, "loss": 0.0, "step": 50898 }, { "epoch": 0.99, "learning_rate": 3.3540360615647384e-05, "loss": 0.0, "step": 50900 }, { "epoch": 0.99, "learning_rate": 3.353971365539015e-05, "loss": 0.0, "step": 50902 }, { "epoch": 0.99, "learning_rate": 3.353906669513292e-05, "loss": 0.0, "step": 50904 }, { "epoch": 0.99, "learning_rate": 3.3538419734875684e-05, "loss": 0.0004, "step": 50906 }, { "epoch": 0.99, "learning_rate": 3.353777277461846e-05, "loss": 0.0, "step": 50908 }, { "epoch": 0.99, "learning_rate": 3.353712581436122e-05, "loss": 0.0, "step": 50910 }, { "epoch": 0.99, "learning_rate": 3.353647885410399e-05, "loss": 0.0, "step": 50912 }, { "epoch": 0.99, "learning_rate": 3.353583189384677e-05, "loss": 0.0, "step": 50914 }, { "epoch": 0.99, "learning_rate": 3.353518493358953e-05, "loss": 0.0, "step": 50916 }, { "epoch": 0.99, "learning_rate": 3.35345379733323e-05, "loss": 0.0, "step": 50918 }, { "epoch": 0.99, "learning_rate": 3.353389101307507e-05, "loss": 0.0, "step": 50920 }, { "epoch": 0.99, "learning_rate": 3.353324405281784e-05, "loss": 0.0, "step": 50922 }, { "epoch": 0.99, "learning_rate": 3.3532597092560606e-05, "loss": 0.0, "step": 50924 }, { "epoch": 0.99, "learning_rate": 3.3531950132303375e-05, "loss": 0.003, "step": 50926 }, { "epoch": 0.99, "learning_rate": 3.3531303172046144e-05, "loss": 0.0095, "step": 50928 }, { "epoch": 0.99, "learning_rate": 3.353065621178891e-05, "loss": 0.0, "step": 50930 }, { "epoch": 0.99, "learning_rate": 3.353000925153168e-05, "loss": 0.0, "step": 50932 }, { "epoch": 0.99, "learning_rate": 3.3529362291274445e-05, "loss": 0.0, "step": 50934 }, { "epoch": 0.99, "learning_rate": 3.352871533101722e-05, "loss": 0.0, "step": 50936 }, { "epoch": 0.99, "learning_rate": 3.352806837075998e-05, "loss": 0.0, "step": 50938 }, { "epoch": 0.99, "learning_rate": 3.352742141050275e-05, "loss": 0.0033, "step": 50940 }, { "epoch": 0.99, "learning_rate": 3.352677445024552e-05, "loss": 0.0, "step": 50942 }, { "epoch": 0.99, "learning_rate": 3.352612748998829e-05, "loss": 0.0, "step": 50944 }, { "epoch": 0.99, "learning_rate": 3.3525480529731066e-05, "loss": 0.0, "step": 50946 }, { "epoch": 0.99, "learning_rate": 3.352483356947383e-05, "loss": 0.0, "step": 50948 }, { "epoch": 0.99, "learning_rate": 3.35241866092166e-05, "loss": 0.0, "step": 50950 }, { "epoch": 0.99, "learning_rate": 3.3523539648959366e-05, "loss": 0.0, "step": 50952 }, { "epoch": 0.99, "learning_rate": 3.3522892688702136e-05, "loss": 0.0, "step": 50954 }, { "epoch": 0.99, "learning_rate": 3.35222457284449e-05, "loss": 0.0, "step": 50956 }, { "epoch": 0.99, "learning_rate": 3.3521598768187674e-05, "loss": 0.0116, "step": 50958 }, { "epoch": 0.99, "learning_rate": 3.352095180793044e-05, "loss": 0.0057, "step": 50960 }, { "epoch": 0.99, "learning_rate": 3.352030484767321e-05, "loss": 0.0, "step": 50962 }, { "epoch": 0.99, "learning_rate": 3.351965788741598e-05, "loss": 0.0, "step": 50964 }, { "epoch": 0.99, "learning_rate": 3.351901092715874e-05, "loss": 0.0097, "step": 50966 }, { "epoch": 0.99, "learning_rate": 3.351836396690152e-05, "loss": 0.0038, "step": 50968 }, { "epoch": 0.99, "learning_rate": 3.351771700664428e-05, "loss": 0.0, "step": 50970 }, { "epoch": 0.99, "learning_rate": 3.351707004638705e-05, "loss": 0.0, "step": 50972 }, { "epoch": 0.99, "learning_rate": 3.351642308612982e-05, "loss": 0.0004, "step": 50974 }, { "epoch": 0.99, "learning_rate": 3.351577612587259e-05, "loss": 0.0, "step": 50976 }, { "epoch": 0.99, "learning_rate": 3.351512916561536e-05, "loss": 0.0, "step": 50978 }, { "epoch": 0.99, "learning_rate": 3.351448220535813e-05, "loss": 0.0, "step": 50980 }, { "epoch": 0.99, "learning_rate": 3.3513835245100896e-05, "loss": 0.0, "step": 50982 }, { "epoch": 0.99, "learning_rate": 3.3513188284843665e-05, "loss": 0.0, "step": 50984 }, { "epoch": 0.99, "learning_rate": 3.3512541324586434e-05, "loss": 0.0132, "step": 50986 }, { "epoch": 0.99, "learning_rate": 3.3511894364329196e-05, "loss": 0.0, "step": 50988 }, { "epoch": 0.99, "learning_rate": 3.351124740407197e-05, "loss": 0.0, "step": 50990 }, { "epoch": 0.99, "learning_rate": 3.351060044381474e-05, "loss": 0.0, "step": 50992 }, { "epoch": 0.99, "learning_rate": 3.3509953483557504e-05, "loss": 0.0004, "step": 50994 }, { "epoch": 0.99, "learning_rate": 3.350930652330028e-05, "loss": 0.0, "step": 50996 }, { "epoch": 0.99, "learning_rate": 3.350865956304304e-05, "loss": 0.0007, "step": 50998 }, { "epoch": 0.99, "learning_rate": 3.350801260278581e-05, "loss": 0.0, "step": 51000 }, { "epoch": 0.99, "learning_rate": 3.350736564252858e-05, "loss": 0.0, "step": 51002 }, { "epoch": 0.99, "learning_rate": 3.350671868227135e-05, "loss": 0.0063, "step": 51004 }, { "epoch": 0.99, "learning_rate": 3.350607172201412e-05, "loss": 0.0001, "step": 51006 }, { "epoch": 0.99, "learning_rate": 3.350542476175689e-05, "loss": 0.0007, "step": 51008 }, { "epoch": 0.99, "learning_rate": 3.3504777801499656e-05, "loss": 0.0015, "step": 51010 }, { "epoch": 0.99, "learning_rate": 3.3504130841242425e-05, "loss": 0.0, "step": 51012 }, { "epoch": 0.99, "learning_rate": 3.3503483880985194e-05, "loss": 0.0001, "step": 51014 }, { "epoch": 0.99, "learning_rate": 3.350283692072796e-05, "loss": 0.0001, "step": 51016 }, { "epoch": 0.99, "learning_rate": 3.350218996047073e-05, "loss": 0.0114, "step": 51018 }, { "epoch": 0.99, "learning_rate": 3.3501543000213495e-05, "loss": 0.0002, "step": 51020 }, { "epoch": 0.99, "learning_rate": 3.3500896039956264e-05, "loss": 0.0, "step": 51022 }, { "epoch": 0.99, "learning_rate": 3.350024907969903e-05, "loss": 0.0, "step": 51024 }, { "epoch": 0.99, "learning_rate": 3.34996021194418e-05, "loss": 0.0096, "step": 51026 }, { "epoch": 0.99, "learning_rate": 3.349895515918458e-05, "loss": 0.0008, "step": 51028 }, { "epoch": 0.99, "learning_rate": 3.349830819892734e-05, "loss": 0.0, "step": 51030 }, { "epoch": 0.99, "learning_rate": 3.349766123867011e-05, "loss": 0.0001, "step": 51032 }, { "epoch": 0.99, "learning_rate": 3.349701427841288e-05, "loss": 0.0, "step": 51034 }, { "epoch": 0.99, "learning_rate": 3.349636731815565e-05, "loss": 0.0006, "step": 51036 }, { "epoch": 0.99, "learning_rate": 3.349572035789842e-05, "loss": 0.0, "step": 51038 }, { "epoch": 0.99, "learning_rate": 3.3495073397641186e-05, "loss": 0.0, "step": 51040 }, { "epoch": 0.99, "learning_rate": 3.3494426437383955e-05, "loss": 0.0062, "step": 51042 }, { "epoch": 0.99, "learning_rate": 3.3493779477126724e-05, "loss": 0.0, "step": 51044 }, { "epoch": 0.99, "learning_rate": 3.349313251686949e-05, "loss": 0.0, "step": 51046 }, { "epoch": 0.99, "learning_rate": 3.3492485556612255e-05, "loss": 0.0019, "step": 51048 }, { "epoch": 0.99, "learning_rate": 3.349183859635503e-05, "loss": 0.0001, "step": 51050 }, { "epoch": 0.99, "learning_rate": 3.3491191636097794e-05, "loss": 0.0, "step": 51052 }, { "epoch": 0.99, "learning_rate": 3.349054467584056e-05, "loss": 0.0003, "step": 51054 }, { "epoch": 0.99, "learning_rate": 3.348989771558333e-05, "loss": 0.0001, "step": 51056 }, { "epoch": 0.99, "learning_rate": 3.34892507553261e-05, "loss": 0.0, "step": 51058 }, { "epoch": 0.99, "learning_rate": 3.348860379506887e-05, "loss": 0.0, "step": 51060 }, { "epoch": 0.99, "learning_rate": 3.348795683481164e-05, "loss": 0.0, "step": 51062 }, { "epoch": 0.99, "learning_rate": 3.348730987455441e-05, "loss": 0.0, "step": 51064 }, { "epoch": 0.99, "learning_rate": 3.348666291429718e-05, "loss": 0.0, "step": 51066 }, { "epoch": 0.99, "learning_rate": 3.3486015954039946e-05, "loss": 0.0, "step": 51068 }, { "epoch": 0.99, "learning_rate": 3.3485368993782715e-05, "loss": 0.0013, "step": 51070 }, { "epoch": 0.99, "learning_rate": 3.3484722033525484e-05, "loss": 0.0579, "step": 51072 }, { "epoch": 0.99, "learning_rate": 3.3484075073268253e-05, "loss": 0.0018, "step": 51074 }, { "epoch": 0.99, "learning_rate": 3.3483428113011016e-05, "loss": 0.0001, "step": 51076 }, { "epoch": 0.99, "learning_rate": 3.348278115275379e-05, "loss": 0.0056, "step": 51078 }, { "epoch": 0.99, "learning_rate": 3.3482134192496554e-05, "loss": 0.0, "step": 51080 }, { "epoch": 0.99, "learning_rate": 3.348148723223932e-05, "loss": 0.0026, "step": 51082 }, { "epoch": 0.99, "learning_rate": 3.348084027198209e-05, "loss": 0.0021, "step": 51084 }, { "epoch": 0.99, "learning_rate": 3.348019331172486e-05, "loss": 0.0001, "step": 51086 }, { "epoch": 0.99, "learning_rate": 3.347954635146763e-05, "loss": 0.0, "step": 51088 }, { "epoch": 0.99, "learning_rate": 3.34788993912104e-05, "loss": 0.0, "step": 51090 }, { "epoch": 0.99, "learning_rate": 3.347825243095317e-05, "loss": 0.0, "step": 51092 }, { "epoch": 0.99, "learning_rate": 3.347760547069594e-05, "loss": 0.0001, "step": 51094 }, { "epoch": 0.99, "learning_rate": 3.347695851043871e-05, "loss": 0.0001, "step": 51096 }, { "epoch": 0.99, "learning_rate": 3.347631155018147e-05, "loss": 0.0004, "step": 51098 }, { "epoch": 0.99, "learning_rate": 3.3475664589924245e-05, "loss": 0.0003, "step": 51100 }, { "epoch": 0.99, "learning_rate": 3.347501762966701e-05, "loss": 0.0001, "step": 51102 }, { "epoch": 0.99, "learning_rate": 3.347437066940978e-05, "loss": 0.0, "step": 51104 }, { "epoch": 0.99, "learning_rate": 3.347372370915255e-05, "loss": 0.0, "step": 51106 }, { "epoch": 0.99, "learning_rate": 3.3473076748895314e-05, "loss": 0.0001, "step": 51108 }, { "epoch": 0.99, "learning_rate": 3.347242978863809e-05, "loss": 0.0107, "step": 51110 }, { "epoch": 0.99, "learning_rate": 3.347178282838085e-05, "loss": 0.0, "step": 51112 }, { "epoch": 0.99, "learning_rate": 3.347113586812362e-05, "loss": 0.0, "step": 51114 }, { "epoch": 0.99, "learning_rate": 3.347048890786639e-05, "loss": 0.0, "step": 51116 }, { "epoch": 0.99, "learning_rate": 3.346984194760916e-05, "loss": 0.0001, "step": 51118 }, { "epoch": 0.99, "learning_rate": 3.346919498735193e-05, "loss": 0.0, "step": 51120 }, { "epoch": 0.99, "learning_rate": 3.34685480270947e-05, "loss": 0.0, "step": 51122 }, { "epoch": 0.99, "learning_rate": 3.346790106683747e-05, "loss": 0.0003, "step": 51124 }, { "epoch": 0.99, "learning_rate": 3.3467254106580236e-05, "loss": 0.0, "step": 51126 }, { "epoch": 0.99, "learning_rate": 3.3466607146323005e-05, "loss": 0.0, "step": 51128 }, { "epoch": 0.99, "learning_rate": 3.346596018606577e-05, "loss": 0.0, "step": 51130 }, { "epoch": 0.99, "learning_rate": 3.346531322580854e-05, "loss": 0.0014, "step": 51132 }, { "epoch": 0.99, "learning_rate": 3.3464666265551306e-05, "loss": 0.0, "step": 51134 }, { "epoch": 0.99, "learning_rate": 3.3464019305294075e-05, "loss": 0.0002, "step": 51136 }, { "epoch": 0.99, "learning_rate": 3.346337234503685e-05, "loss": 0.0005, "step": 51138 }, { "epoch": 0.99, "learning_rate": 3.346272538477961e-05, "loss": 0.0002, "step": 51140 }, { "epoch": 0.99, "learning_rate": 3.346207842452238e-05, "loss": 0.0, "step": 51142 }, { "epoch": 0.99, "learning_rate": 3.346143146426515e-05, "loss": 0.0001, "step": 51144 }, { "epoch": 0.99, "learning_rate": 3.346078450400792e-05, "loss": 0.0, "step": 51146 }, { "epoch": 0.99, "learning_rate": 3.346013754375069e-05, "loss": 0.0019, "step": 51148 }, { "epoch": 0.99, "learning_rate": 3.345949058349346e-05, "loss": 0.0, "step": 51150 }, { "epoch": 0.99, "learning_rate": 3.345884362323623e-05, "loss": 0.0859, "step": 51152 }, { "epoch": 0.99, "learning_rate": 3.3458196662978997e-05, "loss": 0.0, "step": 51154 }, { "epoch": 0.99, "learning_rate": 3.3457549702721766e-05, "loss": 0.0038, "step": 51156 }, { "epoch": 0.99, "learning_rate": 3.345690274246453e-05, "loss": 0.0001, "step": 51158 }, { "epoch": 0.99, "learning_rate": 3.3456255782207304e-05, "loss": 0.0, "step": 51160 }, { "epoch": 0.99, "learning_rate": 3.3455608821950066e-05, "loss": 0.0003, "step": 51162 }, { "epoch": 0.99, "learning_rate": 3.3454961861692835e-05, "loss": 0.0, "step": 51164 }, { "epoch": 0.99, "learning_rate": 3.3454314901435604e-05, "loss": 0.0038, "step": 51166 }, { "epoch": 0.99, "learning_rate": 3.345366794117837e-05, "loss": 0.0, "step": 51168 }, { "epoch": 0.99, "learning_rate": 3.345302098092115e-05, "loss": 0.0012, "step": 51170 }, { "epoch": 0.99, "learning_rate": 3.345237402066391e-05, "loss": 0.0, "step": 51172 }, { "epoch": 0.99, "learning_rate": 3.345172706040668e-05, "loss": 0.0, "step": 51174 }, { "epoch": 0.99, "learning_rate": 3.345108010014945e-05, "loss": 0.0, "step": 51176 }, { "epoch": 0.99, "learning_rate": 3.345043313989222e-05, "loss": 0.0, "step": 51178 }, { "epoch": 0.99, "learning_rate": 3.344978617963498e-05, "loss": 0.0, "step": 51180 }, { "epoch": 0.99, "learning_rate": 3.344913921937776e-05, "loss": 0.0, "step": 51182 }, { "epoch": 0.99, "learning_rate": 3.3448492259120526e-05, "loss": 0.0001, "step": 51184 }, { "epoch": 0.99, "learning_rate": 3.3447845298863295e-05, "loss": 0.0, "step": 51186 }, { "epoch": 0.99, "learning_rate": 3.3447198338606064e-05, "loss": 0.0, "step": 51188 }, { "epoch": 0.99, "learning_rate": 3.3446551378348826e-05, "loss": 0.0, "step": 51190 }, { "epoch": 0.99, "learning_rate": 3.34459044180916e-05, "loss": 0.0, "step": 51192 }, { "epoch": 0.99, "learning_rate": 3.3445257457834365e-05, "loss": 0.0065, "step": 51194 }, { "epoch": 0.99, "learning_rate": 3.3444610497577134e-05, "loss": 0.0, "step": 51196 }, { "epoch": 0.99, "learning_rate": 3.34439635373199e-05, "loss": 0.0, "step": 51198 }, { "epoch": 0.99, "learning_rate": 3.344331657706267e-05, "loss": 0.0, "step": 51200 }, { "epoch": 0.99, "learning_rate": 3.344266961680544e-05, "loss": 0.0, "step": 51202 }, { "epoch": 0.99, "learning_rate": 3.344202265654821e-05, "loss": 0.0, "step": 51204 }, { "epoch": 0.99, "learning_rate": 3.344137569629098e-05, "loss": 0.0005, "step": 51206 }, { "epoch": 0.99, "learning_rate": 3.344072873603375e-05, "loss": 0.0, "step": 51208 }, { "epoch": 0.99, "learning_rate": 3.344008177577652e-05, "loss": 0.0001, "step": 51210 }, { "epoch": 0.99, "learning_rate": 3.343943481551928e-05, "loss": 0.0, "step": 51212 }, { "epoch": 0.99, "learning_rate": 3.3438787855262055e-05, "loss": 0.0, "step": 51214 }, { "epoch": 0.99, "learning_rate": 3.3438140895004825e-05, "loss": 0.0013, "step": 51216 }, { "epoch": 0.99, "learning_rate": 3.343749393474759e-05, "loss": 0.0, "step": 51218 }, { "epoch": 0.99, "learning_rate": 3.343684697449036e-05, "loss": 0.0, "step": 51220 }, { "epoch": 0.99, "learning_rate": 3.3436200014233125e-05, "loss": 0.0, "step": 51222 }, { "epoch": 0.99, "learning_rate": 3.3435553053975894e-05, "loss": 0.0014, "step": 51224 }, { "epoch": 0.99, "learning_rate": 3.343490609371866e-05, "loss": 0.0, "step": 51226 }, { "epoch": 0.99, "learning_rate": 3.343425913346143e-05, "loss": 0.0, "step": 51228 }, { "epoch": 0.99, "learning_rate": 3.34336121732042e-05, "loss": 0.0, "step": 51230 }, { "epoch": 0.99, "learning_rate": 3.343296521294697e-05, "loss": 0.0, "step": 51232 }, { "epoch": 0.99, "learning_rate": 3.343231825268974e-05, "loss": 0.0, "step": 51234 }, { "epoch": 0.99, "learning_rate": 3.343167129243251e-05, "loss": 0.0001, "step": 51236 }, { "epoch": 0.99, "learning_rate": 3.343102433217528e-05, "loss": 0.0, "step": 51238 }, { "epoch": 0.99, "learning_rate": 3.343037737191804e-05, "loss": 0.0, "step": 51240 }, { "epoch": 0.99, "learning_rate": 3.3429730411660816e-05, "loss": 0.0, "step": 51242 }, { "epoch": 0.99, "learning_rate": 3.342908345140358e-05, "loss": 0.0, "step": 51244 }, { "epoch": 0.99, "learning_rate": 3.3428436491146354e-05, "loss": 0.0032, "step": 51246 }, { "epoch": 0.99, "learning_rate": 3.342778953088912e-05, "loss": 0.0001, "step": 51248 }, { "epoch": 0.99, "learning_rate": 3.3427142570631885e-05, "loss": 0.0, "step": 51250 }, { "epoch": 0.99, "learning_rate": 3.342649561037466e-05, "loss": 0.0223, "step": 51252 }, { "epoch": 0.99, "learning_rate": 3.3425848650117424e-05, "loss": 0.0, "step": 51254 }, { "epoch": 0.99, "learning_rate": 3.342520168986019e-05, "loss": 0.0, "step": 51256 }, { "epoch": 0.99, "learning_rate": 3.342455472960296e-05, "loss": 0.0606, "step": 51258 }, { "epoch": 0.99, "learning_rate": 3.342390776934573e-05, "loss": 0.0, "step": 51260 }, { "epoch": 0.99, "learning_rate": 3.34232608090885e-05, "loss": 0.0001, "step": 51262 }, { "epoch": 0.99, "learning_rate": 3.342261384883127e-05, "loss": 0.0065, "step": 51264 }, { "epoch": 1.0, "learning_rate": 3.342196688857404e-05, "loss": 0.0091, "step": 51266 }, { "epoch": 1.0, "learning_rate": 3.342131992831681e-05, "loss": 0.0, "step": 51268 }, { "epoch": 1.0, "learning_rate": 3.3420672968059576e-05, "loss": 0.0004, "step": 51270 }, { "epoch": 1.0, "learning_rate": 3.342002600780234e-05, "loss": 0.0, "step": 51272 }, { "epoch": 1.0, "learning_rate": 3.3419379047545114e-05, "loss": 0.0, "step": 51274 }, { "epoch": 1.0, "learning_rate": 3.341873208728788e-05, "loss": 0.0064, "step": 51276 }, { "epoch": 1.0, "learning_rate": 3.3418085127030646e-05, "loss": 0.0, "step": 51278 }, { "epoch": 1.0, "learning_rate": 3.3417438166773415e-05, "loss": 0.0, "step": 51280 }, { "epoch": 1.0, "learning_rate": 3.3416791206516184e-05, "loss": 0.0, "step": 51282 }, { "epoch": 1.0, "learning_rate": 3.341614424625895e-05, "loss": 0.0001, "step": 51284 }, { "epoch": 1.0, "learning_rate": 3.341549728600172e-05, "loss": 0.0, "step": 51286 }, { "epoch": 1.0, "learning_rate": 3.341485032574449e-05, "loss": 0.0, "step": 51288 }, { "epoch": 1.0, "learning_rate": 3.341420336548726e-05, "loss": 0.0, "step": 51290 }, { "epoch": 1.0, "learning_rate": 3.341355640523003e-05, "loss": 0.0, "step": 51292 }, { "epoch": 1.0, "learning_rate": 3.34129094449728e-05, "loss": 0.0002, "step": 51294 }, { "epoch": 1.0, "learning_rate": 3.341226248471557e-05, "loss": 0.0, "step": 51296 }, { "epoch": 1.0, "learning_rate": 3.341161552445834e-05, "loss": 0.0, "step": 51298 }, { "epoch": 1.0, "learning_rate": 3.34109685642011e-05, "loss": 0.0001, "step": 51300 }, { "epoch": 1.0, "learning_rate": 3.3410321603943875e-05, "loss": 0.0, "step": 51302 }, { "epoch": 1.0, "learning_rate": 3.340967464368664e-05, "loss": 0.0, "step": 51304 }, { "epoch": 1.0, "learning_rate": 3.3409027683429406e-05, "loss": 0.0, "step": 51306 }, { "epoch": 1.0, "learning_rate": 3.3408380723172175e-05, "loss": 0.003, "step": 51308 }, { "epoch": 1.0, "learning_rate": 3.3407733762914944e-05, "loss": 0.0041, "step": 51310 }, { "epoch": 1.0, "learning_rate": 3.3407086802657714e-05, "loss": 0.0003, "step": 51312 }, { "epoch": 1.0, "learning_rate": 3.340643984240048e-05, "loss": 0.0, "step": 51314 }, { "epoch": 1.0, "learning_rate": 3.340579288214325e-05, "loss": 0.0, "step": 51316 }, { "epoch": 1.0, "learning_rate": 3.340514592188602e-05, "loss": 0.0, "step": 51318 }, { "epoch": 1.0, "learning_rate": 3.340449896162879e-05, "loss": 0.0, "step": 51320 }, { "epoch": 1.0, "learning_rate": 3.340385200137155e-05, "loss": 0.0012, "step": 51322 }, { "epoch": 1.0, "learning_rate": 3.340320504111433e-05, "loss": 0.0012, "step": 51324 }, { "epoch": 1.0, "learning_rate": 3.340255808085709e-05, "loss": 0.0, "step": 51326 }, { "epoch": 1.0, "learning_rate": 3.3401911120599866e-05, "loss": 0.0008, "step": 51328 }, { "epoch": 1.0, "learning_rate": 3.3401264160342635e-05, "loss": 0.0, "step": 51330 }, { "epoch": 1.0, "learning_rate": 3.34006172000854e-05, "loss": 0.0, "step": 51332 }, { "epoch": 1.0, "learning_rate": 3.3399970239828173e-05, "loss": 0.0, "step": 51334 }, { "epoch": 1.0, "learning_rate": 3.3399323279570936e-05, "loss": 0.0, "step": 51336 }, { "epoch": 1.0, "learning_rate": 3.3398676319313705e-05, "loss": 0.0009, "step": 51338 }, { "epoch": 1.0, "learning_rate": 3.3398029359056474e-05, "loss": 0.0044, "step": 51340 }, { "epoch": 1.0, "learning_rate": 3.339738239879924e-05, "loss": 0.0, "step": 51342 }, { "epoch": 1.0, "learning_rate": 3.339673543854201e-05, "loss": 0.0, "step": 51344 }, { "epoch": 1.0, "learning_rate": 3.339608847828478e-05, "loss": 0.0001, "step": 51346 }, { "epoch": 1.0, "learning_rate": 3.339544151802755e-05, "loss": 0.0007, "step": 51348 }, { "epoch": 1.0, "learning_rate": 3.339479455777032e-05, "loss": 0.0001, "step": 51350 }, { "epoch": 1.0, "learning_rate": 3.339414759751309e-05, "loss": 0.0, "step": 51352 }, { "epoch": 1.0, "learning_rate": 3.339350063725585e-05, "loss": 0.0003, "step": 51354 }, { "epoch": 1.0, "learning_rate": 3.3392853676998627e-05, "loss": 0.0311, "step": 51356 }, { "epoch": 1.0, "learning_rate": 3.339220671674139e-05, "loss": 0.0004, "step": 51358 }, { "epoch": 1.0, "learning_rate": 3.339155975648416e-05, "loss": 0.0, "step": 51360 }, { "epoch": 1.0, "learning_rate": 3.3390912796226934e-05, "loss": 0.0, "step": 51362 }, { "epoch": 1.0, "learning_rate": 3.3390265835969696e-05, "loss": 0.0, "step": 51364 }, { "epoch": 1.0, "learning_rate": 3.3389618875712465e-05, "loss": 0.0, "step": 51366 }, { "epoch": 1.0, "learning_rate": 3.3388971915455234e-05, "loss": 0.0, "step": 51368 }, { "epoch": 1.0, "learning_rate": 3.3388324955198003e-05, "loss": 0.0, "step": 51370 }, { "epoch": 1.0, "learning_rate": 3.338767799494077e-05, "loss": 0.0, "step": 51372 }, { "epoch": 1.0, "learning_rate": 3.338703103468354e-05, "loss": 0.0001, "step": 51374 }, { "epoch": 1.0, "learning_rate": 3.338638407442631e-05, "loss": 0.0, "step": 51376 }, { "epoch": 1.0, "learning_rate": 3.338573711416908e-05, "loss": 0.0, "step": 51378 }, { "epoch": 1.0, "learning_rate": 3.338509015391185e-05, "loss": 0.0003, "step": 51380 }, { "epoch": 1.0, "learning_rate": 3.338444319365461e-05, "loss": 0.0, "step": 51382 }, { "epoch": 1.0, "learning_rate": 3.338379623339739e-05, "loss": 0.0, "step": 51384 }, { "epoch": 1.0, "learning_rate": 3.338314927314015e-05, "loss": 0.0, "step": 51386 }, { "epoch": 1.0, "learning_rate": 3.3382502312882925e-05, "loss": 0.0, "step": 51388 }, { "epoch": 1.0, "learning_rate": 3.338185535262569e-05, "loss": 0.0, "step": 51390 }, { "epoch": 1.0, "learning_rate": 3.3381208392368457e-05, "loss": 0.0001, "step": 51392 }, { "epoch": 1.0, "learning_rate": 3.338056143211123e-05, "loss": 0.0, "step": 51394 }, { "epoch": 1.0, "learning_rate": 3.3379914471853995e-05, "loss": 0.0, "step": 51396 }, { "epoch": 1.0, "learning_rate": 3.3379267511596764e-05, "loss": 0.0002, "step": 51398 }, { "epoch": 1.0, "learning_rate": 3.337862055133953e-05, "loss": 0.0025, "step": 51400 }, { "epoch": 1.0, "learning_rate": 3.33779735910823e-05, "loss": 0.0, "step": 51402 }, { "epoch": 1.0, "learning_rate": 3.3377326630825064e-05, "loss": 0.0, "step": 51404 }, { "epoch": 1.0, "learning_rate": 3.337667967056784e-05, "loss": 0.0, "step": 51406 }, { "epoch": 1.0, "learning_rate": 3.337603271031061e-05, "loss": 0.0, "step": 51408 }, { "epoch": 1.0, "learning_rate": 3.337538575005338e-05, "loss": 0.0001, "step": 51410 }, { "epoch": 1.0, "learning_rate": 3.337473878979615e-05, "loss": 0.0072, "step": 51412 }, { "epoch": 1.0, "learning_rate": 3.337409182953891e-05, "loss": 0.0, "step": 51414 }, { "epoch": 1.0, "learning_rate": 3.3373444869281686e-05, "loss": 0.0, "step": 51416 }, { "epoch": 1.0, "learning_rate": 3.337279790902445e-05, "loss": 0.0001, "step": 51418 }, { "epoch": 1.0, "learning_rate": 3.337215094876722e-05, "loss": 0.0003, "step": 51420 }, { "epoch": 1.0, "learning_rate": 3.3371503988509986e-05, "loss": 0.0016, "step": 51422 }, { "epoch": 1.0, "learning_rate": 3.3370857028252755e-05, "loss": 0.0, "step": 51424 }, { "epoch": 1.0, "learning_rate": 3.3370210067995524e-05, "loss": 0.0, "step": 51426 }, { "epoch": 1.0, "learning_rate": 3.336956310773829e-05, "loss": 0.0, "step": 51428 }, { "epoch": 1.0, "learning_rate": 3.336891614748106e-05, "loss": 0.0002, "step": 51430 }, { "epoch": 1.0, "learning_rate": 3.336826918722383e-05, "loss": 0.0, "step": 51432 }, { "epoch": 1.0, "learning_rate": 3.33676222269666e-05, "loss": 0.0, "step": 51434 }, { "epoch": 1.0, "learning_rate": 3.336697526670936e-05, "loss": 0.0155, "step": 51436 }, { "epoch": 1.0, "learning_rate": 3.336632830645214e-05, "loss": 0.0, "step": 51438 }, { "epoch": 1.0, "learning_rate": 3.336568134619491e-05, "loss": 0.0, "step": 51440 }, { "epoch": 1.0, "learning_rate": 3.336503438593767e-05, "loss": 0.0, "step": 51442 }, { "epoch": 1.0, "learning_rate": 3.3364387425680446e-05, "loss": 0.0, "step": 51444 }, { "epoch": 1.0, "learning_rate": 3.336374046542321e-05, "loss": 0.0, "step": 51446 }, { "epoch": 1.0, "learning_rate": 3.336309350516598e-05, "loss": 0.0173, "step": 51448 }, { "epoch": 1.0, "learning_rate": 3.3362446544908746e-05, "loss": 0.0024, "step": 51450 }, { "epoch": 1.0, "learning_rate": 3.3361799584651516e-05, "loss": 0.0, "step": 51452 }, { "epoch": 1.0, "learning_rate": 3.3361152624394285e-05, "loss": 0.0, "step": 51454 }, { "epoch": 1.0, "learning_rate": 3.3360505664137054e-05, "loss": 0.0, "step": 51456 }, { "epoch": 1.0, "learning_rate": 3.335985870387982e-05, "loss": 0.0002, "step": 51458 }, { "epoch": 1.0, "learning_rate": 3.335921174362259e-05, "loss": 0.0, "step": 51460 }, { "epoch": 1.0, "learning_rate": 3.335856478336536e-05, "loss": 0.0041, "step": 51462 }, { "epoch": 1.0, "learning_rate": 3.335791782310812e-05, "loss": 0.0, "step": 51464 }, { "epoch": 1.0, "learning_rate": 3.33572708628509e-05, "loss": 0.002, "step": 51466 }, { "epoch": 1.0, "learning_rate": 3.335662390259366e-05, "loss": 0.0, "step": 51468 }, { "epoch": 1.0, "learning_rate": 3.335597694233644e-05, "loss": 0.0002, "step": 51470 }, { "epoch": 1.0, "learning_rate": 3.3355329982079206e-05, "loss": 0.0, "step": 51472 }, { "epoch": 1.0, "learning_rate": 3.335468302182197e-05, "loss": 0.0, "step": 51474 }, { "epoch": 1.0, "learning_rate": 3.3354036061564745e-05, "loss": 0.0011, "step": 51476 }, { "epoch": 1.0, "learning_rate": 3.335338910130751e-05, "loss": 0.0, "step": 51478 }, { "epoch": 1.0, "learning_rate": 3.3352742141050276e-05, "loss": 0.1131, "step": 51480 }, { "epoch": 1.0, "learning_rate": 3.3352095180793045e-05, "loss": 0.0001, "step": 51482 }, { "epoch": 1.0, "learning_rate": 3.3351448220535814e-05, "loss": 0.0, "step": 51484 }, { "epoch": 1.0, "learning_rate": 3.335080126027858e-05, "loss": 0.0, "step": 51486 }, { "epoch": 1.0, "learning_rate": 3.335015430002135e-05, "loss": 0.0001, "step": 51488 }, { "epoch": 1.0, "learning_rate": 3.334950733976412e-05, "loss": 0.0, "step": 51490 }, { "epoch": 1.0, "learning_rate": 3.334886037950689e-05, "loss": 0.0195, "step": 51492 }, { "epoch": 1.0, "learning_rate": 3.334821341924966e-05, "loss": 0.0001, "step": 51494 }, { "epoch": 1.0, "learning_rate": 3.334756645899242e-05, "loss": 0.0, "step": 51496 }, { "epoch": 1.0, "learning_rate": 3.33469194987352e-05, "loss": 0.0, "step": 51498 }, { "epoch": 1.0, "learning_rate": 3.334627253847796e-05, "loss": 0.0063, "step": 51500 }, { "epoch": 1.0, "learning_rate": 3.334562557822073e-05, "loss": 0.0001, "step": 51502 }, { "epoch": 1.0, "learning_rate": 3.33449786179635e-05, "loss": 0.0002, "step": 51504 }, { "epoch": 1.0, "learning_rate": 3.334433165770627e-05, "loss": 0.0256, "step": 51506 }, { "epoch": 1.0, "learning_rate": 3.3343684697449036e-05, "loss": 0.0, "step": 51508 }, { "epoch": 1.0, "learning_rate": 3.3343037737191805e-05, "loss": 0.0404, "step": 51510 }, { "epoch": 1.0, "learning_rate": 3.3342390776934575e-05, "loss": 0.0003, "step": 51512 }, { "epoch": 1.0, "learning_rate": 3.3341743816677344e-05, "loss": 0.0, "step": 51514 }, { "epoch": 1.0, "learning_rate": 3.334109685642011e-05, "loss": 0.0001, "step": 51516 }, { "epoch": 1.0, "learning_rate": 3.334044989616288e-05, "loss": 0.0, "step": 51518 }, { "epoch": 1.0, "learning_rate": 3.333980293590565e-05, "loss": 0.0, "step": 51520 }, { "epoch": 1.0, "learning_rate": 3.333915597564842e-05, "loss": 0.0001, "step": 51522 }, { "epoch": 1.0, "learning_rate": 3.333850901539118e-05, "loss": 0.0, "step": 51524 }, { "epoch": 1.0, "learning_rate": 3.333786205513396e-05, "loss": 0.0002, "step": 51526 }, { "epoch": 1.0, "learning_rate": 3.333721509487672e-05, "loss": 0.0, "step": 51528 }, { "epoch": 1.0, "learning_rate": 3.3336568134619496e-05, "loss": 0.0018, "step": 51530 }, { "epoch": 1.0, "learning_rate": 3.333592117436226e-05, "loss": 0.0, "step": 51532 }, { "epoch": 1.0, "learning_rate": 3.333527421410503e-05, "loss": 0.0, "step": 51534 }, { "epoch": 1.0, "learning_rate": 3.33346272538478e-05, "loss": 0.0, "step": 51536 }, { "epoch": 1.0, "learning_rate": 3.3333980293590566e-05, "loss": 0.0, "step": 51538 }, { "epoch": 1.0, "learning_rate": 3.3333333333333335e-05, "loss": 0.0007, "step": 51540 }, { "epoch": 1.0, "learning_rate": 3.3332686373076104e-05, "loss": 0.0, "step": 51542 }, { "epoch": 1.0, "learning_rate": 3.333203941281887e-05, "loss": 0.0, "step": 51544 }, { "epoch": 1.0, "learning_rate": 3.3331392452561635e-05, "loss": 0.0, "step": 51546 }, { "epoch": 1.0, "learning_rate": 3.333074549230441e-05, "loss": 0.0008, "step": 51548 }, { "epoch": 1.0, "learning_rate": 3.333009853204718e-05, "loss": 0.0034, "step": 51550 }, { "epoch": 1.0, "learning_rate": 3.332945157178995e-05, "loss": 0.0001, "step": 51552 }, { "epoch": 1.0, "learning_rate": 3.332880461153272e-05, "loss": 0.0001, "step": 51554 }, { "epoch": 1.0, "learning_rate": 3.332815765127548e-05, "loss": 0.0, "step": 51556 }, { "epoch": 1.0, "learning_rate": 3.332751069101826e-05, "loss": 0.001, "step": 51558 }, { "epoch": 1.0, "learning_rate": 3.332686373076102e-05, "loss": 0.0, "step": 51560 }, { "epoch": 1.0, "learning_rate": 3.332621677050379e-05, "loss": 0.0, "step": 51562 }, { "epoch": 1.0, "learning_rate": 3.332556981024656e-05, "loss": 0.0, "step": 51564 }, { "epoch": 1.0, "learning_rate": 3.3324922849989326e-05, "loss": 0.0015, "step": 51566 }, { "epoch": 1.0, "learning_rate": 3.3324275889732095e-05, "loss": 0.0, "step": 51568 }, { "epoch": 1.0, "learning_rate": 3.3323628929474864e-05, "loss": 0.0, "step": 51570 }, { "epoch": 1.0, "learning_rate": 3.3322981969217634e-05, "loss": 0.0, "step": 51572 }, { "epoch": 1.0, "learning_rate": 3.33223350089604e-05, "loss": 0.0, "step": 51574 }, { "epoch": 1.0, "learning_rate": 3.332168804870317e-05, "loss": 0.0054, "step": 51576 }, { "epoch": 1.0, "learning_rate": 3.3321041088445934e-05, "loss": 0.0, "step": 51578 }, { "epoch": 1.0, "learning_rate": 3.332039412818871e-05, "loss": 0.0004, "step": 51580 }, { "epoch": 1.0, "learning_rate": 3.331974716793147e-05, "loss": 0.0001, "step": 51582 }, { "epoch": 1.0, "learning_rate": 3.331910020767424e-05, "loss": 0.0, "step": 51584 }, { "epoch": 1.0, "learning_rate": 3.331845324741702e-05, "loss": 0.0, "step": 51586 }, { "epoch": 1.0, "learning_rate": 3.331780628715978e-05, "loss": 0.0006, "step": 51588 }, { "epoch": 1.0, "learning_rate": 3.331715932690255e-05, "loss": 0.0, "step": 51590 }, { "epoch": 1.0, "learning_rate": 3.331651236664532e-05, "loss": 0.0, "step": 51592 }, { "epoch": 1.0, "learning_rate": 3.331586540638809e-05, "loss": 0.0, "step": 51594 }, { "epoch": 1.0, "learning_rate": 3.3315218446130856e-05, "loss": 0.0, "step": 51596 }, { "epoch": 1.0, "learning_rate": 3.3314571485873625e-05, "loss": 0.0, "step": 51598 }, { "epoch": 1.0, "learning_rate": 3.3313924525616394e-05, "loss": 0.0, "step": 51600 }, { "epoch": 1.0, "learning_rate": 3.331327756535916e-05, "loss": 0.0052, "step": 51602 }, { "epoch": 1.0, "learning_rate": 3.331263060510193e-05, "loss": 0.0, "step": 51604 }, { "epoch": 1.0, "learning_rate": 3.3311983644844694e-05, "loss": 0.0, "step": 51606 }, { "epoch": 1.0, "learning_rate": 3.331133668458747e-05, "loss": 0.0, "step": 51608 }, { "epoch": 1.0, "learning_rate": 3.331068972433023e-05, "loss": 0.0, "step": 51610 }, { "epoch": 1.0, "learning_rate": 3.331004276407301e-05, "loss": 0.0, "step": 51612 }, { "epoch": 1.0, "learning_rate": 3.330939580381577e-05, "loss": 0.0009, "step": 51614 }, { "epoch": 1.0, "learning_rate": 3.330874884355854e-05, "loss": 0.0, "step": 51616 }, { "epoch": 1.0, "learning_rate": 3.3308101883301316e-05, "loss": 0.0, "step": 51618 }, { "epoch": 1.0, "learning_rate": 3.330745492304408e-05, "loss": 0.0001, "step": 51620 }, { "epoch": 1.0, "learning_rate": 3.330680796278685e-05, "loss": 0.0, "step": 51622 }, { "epoch": 1.0, "learning_rate": 3.3306161002529616e-05, "loss": 0.0, "step": 51624 }, { "epoch": 1.0, "learning_rate": 3.3305514042272385e-05, "loss": 0.0, "step": 51626 }, { "epoch": 1.0, "learning_rate": 3.3304867082015154e-05, "loss": 0.0027, "step": 51628 }, { "epoch": 1.0, "learning_rate": 3.3304220121757923e-05, "loss": 0.0, "step": 51630 }, { "epoch": 1.0, "learning_rate": 3.330357316150069e-05, "loss": 0.0, "step": 51632 }, { "epoch": 1.0, "learning_rate": 3.330292620124346e-05, "loss": 0.0, "step": 51634 }, { "epoch": 1.0, "learning_rate": 3.330227924098623e-05, "loss": 0.0, "step": 51636 }, { "epoch": 1.0, "learning_rate": 3.330163228072899e-05, "loss": 0.0, "step": 51638 }, { "epoch": 1.0, "learning_rate": 3.330098532047177e-05, "loss": 0.0, "step": 51640 }, { "epoch": 1.0, "learning_rate": 3.330033836021453e-05, "loss": 0.0, "step": 51642 }, { "epoch": 1.0, "learning_rate": 3.32996913999573e-05, "loss": 0.0, "step": 51644 }, { "epoch": 1.0, "learning_rate": 3.329904443970007e-05, "loss": 0.0, "step": 51646 }, { "epoch": 1.0, "learning_rate": 3.329839747944284e-05, "loss": 0.0, "step": 51648 }, { "epoch": 1.0, "learning_rate": 3.329775051918561e-05, "loss": 0.0, "step": 51650 }, { "epoch": 1.0, "learning_rate": 3.3297103558928377e-05, "loss": 0.0, "step": 51652 }, { "epoch": 1.0, "learning_rate": 3.3296456598671146e-05, "loss": 0.0003, "step": 51654 }, { "epoch": 1.0, "learning_rate": 3.3295809638413915e-05, "loss": 0.0002, "step": 51656 }, { "epoch": 1.0, "learning_rate": 3.3295162678156684e-05, "loss": 0.0, "step": 51658 }, { "epoch": 1.0, "learning_rate": 3.3294515717899446e-05, "loss": 0.0, "step": 51660 }, { "epoch": 1.0, "learning_rate": 3.329386875764222e-05, "loss": 0.0037, "step": 51662 }, { "epoch": 1.0, "learning_rate": 3.329322179738499e-05, "loss": 0.0002, "step": 51664 }, { "epoch": 1.0, "learning_rate": 3.329257483712775e-05, "loss": 0.0, "step": 51666 }, { "epoch": 1.0, "learning_rate": 3.329192787687053e-05, "loss": 0.0002, "step": 51668 }, { "epoch": 1.0, "learning_rate": 3.329128091661329e-05, "loss": 0.0, "step": 51670 }, { "epoch": 1.0, "learning_rate": 3.329063395635607e-05, "loss": 0.0001, "step": 51672 }, { "epoch": 1.0, "learning_rate": 3.328998699609883e-05, "loss": 0.0, "step": 51674 }, { "epoch": 1.0, "learning_rate": 3.32893400358416e-05, "loss": 0.0, "step": 51676 }, { "epoch": 1.0, "learning_rate": 3.328869307558437e-05, "loss": 0.0, "step": 51678 }, { "epoch": 1.0, "learning_rate": 3.328804611532714e-05, "loss": 0.0, "step": 51680 }, { "epoch": 1.0, "learning_rate": 3.3287399155069906e-05, "loss": 0.0, "step": 51682 }, { "epoch": 1.0, "learning_rate": 3.3286752194812675e-05, "loss": 0.0, "step": 51684 }, { "epoch": 1.0, "learning_rate": 3.3286105234555444e-05, "loss": 0.0, "step": 51686 }, { "epoch": 1.0, "learning_rate": 3.3285458274298207e-05, "loss": 0.0, "step": 51688 }, { "epoch": 1.0, "learning_rate": 3.328481131404098e-05, "loss": 0.0002, "step": 51690 }, { "epoch": 1.0, "learning_rate": 3.3284164353783745e-05, "loss": 0.0, "step": 51692 }, { "epoch": 1.0, "learning_rate": 3.328351739352652e-05, "loss": 0.0, "step": 51694 }, { "epoch": 1.0, "learning_rate": 3.328287043326929e-05, "loss": 0.0, "step": 51696 }, { "epoch": 1.0, "learning_rate": 3.328222347301205e-05, "loss": 0.0, "step": 51698 }, { "epoch": 1.0, "learning_rate": 3.328157651275483e-05, "loss": 0.0, "step": 51700 }, { "epoch": 1.0, "learning_rate": 3.328092955249759e-05, "loss": 0.0002, "step": 51702 }, { "epoch": 1.0, "learning_rate": 3.328028259224036e-05, "loss": 0.0137, "step": 51704 }, { "epoch": 1.0, "learning_rate": 3.327963563198313e-05, "loss": 0.0003, "step": 51706 }, { "epoch": 1.0, "learning_rate": 3.32789886717259e-05, "loss": 0.0001, "step": 51708 }, { "epoch": 1.0, "learning_rate": 3.3278341711468666e-05, "loss": 0.0342, "step": 51710 }, { "epoch": 1.0, "learning_rate": 3.3277694751211436e-05, "loss": 0.0, "step": 51712 }, { "epoch": 1.0, "learning_rate": 3.3277047790954205e-05, "loss": 0.0004, "step": 51714 }, { "epoch": 1.0, "learning_rate": 3.3276400830696974e-05, "loss": 0.0, "step": 51716 }, { "epoch": 1.0, "learning_rate": 3.327575387043974e-05, "loss": 0.0, "step": 51718 }, { "epoch": 1.0, "learning_rate": 3.3275106910182505e-05, "loss": 0.0013, "step": 51720 }, { "epoch": 1.0, "learning_rate": 3.327445994992528e-05, "loss": 0.0002, "step": 51722 }, { "epoch": 1.0, "learning_rate": 3.327381298966804e-05, "loss": 0.0, "step": 51724 }, { "epoch": 1.0, "learning_rate": 3.327316602941081e-05, "loss": 0.0, "step": 51726 }, { "epoch": 1.0, "learning_rate": 3.327251906915359e-05, "loss": 0.0, "step": 51728 }, { "epoch": 1.0, "learning_rate": 3.327187210889635e-05, "loss": 0.0, "step": 51730 }, { "epoch": 1.0, "learning_rate": 3.327122514863912e-05, "loss": 0.0, "step": 51732 }, { "epoch": 1.0, "learning_rate": 3.327057818838189e-05, "loss": 0.0023, "step": 51734 }, { "epoch": 1.0, "learning_rate": 3.326993122812466e-05, "loss": 0.0006, "step": 51736 }, { "epoch": 1.0, "learning_rate": 3.326928426786743e-05, "loss": 0.0, "step": 51738 }, { "epoch": 1.0, "learning_rate": 3.3268637307610196e-05, "loss": 0.0, "step": 51740 }, { "epoch": 1.0, "learning_rate": 3.3267990347352965e-05, "loss": 0.0001, "step": 51742 }, { "epoch": 1.0, "learning_rate": 3.3267343387095734e-05, "loss": 0.0, "step": 51744 }, { "epoch": 1.0, "learning_rate": 3.32666964268385e-05, "loss": 0.0, "step": 51746 }, { "epoch": 1.0, "learning_rate": 3.3266049466581265e-05, "loss": 0.0001, "step": 51748 }, { "epoch": 1.0, "learning_rate": 3.326540250632404e-05, "loss": 0.0, "step": 51750 }, { "epoch": 1.0, "learning_rate": 3.3264755546066804e-05, "loss": 0.001, "step": 51752 }, { "epoch": 1.0, "learning_rate": 3.326410858580958e-05, "loss": 0.0007, "step": 51754 }, { "epoch": 1.0, "learning_rate": 3.326346162555234e-05, "loss": 0.0, "step": 51756 }, { "epoch": 1.0, "learning_rate": 3.326281466529511e-05, "loss": 0.0054, "step": 51758 }, { "epoch": 1.0, "learning_rate": 3.326216770503788e-05, "loss": 0.0098, "step": 51760 }, { "epoch": 1.0, "learning_rate": 3.326152074478065e-05, "loss": 0.0, "step": 51762 }, { "epoch": 1.0, "learning_rate": 3.326087378452342e-05, "loss": 0.0, "step": 51764 }, { "epoch": 1.0, "learning_rate": 3.326022682426619e-05, "loss": 0.0, "step": 51766 }, { "epoch": 1.0, "learning_rate": 3.3259579864008956e-05, "loss": 0.0, "step": 51768 }, { "epoch": 1.0, "learning_rate": 3.325893290375172e-05, "loss": 0.0, "step": 51770 }, { "epoch": 1.0, "learning_rate": 3.3258285943494495e-05, "loss": 0.0, "step": 51772 }, { "epoch": 1.0, "learning_rate": 3.3257638983237264e-05, "loss": 0.0, "step": 51774 }, { "epoch": 1.0, "learning_rate": 3.325699202298003e-05, "loss": 0.0, "step": 51776 }, { "epoch": 1.0, "learning_rate": 3.32563450627228e-05, "loss": 0.0005, "step": 51778 }, { "epoch": 1.0, "learning_rate": 3.3255698102465564e-05, "loss": 0.0023, "step": 51780 }, { "epoch": 1.01, "learning_rate": 3.325505114220834e-05, "loss": 0.0, "step": 51782 }, { "epoch": 1.01, "learning_rate": 3.32544041819511e-05, "loss": 0.0, "step": 51784 }, { "epoch": 1.01, "learning_rate": 3.325375722169387e-05, "loss": 0.0, "step": 51786 }, { "epoch": 1.01, "learning_rate": 3.325311026143664e-05, "loss": 0.0, "step": 51788 }, { "epoch": 1.01, "learning_rate": 3.325246330117941e-05, "loss": 0.0004, "step": 51790 }, { "epoch": 1.01, "learning_rate": 3.325181634092218e-05, "loss": 0.0, "step": 51792 }, { "epoch": 1.01, "learning_rate": 3.325116938066495e-05, "loss": 0.0, "step": 51794 }, { "epoch": 1.01, "learning_rate": 3.325052242040772e-05, "loss": 0.0, "step": 51796 }, { "epoch": 1.01, "learning_rate": 3.3249875460150486e-05, "loss": 0.0, "step": 51798 }, { "epoch": 1.01, "learning_rate": 3.3249228499893255e-05, "loss": 0.0316, "step": 51800 }, { "epoch": 1.01, "learning_rate": 3.324858153963602e-05, "loss": 0.0, "step": 51802 }, { "epoch": 1.01, "learning_rate": 3.324793457937879e-05, "loss": 0.0001, "step": 51804 }, { "epoch": 1.01, "learning_rate": 3.3247287619121555e-05, "loss": 0.0, "step": 51806 }, { "epoch": 1.01, "learning_rate": 3.3246640658864324e-05, "loss": 0.0, "step": 51808 }, { "epoch": 1.01, "learning_rate": 3.32459936986071e-05, "loss": 0.0, "step": 51810 }, { "epoch": 1.01, "learning_rate": 3.324534673834986e-05, "loss": 0.0, "step": 51812 }, { "epoch": 1.01, "learning_rate": 3.324469977809263e-05, "loss": 0.0001, "step": 51814 }, { "epoch": 1.01, "learning_rate": 3.32440528178354e-05, "loss": 0.0011, "step": 51816 }, { "epoch": 1.01, "learning_rate": 3.324340585757817e-05, "loss": 0.0005, "step": 51818 }, { "epoch": 1.01, "learning_rate": 3.324275889732094e-05, "loss": 0.0, "step": 51820 }, { "epoch": 1.01, "learning_rate": 3.324211193706371e-05, "loss": 0.0004, "step": 51822 }, { "epoch": 1.01, "learning_rate": 3.324146497680648e-05, "loss": 0.0, "step": 51824 }, { "epoch": 1.01, "learning_rate": 3.3240818016549246e-05, "loss": 0.0045, "step": 51826 }, { "epoch": 1.01, "learning_rate": 3.3240171056292015e-05, "loss": 0.0, "step": 51828 }, { "epoch": 1.01, "learning_rate": 3.323952409603478e-05, "loss": 0.0001, "step": 51830 }, { "epoch": 1.01, "learning_rate": 3.3238877135777553e-05, "loss": 0.0096, "step": 51832 }, { "epoch": 1.01, "learning_rate": 3.3238230175520316e-05, "loss": 0.0, "step": 51834 }, { "epoch": 1.01, "learning_rate": 3.323758321526309e-05, "loss": 0.0, "step": 51836 }, { "epoch": 1.01, "learning_rate": 3.3236936255005854e-05, "loss": 0.0, "step": 51838 }, { "epoch": 1.01, "learning_rate": 3.323628929474862e-05, "loss": 0.0, "step": 51840 }, { "epoch": 1.01, "learning_rate": 3.32356423344914e-05, "loss": 0.0, "step": 51842 }, { "epoch": 1.01, "learning_rate": 3.323499537423416e-05, "loss": 0.0, "step": 51844 }, { "epoch": 1.01, "learning_rate": 3.323434841397693e-05, "loss": 0.0047, "step": 51846 }, { "epoch": 1.01, "learning_rate": 3.32337014537197e-05, "loss": 0.0001, "step": 51848 }, { "epoch": 1.01, "learning_rate": 3.323305449346247e-05, "loss": 0.0001, "step": 51850 }, { "epoch": 1.01, "learning_rate": 3.323240753320524e-05, "loss": 0.0, "step": 51852 }, { "epoch": 1.01, "learning_rate": 3.323176057294801e-05, "loss": 0.0, "step": 51854 }, { "epoch": 1.01, "learning_rate": 3.3231113612690776e-05, "loss": 0.0, "step": 51856 }, { "epoch": 1.01, "learning_rate": 3.3230466652433545e-05, "loss": 0.0, "step": 51858 }, { "epoch": 1.01, "learning_rate": 3.3229819692176314e-05, "loss": 0.0, "step": 51860 }, { "epoch": 1.01, "learning_rate": 3.3229172731919076e-05, "loss": 0.0001, "step": 51862 }, { "epoch": 1.01, "learning_rate": 3.322852577166185e-05, "loss": 0.0, "step": 51864 }, { "epoch": 1.01, "learning_rate": 3.3227878811404614e-05, "loss": 0.0, "step": 51866 }, { "epoch": 1.01, "learning_rate": 3.3227231851147383e-05, "loss": 0.0, "step": 51868 }, { "epoch": 1.01, "learning_rate": 3.322658489089015e-05, "loss": 0.0001, "step": 51870 }, { "epoch": 1.01, "learning_rate": 3.322593793063292e-05, "loss": 0.0001, "step": 51872 }, { "epoch": 1.01, "learning_rate": 3.322529097037569e-05, "loss": 0.0, "step": 51874 }, { "epoch": 1.01, "learning_rate": 3.322464401011846e-05, "loss": 0.0001, "step": 51876 }, { "epoch": 1.01, "learning_rate": 3.322399704986123e-05, "loss": 0.0039, "step": 51878 }, { "epoch": 1.01, "learning_rate": 3.3223350089604e-05, "loss": 0.0001, "step": 51880 }, { "epoch": 1.01, "learning_rate": 3.322270312934677e-05, "loss": 0.0, "step": 51882 }, { "epoch": 1.01, "learning_rate": 3.322205616908953e-05, "loss": 0.0001, "step": 51884 }, { "epoch": 1.01, "learning_rate": 3.3221409208832305e-05, "loss": 0.0, "step": 51886 }, { "epoch": 1.01, "learning_rate": 3.3220762248575074e-05, "loss": 0.0, "step": 51888 }, { "epoch": 1.01, "learning_rate": 3.3220115288317837e-05, "loss": 0.0, "step": 51890 }, { "epoch": 1.01, "learning_rate": 3.321946832806061e-05, "loss": 0.0001, "step": 51892 }, { "epoch": 1.01, "learning_rate": 3.3218821367803375e-05, "loss": 0.0, "step": 51894 }, { "epoch": 1.01, "learning_rate": 3.321817440754615e-05, "loss": 0.0, "step": 51896 }, { "epoch": 1.01, "learning_rate": 3.321752744728891e-05, "loss": 0.0, "step": 51898 }, { "epoch": 1.01, "learning_rate": 3.321688048703168e-05, "loss": 0.0425, "step": 51900 }, { "epoch": 1.01, "learning_rate": 3.321623352677445e-05, "loss": 0.0, "step": 51902 }, { "epoch": 1.01, "learning_rate": 3.321558656651722e-05, "loss": 0.0001, "step": 51904 }, { "epoch": 1.01, "learning_rate": 3.321493960625999e-05, "loss": 0.0001, "step": 51906 }, { "epoch": 1.01, "learning_rate": 3.321429264600276e-05, "loss": 0.0, "step": 51908 }, { "epoch": 1.01, "learning_rate": 3.321364568574553e-05, "loss": 0.0001, "step": 51910 }, { "epoch": 1.01, "learning_rate": 3.321299872548829e-05, "loss": 0.0, "step": 51912 }, { "epoch": 1.01, "learning_rate": 3.3212351765231066e-05, "loss": 0.0, "step": 51914 }, { "epoch": 1.01, "learning_rate": 3.321170480497383e-05, "loss": 0.0, "step": 51916 }, { "epoch": 1.01, "learning_rate": 3.3211057844716604e-05, "loss": 0.0, "step": 51918 }, { "epoch": 1.01, "learning_rate": 3.321041088445937e-05, "loss": 0.0, "step": 51920 }, { "epoch": 1.01, "learning_rate": 3.3209763924202135e-05, "loss": 0.0024, "step": 51922 }, { "epoch": 1.01, "learning_rate": 3.320911696394491e-05, "loss": 0.0, "step": 51924 }, { "epoch": 1.01, "learning_rate": 3.320847000368767e-05, "loss": 0.0, "step": 51926 }, { "epoch": 1.01, "learning_rate": 3.320782304343044e-05, "loss": 0.0, "step": 51928 }, { "epoch": 1.01, "learning_rate": 3.320717608317321e-05, "loss": 0.0, "step": 51930 }, { "epoch": 1.01, "learning_rate": 3.320652912291598e-05, "loss": 0.0001, "step": 51932 }, { "epoch": 1.01, "learning_rate": 3.320588216265875e-05, "loss": 0.0, "step": 51934 }, { "epoch": 1.01, "learning_rate": 3.320523520240152e-05, "loss": 0.0, "step": 51936 }, { "epoch": 1.01, "learning_rate": 3.320458824214429e-05, "loss": 0.0, "step": 51938 }, { "epoch": 1.01, "learning_rate": 3.320394128188706e-05, "loss": 0.0, "step": 51940 }, { "epoch": 1.01, "learning_rate": 3.3203294321629826e-05, "loss": 0.0, "step": 51942 }, { "epoch": 1.01, "learning_rate": 3.320264736137259e-05, "loss": 0.0, "step": 51944 }, { "epoch": 1.01, "learning_rate": 3.3202000401115364e-05, "loss": 0.0001, "step": 51946 }, { "epoch": 1.01, "learning_rate": 3.3201353440858126e-05, "loss": 0.0, "step": 51948 }, { "epoch": 1.01, "learning_rate": 3.3200706480600896e-05, "loss": 0.0002, "step": 51950 }, { "epoch": 1.01, "learning_rate": 3.320005952034367e-05, "loss": 0.0001, "step": 51952 }, { "epoch": 1.01, "learning_rate": 3.3199412560086434e-05, "loss": 0.0037, "step": 51954 }, { "epoch": 1.01, "learning_rate": 3.31987655998292e-05, "loss": 0.001, "step": 51956 }, { "epoch": 1.01, "learning_rate": 3.319811863957197e-05, "loss": 0.0006, "step": 51958 }, { "epoch": 1.01, "learning_rate": 3.319747167931474e-05, "loss": 0.0035, "step": 51960 }, { "epoch": 1.01, "learning_rate": 3.319682471905751e-05, "loss": 0.0, "step": 51962 }, { "epoch": 1.01, "learning_rate": 3.319617775880028e-05, "loss": 0.0, "step": 51964 }, { "epoch": 1.01, "learning_rate": 3.319553079854305e-05, "loss": 0.0003, "step": 51966 }, { "epoch": 1.01, "learning_rate": 3.319488383828582e-05, "loss": 0.0, "step": 51968 }, { "epoch": 1.01, "learning_rate": 3.3194236878028586e-05, "loss": 0.0001, "step": 51970 }, { "epoch": 1.01, "learning_rate": 3.319358991777135e-05, "loss": 0.0, "step": 51972 }, { "epoch": 1.01, "learning_rate": 3.3192942957514125e-05, "loss": 0.0, "step": 51974 }, { "epoch": 1.01, "learning_rate": 3.319229599725689e-05, "loss": 0.0, "step": 51976 }, { "epoch": 1.01, "learning_rate": 3.319164903699966e-05, "loss": 0.0005, "step": 51978 }, { "epoch": 1.01, "learning_rate": 3.3191002076742425e-05, "loss": 0.0, "step": 51980 }, { "epoch": 1.01, "learning_rate": 3.3190355116485194e-05, "loss": 0.0, "step": 51982 }, { "epoch": 1.01, "learning_rate": 3.318970815622796e-05, "loss": 0.0, "step": 51984 }, { "epoch": 1.01, "learning_rate": 3.318906119597073e-05, "loss": 0.0, "step": 51986 }, { "epoch": 1.01, "learning_rate": 3.31884142357135e-05, "loss": 0.0, "step": 51988 }, { "epoch": 1.01, "learning_rate": 3.318776727545627e-05, "loss": 0.0001, "step": 51990 }, { "epoch": 1.01, "learning_rate": 3.318712031519904e-05, "loss": 0.0, "step": 51992 }, { "epoch": 1.01, "learning_rate": 3.31864733549418e-05, "loss": 0.0004, "step": 51994 }, { "epoch": 1.01, "learning_rate": 3.318582639468458e-05, "loss": 0.0, "step": 51996 }, { "epoch": 1.01, "learning_rate": 3.318517943442735e-05, "loss": 0.0, "step": 51998 }, { "epoch": 1.01, "learning_rate": 3.3184532474170116e-05, "loss": 0.0, "step": 52000 }, { "epoch": 1.01, "learning_rate": 3.3183885513912885e-05, "loss": 0.0107, "step": 52002 }, { "epoch": 1.01, "learning_rate": 3.318323855365565e-05, "loss": 0.0001, "step": 52004 }, { "epoch": 1.01, "learning_rate": 3.318259159339842e-05, "loss": 0.0, "step": 52006 }, { "epoch": 1.01, "learning_rate": 3.3181944633141185e-05, "loss": 0.0, "step": 52008 }, { "epoch": 1.01, "learning_rate": 3.3181297672883955e-05, "loss": 0.0001, "step": 52010 }, { "epoch": 1.01, "learning_rate": 3.3180650712626724e-05, "loss": 0.0, "step": 52012 }, { "epoch": 1.01, "learning_rate": 3.318000375236949e-05, "loss": 0.0057, "step": 52014 }, { "epoch": 1.01, "learning_rate": 3.317935679211226e-05, "loss": 0.0003, "step": 52016 }, { "epoch": 1.01, "learning_rate": 3.317870983185503e-05, "loss": 0.0, "step": 52018 }, { "epoch": 1.01, "learning_rate": 3.31780628715978e-05, "loss": 0.0, "step": 52020 }, { "epoch": 1.01, "learning_rate": 3.317741591134057e-05, "loss": 0.0, "step": 52022 }, { "epoch": 1.01, "learning_rate": 3.317676895108334e-05, "loss": 0.0, "step": 52024 }, { "epoch": 1.01, "learning_rate": 3.31761219908261e-05, "loss": 0.0, "step": 52026 }, { "epoch": 1.01, "learning_rate": 3.3175475030568876e-05, "loss": 0.0002, "step": 52028 }, { "epoch": 1.01, "learning_rate": 3.3174828070311645e-05, "loss": 0.0, "step": 52030 }, { "epoch": 1.01, "learning_rate": 3.317418111005441e-05, "loss": 0.0, "step": 52032 }, { "epoch": 1.01, "learning_rate": 3.3173534149797184e-05, "loss": 0.0001, "step": 52034 }, { "epoch": 1.01, "learning_rate": 3.3172887189539946e-05, "loss": 0.0, "step": 52036 }, { "epoch": 1.01, "learning_rate": 3.317224022928272e-05, "loss": 0.0, "step": 52038 }, { "epoch": 1.01, "learning_rate": 3.3171593269025484e-05, "loss": 0.0001, "step": 52040 }, { "epoch": 1.01, "learning_rate": 3.317094630876825e-05, "loss": 0.0017, "step": 52042 }, { "epoch": 1.01, "learning_rate": 3.317029934851102e-05, "loss": 0.0002, "step": 52044 }, { "epoch": 1.01, "learning_rate": 3.316965238825379e-05, "loss": 0.0, "step": 52046 }, { "epoch": 1.01, "learning_rate": 3.316900542799656e-05, "loss": 0.0, "step": 52048 }, { "epoch": 1.01, "learning_rate": 3.316835846773933e-05, "loss": 0.0, "step": 52050 }, { "epoch": 1.01, "learning_rate": 3.31677115074821e-05, "loss": 0.0078, "step": 52052 }, { "epoch": 1.01, "learning_rate": 3.316706454722486e-05, "loss": 0.0, "step": 52054 }, { "epoch": 1.01, "learning_rate": 3.316641758696764e-05, "loss": 0.0, "step": 52056 }, { "epoch": 1.01, "learning_rate": 3.31657706267104e-05, "loss": 0.0047, "step": 52058 }, { "epoch": 1.01, "learning_rate": 3.3165123666453175e-05, "loss": 0.0063, "step": 52060 }, { "epoch": 1.01, "learning_rate": 3.316447670619594e-05, "loss": 0.0, "step": 52062 }, { "epoch": 1.01, "learning_rate": 3.3163829745938706e-05, "loss": 0.0, "step": 52064 }, { "epoch": 1.01, "learning_rate": 3.316318278568148e-05, "loss": 0.0, "step": 52066 }, { "epoch": 1.01, "learning_rate": 3.3162535825424244e-05, "loss": 0.0026, "step": 52068 }, { "epoch": 1.01, "learning_rate": 3.3161888865167014e-05, "loss": 0.0001, "step": 52070 }, { "epoch": 1.01, "learning_rate": 3.316124190490978e-05, "loss": 0.0, "step": 52072 }, { "epoch": 1.01, "learning_rate": 3.316059494465255e-05, "loss": 0.0, "step": 52074 }, { "epoch": 1.01, "learning_rate": 3.315994798439532e-05, "loss": 0.0001, "step": 52076 }, { "epoch": 1.01, "learning_rate": 3.315930102413809e-05, "loss": 0.0025, "step": 52078 }, { "epoch": 1.01, "learning_rate": 3.315865406388086e-05, "loss": 0.0, "step": 52080 }, { "epoch": 1.01, "learning_rate": 3.315800710362363e-05, "loss": 0.0, "step": 52082 }, { "epoch": 1.01, "learning_rate": 3.31573601433664e-05, "loss": 0.0, "step": 52084 }, { "epoch": 1.01, "learning_rate": 3.315671318310916e-05, "loss": 0.0114, "step": 52086 }, { "epoch": 1.01, "learning_rate": 3.3156066222851935e-05, "loss": 0.0, "step": 52088 }, { "epoch": 1.01, "learning_rate": 3.31554192625947e-05, "loss": 0.0, "step": 52090 }, { "epoch": 1.01, "learning_rate": 3.315477230233747e-05, "loss": 0.0001, "step": 52092 }, { "epoch": 1.01, "learning_rate": 3.3154125342080236e-05, "loss": 0.0, "step": 52094 }, { "epoch": 1.01, "learning_rate": 3.3153478381823005e-05, "loss": 0.0, "step": 52096 }, { "epoch": 1.01, "learning_rate": 3.3152831421565774e-05, "loss": 0.0, "step": 52098 }, { "epoch": 1.01, "learning_rate": 3.315218446130854e-05, "loss": 0.0, "step": 52100 }, { "epoch": 1.01, "learning_rate": 3.315153750105131e-05, "loss": 0.0047, "step": 52102 }, { "epoch": 1.01, "learning_rate": 3.315089054079408e-05, "loss": 0.0, "step": 52104 }, { "epoch": 1.01, "learning_rate": 3.315024358053685e-05, "loss": 0.0, "step": 52106 }, { "epoch": 1.01, "learning_rate": 3.314959662027962e-05, "loss": 0.0056, "step": 52108 }, { "epoch": 1.01, "learning_rate": 3.314894966002239e-05, "loss": 0.0, "step": 52110 }, { "epoch": 1.01, "learning_rate": 3.314830269976516e-05, "loss": 0.0, "step": 52112 }, { "epoch": 1.01, "learning_rate": 3.314765573950792e-05, "loss": 0.0, "step": 52114 }, { "epoch": 1.01, "learning_rate": 3.3147008779250696e-05, "loss": 0.0, "step": 52116 }, { "epoch": 1.01, "learning_rate": 3.314636181899346e-05, "loss": 0.0062, "step": 52118 }, { "epoch": 1.01, "learning_rate": 3.3145714858736234e-05, "loss": 0.0004, "step": 52120 }, { "epoch": 1.01, "learning_rate": 3.3145067898478996e-05, "loss": 0.0, "step": 52122 }, { "epoch": 1.01, "learning_rate": 3.3144420938221765e-05, "loss": 0.0, "step": 52124 }, { "epoch": 1.01, "learning_rate": 3.3143773977964534e-05, "loss": 0.0001, "step": 52126 }, { "epoch": 1.01, "learning_rate": 3.3143127017707303e-05, "loss": 0.0001, "step": 52128 }, { "epoch": 1.01, "learning_rate": 3.314248005745007e-05, "loss": 0.0, "step": 52130 }, { "epoch": 1.01, "learning_rate": 3.314183309719284e-05, "loss": 0.0, "step": 52132 }, { "epoch": 1.01, "learning_rate": 3.314118613693561e-05, "loss": 0.0, "step": 52134 }, { "epoch": 1.01, "learning_rate": 3.314053917667837e-05, "loss": 0.0, "step": 52136 }, { "epoch": 1.01, "learning_rate": 3.313989221642115e-05, "loss": 0.0, "step": 52138 }, { "epoch": 1.01, "learning_rate": 3.313924525616391e-05, "loss": 0.0, "step": 52140 }, { "epoch": 1.01, "learning_rate": 3.313859829590669e-05, "loss": 0.0, "step": 52142 }, { "epoch": 1.01, "learning_rate": 3.3137951335649456e-05, "loss": 0.0, "step": 52144 }, { "epoch": 1.01, "learning_rate": 3.313730437539222e-05, "loss": 0.0, "step": 52146 }, { "epoch": 1.01, "learning_rate": 3.3136657415134994e-05, "loss": 0.0, "step": 52148 }, { "epoch": 1.01, "learning_rate": 3.3136010454877757e-05, "loss": 0.0, "step": 52150 }, { "epoch": 1.01, "learning_rate": 3.3135363494620526e-05, "loss": 0.0, "step": 52152 }, { "epoch": 1.01, "learning_rate": 3.3134716534363295e-05, "loss": 0.0, "step": 52154 }, { "epoch": 1.01, "learning_rate": 3.3134069574106064e-05, "loss": 0.0, "step": 52156 }, { "epoch": 1.01, "learning_rate": 3.313342261384883e-05, "loss": 0.0, "step": 52158 }, { "epoch": 1.01, "learning_rate": 3.31327756535916e-05, "loss": 0.0, "step": 52160 }, { "epoch": 1.01, "learning_rate": 3.313212869333437e-05, "loss": 0.0059, "step": 52162 }, { "epoch": 1.01, "learning_rate": 3.313148173307714e-05, "loss": 0.0001, "step": 52164 }, { "epoch": 1.01, "learning_rate": 3.313083477281991e-05, "loss": 0.0, "step": 52166 }, { "epoch": 1.01, "learning_rate": 3.313018781256267e-05, "loss": 0.0, "step": 52168 }, { "epoch": 1.01, "learning_rate": 3.312954085230545e-05, "loss": 0.0, "step": 52170 }, { "epoch": 1.01, "learning_rate": 3.312889389204821e-05, "loss": 0.0, "step": 52172 }, { "epoch": 1.01, "learning_rate": 3.312824693179098e-05, "loss": 0.0, "step": 52174 }, { "epoch": 1.01, "learning_rate": 3.3127599971533755e-05, "loss": 0.0002, "step": 52176 }, { "epoch": 1.01, "learning_rate": 3.312695301127652e-05, "loss": 0.0, "step": 52178 }, { "epoch": 1.01, "learning_rate": 3.312630605101929e-05, "loss": 0.0, "step": 52180 }, { "epoch": 1.01, "learning_rate": 3.3125659090762055e-05, "loss": 0.0, "step": 52182 }, { "epoch": 1.01, "learning_rate": 3.3125012130504824e-05, "loss": 0.0, "step": 52184 }, { "epoch": 1.01, "learning_rate": 3.312436517024759e-05, "loss": 0.0025, "step": 52186 }, { "epoch": 1.01, "learning_rate": 3.312371820999036e-05, "loss": 0.0, "step": 52188 }, { "epoch": 1.01, "learning_rate": 3.312307124973313e-05, "loss": 0.0, "step": 52190 }, { "epoch": 1.01, "learning_rate": 3.31224242894759e-05, "loss": 0.0, "step": 52192 }, { "epoch": 1.01, "learning_rate": 3.312177732921867e-05, "loss": 0.0, "step": 52194 }, { "epoch": 1.01, "learning_rate": 3.312113036896143e-05, "loss": 0.0, "step": 52196 }, { "epoch": 1.01, "learning_rate": 3.312048340870421e-05, "loss": 0.0001, "step": 52198 }, { "epoch": 1.01, "learning_rate": 3.311983644844697e-05, "loss": 0.0015, "step": 52200 }, { "epoch": 1.01, "learning_rate": 3.3119189488189746e-05, "loss": 0.0, "step": 52202 }, { "epoch": 1.01, "learning_rate": 3.311854252793251e-05, "loss": 0.0013, "step": 52204 }, { "epoch": 1.01, "learning_rate": 3.311789556767528e-05, "loss": 0.0001, "step": 52206 }, { "epoch": 1.01, "learning_rate": 3.311724860741805e-05, "loss": 0.0064, "step": 52208 }, { "epoch": 1.01, "learning_rate": 3.3116601647160816e-05, "loss": 0.0, "step": 52210 }, { "epoch": 1.01, "learning_rate": 3.3115954686903585e-05, "loss": 0.0, "step": 52212 }, { "epoch": 1.01, "learning_rate": 3.3115307726646354e-05, "loss": 0.0, "step": 52214 }, { "epoch": 1.01, "learning_rate": 3.311466076638912e-05, "loss": 0.0042, "step": 52216 }, { "epoch": 1.01, "learning_rate": 3.3114013806131885e-05, "loss": 0.0001, "step": 52218 }, { "epoch": 1.01, "learning_rate": 3.311336684587466e-05, "loss": 0.0005, "step": 52220 }, { "epoch": 1.01, "learning_rate": 3.311271988561743e-05, "loss": 0.0, "step": 52222 }, { "epoch": 1.01, "learning_rate": 3.31120729253602e-05, "loss": 0.0001, "step": 52224 }, { "epoch": 1.01, "learning_rate": 3.311142596510297e-05, "loss": 0.0, "step": 52226 }, { "epoch": 1.01, "learning_rate": 3.311077900484573e-05, "loss": 0.0, "step": 52228 }, { "epoch": 1.01, "learning_rate": 3.3110132044588506e-05, "loss": 0.0, "step": 52230 }, { "epoch": 1.01, "learning_rate": 3.310948508433127e-05, "loss": 0.0, "step": 52232 }, { "epoch": 1.01, "learning_rate": 3.310883812407404e-05, "loss": 0.0001, "step": 52234 }, { "epoch": 1.01, "learning_rate": 3.310819116381681e-05, "loss": 0.0, "step": 52236 }, { "epoch": 1.01, "learning_rate": 3.3107544203559576e-05, "loss": 0.0, "step": 52238 }, { "epoch": 1.01, "learning_rate": 3.3106897243302345e-05, "loss": 0.0, "step": 52240 }, { "epoch": 1.01, "learning_rate": 3.3106250283045114e-05, "loss": 0.0, "step": 52242 }, { "epoch": 1.01, "learning_rate": 3.310560332278788e-05, "loss": 0.0, "step": 52244 }, { "epoch": 1.01, "learning_rate": 3.310495636253065e-05, "loss": 0.0, "step": 52246 }, { "epoch": 1.01, "learning_rate": 3.310430940227342e-05, "loss": 0.0003, "step": 52248 }, { "epoch": 1.01, "learning_rate": 3.3103662442016184e-05, "loss": 0.0, "step": 52250 }, { "epoch": 1.01, "learning_rate": 3.310301548175896e-05, "loss": 0.0, "step": 52252 }, { "epoch": 1.01, "learning_rate": 3.310236852150173e-05, "loss": 0.006, "step": 52254 }, { "epoch": 1.01, "learning_rate": 3.310172156124449e-05, "loss": 0.0, "step": 52256 }, { "epoch": 1.01, "learning_rate": 3.310107460098727e-05, "loss": 0.0003, "step": 52258 }, { "epoch": 1.01, "learning_rate": 3.310042764073003e-05, "loss": 0.0007, "step": 52260 }, { "epoch": 1.01, "learning_rate": 3.3099780680472805e-05, "loss": 0.0, "step": 52262 }, { "epoch": 1.01, "learning_rate": 3.309913372021557e-05, "loss": 0.001, "step": 52264 }, { "epoch": 1.01, "learning_rate": 3.3098486759958336e-05, "loss": 0.0, "step": 52266 }, { "epoch": 1.01, "learning_rate": 3.3097839799701105e-05, "loss": 0.0006, "step": 52268 }, { "epoch": 1.01, "learning_rate": 3.3097192839443875e-05, "loss": 0.0, "step": 52270 }, { "epoch": 1.01, "learning_rate": 3.3096545879186644e-05, "loss": 0.0, "step": 52272 }, { "epoch": 1.01, "learning_rate": 3.309589891892941e-05, "loss": 0.0, "step": 52274 }, { "epoch": 1.01, "learning_rate": 3.309525195867218e-05, "loss": 0.0, "step": 52276 }, { "epoch": 1.01, "learning_rate": 3.3094604998414944e-05, "loss": 0.0, "step": 52278 }, { "epoch": 1.01, "learning_rate": 3.309395803815772e-05, "loss": 0.0, "step": 52280 }, { "epoch": 1.01, "learning_rate": 3.309331107790048e-05, "loss": 0.0001, "step": 52282 }, { "epoch": 1.01, "learning_rate": 3.309266411764326e-05, "loss": 0.0001, "step": 52284 }, { "epoch": 1.01, "learning_rate": 3.309201715738602e-05, "loss": 0.0, "step": 52286 }, { "epoch": 1.01, "learning_rate": 3.309137019712879e-05, "loss": 0.0, "step": 52288 }, { "epoch": 1.01, "learning_rate": 3.3090723236871565e-05, "loss": 0.0, "step": 52290 }, { "epoch": 1.01, "learning_rate": 3.309007627661433e-05, "loss": 0.0001, "step": 52292 }, { "epoch": 1.01, "learning_rate": 3.30894293163571e-05, "loss": 0.0, "step": 52294 }, { "epoch": 1.02, "learning_rate": 3.3088782356099866e-05, "loss": 0.0011, "step": 52296 }, { "epoch": 1.02, "learning_rate": 3.3088135395842635e-05, "loss": 0.0, "step": 52298 }, { "epoch": 1.02, "learning_rate": 3.3087488435585404e-05, "loss": 0.0, "step": 52300 }, { "epoch": 1.02, "learning_rate": 3.308684147532817e-05, "loss": 0.0, "step": 52302 }, { "epoch": 1.02, "learning_rate": 3.308619451507094e-05, "loss": 0.0, "step": 52304 }, { "epoch": 1.02, "learning_rate": 3.308554755481371e-05, "loss": 0.0, "step": 52306 }, { "epoch": 1.02, "learning_rate": 3.308490059455648e-05, "loss": 0.0, "step": 52308 }, { "epoch": 1.02, "learning_rate": 3.308425363429924e-05, "loss": 0.0001, "step": 52310 }, { "epoch": 1.02, "learning_rate": 3.308360667404202e-05, "loss": 0.0, "step": 52312 }, { "epoch": 1.02, "learning_rate": 3.308295971378478e-05, "loss": 0.0001, "step": 52314 }, { "epoch": 1.02, "learning_rate": 3.308231275352755e-05, "loss": 0.0, "step": 52316 }, { "epoch": 1.02, "learning_rate": 3.308166579327032e-05, "loss": 0.0, "step": 52318 }, { "epoch": 1.02, "learning_rate": 3.308101883301309e-05, "loss": 0.0, "step": 52320 }, { "epoch": 1.02, "learning_rate": 3.3080371872755864e-05, "loss": 0.0, "step": 52322 }, { "epoch": 1.02, "learning_rate": 3.3079724912498626e-05, "loss": 0.0091, "step": 52324 }, { "epoch": 1.02, "learning_rate": 3.3079077952241395e-05, "loss": 0.0004, "step": 52326 }, { "epoch": 1.02, "learning_rate": 3.3078430991984164e-05, "loss": 0.0, "step": 52328 }, { "epoch": 1.02, "learning_rate": 3.3077784031726934e-05, "loss": 0.0008, "step": 52330 }, { "epoch": 1.02, "learning_rate": 3.30771370714697e-05, "loss": 0.0004, "step": 52332 }, { "epoch": 1.02, "learning_rate": 3.307649011121247e-05, "loss": 0.0, "step": 52334 }, { "epoch": 1.02, "learning_rate": 3.307584315095524e-05, "loss": 0.0001, "step": 52336 }, { "epoch": 1.02, "learning_rate": 3.3075196190698e-05, "loss": 0.0, "step": 52338 }, { "epoch": 1.02, "learning_rate": 3.307454923044078e-05, "loss": 0.0, "step": 52340 }, { "epoch": 1.02, "learning_rate": 3.307390227018354e-05, "loss": 0.0001, "step": 52342 }, { "epoch": 1.02, "learning_rate": 3.307325530992632e-05, "loss": 0.0001, "step": 52344 }, { "epoch": 1.02, "learning_rate": 3.307260834966908e-05, "loss": 0.0, "step": 52346 }, { "epoch": 1.02, "learning_rate": 3.307196138941185e-05, "loss": 0.0, "step": 52348 }, { "epoch": 1.02, "learning_rate": 3.307131442915462e-05, "loss": 0.0, "step": 52350 }, { "epoch": 1.02, "learning_rate": 3.307066746889739e-05, "loss": 0.0, "step": 52352 }, { "epoch": 1.02, "learning_rate": 3.3070020508640156e-05, "loss": 0.0413, "step": 52354 }, { "epoch": 1.02, "learning_rate": 3.3069373548382925e-05, "loss": 0.0, "step": 52356 }, { "epoch": 1.02, "learning_rate": 3.3068726588125694e-05, "loss": 0.0008, "step": 52358 }, { "epoch": 1.02, "learning_rate": 3.3068079627868456e-05, "loss": 0.0027, "step": 52360 }, { "epoch": 1.02, "learning_rate": 3.306743266761123e-05, "loss": 0.0, "step": 52362 }, { "epoch": 1.02, "learning_rate": 3.3066785707353994e-05, "loss": 0.0, "step": 52364 }, { "epoch": 1.02, "learning_rate": 3.306613874709677e-05, "loss": 0.0, "step": 52366 }, { "epoch": 1.02, "learning_rate": 3.306549178683954e-05, "loss": 0.0003, "step": 52368 }, { "epoch": 1.02, "learning_rate": 3.30648448265823e-05, "loss": 0.0, "step": 52370 }, { "epoch": 1.02, "learning_rate": 3.306419786632508e-05, "loss": 0.0001, "step": 52372 }, { "epoch": 1.02, "learning_rate": 3.306355090606784e-05, "loss": 0.0, "step": 52374 }, { "epoch": 1.02, "learning_rate": 3.306290394581061e-05, "loss": 0.0, "step": 52376 }, { "epoch": 1.02, "learning_rate": 3.306225698555338e-05, "loss": 0.0002, "step": 52378 }, { "epoch": 1.02, "learning_rate": 3.306161002529615e-05, "loss": 0.0, "step": 52380 }, { "epoch": 1.02, "learning_rate": 3.3060963065038916e-05, "loss": 0.0002, "step": 52382 }, { "epoch": 1.02, "learning_rate": 3.3060316104781685e-05, "loss": 0.0, "step": 52384 }, { "epoch": 1.02, "learning_rate": 3.3059669144524454e-05, "loss": 0.0, "step": 52386 }, { "epoch": 1.02, "learning_rate": 3.3059022184267223e-05, "loss": 0.0008, "step": 52388 }, { "epoch": 1.02, "learning_rate": 3.305837522400999e-05, "loss": 0.0002, "step": 52390 }, { "epoch": 1.02, "learning_rate": 3.3057728263752755e-05, "loss": 0.0, "step": 52392 }, { "epoch": 1.02, "learning_rate": 3.305708130349553e-05, "loss": 0.0, "step": 52394 }, { "epoch": 1.02, "learning_rate": 3.305643434323829e-05, "loss": 0.0001, "step": 52396 }, { "epoch": 1.02, "learning_rate": 3.305578738298106e-05, "loss": 0.0085, "step": 52398 }, { "epoch": 1.02, "learning_rate": 3.305514042272384e-05, "loss": 0.0, "step": 52400 }, { "epoch": 1.02, "learning_rate": 3.30544934624666e-05, "loss": 0.0, "step": 52402 }, { "epoch": 1.02, "learning_rate": 3.3053846502209376e-05, "loss": 0.0012, "step": 52404 }, { "epoch": 1.02, "learning_rate": 3.305319954195214e-05, "loss": 0.0055, "step": 52406 }, { "epoch": 1.02, "learning_rate": 3.305255258169491e-05, "loss": 0.0, "step": 52408 }, { "epoch": 1.02, "learning_rate": 3.3051905621437677e-05, "loss": 0.0, "step": 52410 }, { "epoch": 1.02, "learning_rate": 3.3051258661180446e-05, "loss": 0.0, "step": 52412 }, { "epoch": 1.02, "learning_rate": 3.3050611700923215e-05, "loss": 0.0002, "step": 52414 }, { "epoch": 1.02, "learning_rate": 3.3049964740665984e-05, "loss": 0.0, "step": 52416 }, { "epoch": 1.02, "learning_rate": 3.304931778040875e-05, "loss": 0.0393, "step": 52418 }, { "epoch": 1.02, "learning_rate": 3.3048670820151515e-05, "loss": 0.0, "step": 52420 }, { "epoch": 1.02, "learning_rate": 3.304802385989429e-05, "loss": 0.0, "step": 52422 }, { "epoch": 1.02, "learning_rate": 3.304737689963705e-05, "loss": 0.0, "step": 52424 }, { "epoch": 1.02, "learning_rate": 3.304672993937983e-05, "loss": 0.0027, "step": 52426 }, { "epoch": 1.02, "learning_rate": 3.304608297912259e-05, "loss": 0.0, "step": 52428 }, { "epoch": 1.02, "learning_rate": 3.304543601886536e-05, "loss": 0.0, "step": 52430 }, { "epoch": 1.02, "learning_rate": 3.3044789058608136e-05, "loss": 0.0002, "step": 52432 }, { "epoch": 1.02, "learning_rate": 3.30441420983509e-05, "loss": 0.0, "step": 52434 }, { "epoch": 1.02, "learning_rate": 3.304349513809367e-05, "loss": 0.0, "step": 52436 }, { "epoch": 1.02, "learning_rate": 3.304284817783644e-05, "loss": 0.0, "step": 52438 }, { "epoch": 1.02, "learning_rate": 3.3042201217579206e-05, "loss": 0.0, "step": 52440 }, { "epoch": 1.02, "learning_rate": 3.304155425732197e-05, "loss": 0.0, "step": 52442 }, { "epoch": 1.02, "learning_rate": 3.3040907297064744e-05, "loss": 0.0, "step": 52444 }, { "epoch": 1.02, "learning_rate": 3.304026033680751e-05, "loss": 0.0, "step": 52446 }, { "epoch": 1.02, "learning_rate": 3.303961337655028e-05, "loss": 0.0, "step": 52448 }, { "epoch": 1.02, "learning_rate": 3.303896641629305e-05, "loss": 0.0037, "step": 52450 }, { "epoch": 1.02, "learning_rate": 3.3038319456035814e-05, "loss": 0.0, "step": 52452 }, { "epoch": 1.02, "learning_rate": 3.303767249577859e-05, "loss": 0.0, "step": 52454 }, { "epoch": 1.02, "learning_rate": 3.303702553552135e-05, "loss": 0.0, "step": 52456 }, { "epoch": 1.02, "learning_rate": 3.303637857526412e-05, "loss": 0.001, "step": 52458 }, { "epoch": 1.02, "learning_rate": 3.303573161500689e-05, "loss": 0.0003, "step": 52460 }, { "epoch": 1.02, "learning_rate": 3.303508465474966e-05, "loss": 0.0, "step": 52462 }, { "epoch": 1.02, "learning_rate": 3.303443769449243e-05, "loss": 0.0, "step": 52464 }, { "epoch": 1.02, "learning_rate": 3.30337907342352e-05, "loss": 0.0001, "step": 52466 }, { "epoch": 1.02, "learning_rate": 3.3033143773977966e-05, "loss": 0.0001, "step": 52468 }, { "epoch": 1.02, "learning_rate": 3.3032496813720736e-05, "loss": 0.0025, "step": 52470 }, { "epoch": 1.02, "learning_rate": 3.3031849853463505e-05, "loss": 0.0, "step": 52472 }, { "epoch": 1.02, "learning_rate": 3.303120289320627e-05, "loss": 0.0, "step": 52474 }, { "epoch": 1.02, "learning_rate": 3.303055593294904e-05, "loss": 0.0, "step": 52476 }, { "epoch": 1.02, "learning_rate": 3.302990897269181e-05, "loss": 0.0, "step": 52478 }, { "epoch": 1.02, "learning_rate": 3.3029262012434574e-05, "loss": 0.0, "step": 52480 }, { "epoch": 1.02, "learning_rate": 3.302861505217735e-05, "loss": 0.0, "step": 52482 }, { "epoch": 1.02, "learning_rate": 3.302796809192011e-05, "loss": 0.0046, "step": 52484 }, { "epoch": 1.02, "learning_rate": 3.302732113166289e-05, "loss": 0.0, "step": 52486 }, { "epoch": 1.02, "learning_rate": 3.302667417140565e-05, "loss": 0.0034, "step": 52488 }, { "epoch": 1.02, "learning_rate": 3.302602721114842e-05, "loss": 0.0009, "step": 52490 }, { "epoch": 1.02, "learning_rate": 3.302538025089119e-05, "loss": 0.0004, "step": 52492 }, { "epoch": 1.02, "learning_rate": 3.302473329063396e-05, "loss": 0.0, "step": 52494 }, { "epoch": 1.02, "learning_rate": 3.302408633037673e-05, "loss": 0.0, "step": 52496 }, { "epoch": 1.02, "learning_rate": 3.3023439370119496e-05, "loss": 0.0, "step": 52498 }, { "epoch": 1.02, "learning_rate": 3.3022792409862265e-05, "loss": 0.0, "step": 52500 }, { "epoch": 1.02, "learning_rate": 3.302214544960503e-05, "loss": 0.0, "step": 52502 }, { "epoch": 1.02, "learning_rate": 3.30214984893478e-05, "loss": 0.0046, "step": 52504 }, { "epoch": 1.02, "learning_rate": 3.3020851529090566e-05, "loss": 0.0, "step": 52506 }, { "epoch": 1.02, "learning_rate": 3.302020456883334e-05, "loss": 0.0009, "step": 52508 }, { "epoch": 1.02, "learning_rate": 3.301955760857611e-05, "loss": 0.0001, "step": 52510 }, { "epoch": 1.02, "learning_rate": 3.301891064831887e-05, "loss": 0.0001, "step": 52512 }, { "epoch": 1.02, "learning_rate": 3.301826368806165e-05, "loss": 0.0, "step": 52514 }, { "epoch": 1.02, "learning_rate": 3.301761672780441e-05, "loss": 0.0, "step": 52516 }, { "epoch": 1.02, "learning_rate": 3.301696976754718e-05, "loss": 0.0071, "step": 52518 }, { "epoch": 1.02, "learning_rate": 3.301632280728995e-05, "loss": 0.0, "step": 52520 }, { "epoch": 1.02, "learning_rate": 3.301567584703272e-05, "loss": 0.0003, "step": 52522 }, { "epoch": 1.02, "learning_rate": 3.301502888677549e-05, "loss": 0.0, "step": 52524 }, { "epoch": 1.02, "learning_rate": 3.3014381926518256e-05, "loss": 0.0, "step": 52526 }, { "epoch": 1.02, "learning_rate": 3.3013734966261025e-05, "loss": 0.0029, "step": 52528 }, { "epoch": 1.02, "learning_rate": 3.3013088006003795e-05, "loss": 0.0, "step": 52530 }, { "epoch": 1.02, "learning_rate": 3.3012441045746564e-05, "loss": 0.0, "step": 52532 }, { "epoch": 1.02, "learning_rate": 3.3011794085489326e-05, "loss": 0.0, "step": 52534 }, { "epoch": 1.02, "learning_rate": 3.30111471252321e-05, "loss": 0.0, "step": 52536 }, { "epoch": 1.02, "learning_rate": 3.3010500164974864e-05, "loss": 0.0005, "step": 52538 }, { "epoch": 1.02, "learning_rate": 3.300985320471763e-05, "loss": 0.0, "step": 52540 }, { "epoch": 1.02, "learning_rate": 3.30092062444604e-05, "loss": 0.0001, "step": 52542 }, { "epoch": 1.02, "learning_rate": 3.300855928420317e-05, "loss": 0.001, "step": 52544 }, { "epoch": 1.02, "learning_rate": 3.300791232394595e-05, "loss": 0.0, "step": 52546 }, { "epoch": 1.02, "learning_rate": 3.300726536368871e-05, "loss": 0.0008, "step": 52548 }, { "epoch": 1.02, "learning_rate": 3.300661840343148e-05, "loss": 0.0, "step": 52550 }, { "epoch": 1.02, "learning_rate": 3.300597144317425e-05, "loss": 0.003, "step": 52552 }, { "epoch": 1.02, "learning_rate": 3.300532448291702e-05, "loss": 0.0002, "step": 52554 }, { "epoch": 1.02, "learning_rate": 3.3004677522659786e-05, "loss": 0.0001, "step": 52556 }, { "epoch": 1.02, "learning_rate": 3.3004030562402555e-05, "loss": 0.0, "step": 52558 }, { "epoch": 1.02, "learning_rate": 3.3003383602145324e-05, "loss": 0.0, "step": 52560 }, { "epoch": 1.02, "learning_rate": 3.3002736641888086e-05, "loss": 0.0, "step": 52562 }, { "epoch": 1.02, "learning_rate": 3.300208968163086e-05, "loss": 0.0067, "step": 52564 }, { "epoch": 1.02, "learning_rate": 3.3001442721373624e-05, "loss": 0.0, "step": 52566 }, { "epoch": 1.02, "learning_rate": 3.30007957611164e-05, "loss": 0.0, "step": 52568 }, { "epoch": 1.02, "learning_rate": 3.300014880085916e-05, "loss": 0.0005, "step": 52570 }, { "epoch": 1.02, "learning_rate": 3.299950184060193e-05, "loss": 0.0, "step": 52572 }, { "epoch": 1.02, "learning_rate": 3.29988548803447e-05, "loss": 0.0, "step": 52574 }, { "epoch": 1.02, "learning_rate": 3.299820792008747e-05, "loss": 0.0013, "step": 52576 }, { "epoch": 1.02, "learning_rate": 3.299756095983024e-05, "loss": 0.0, "step": 52578 }, { "epoch": 1.02, "learning_rate": 3.299691399957301e-05, "loss": 0.0001, "step": 52580 }, { "epoch": 1.02, "learning_rate": 3.299626703931578e-05, "loss": 0.0, "step": 52582 }, { "epoch": 1.02, "learning_rate": 3.299562007905854e-05, "loss": 0.0012, "step": 52584 }, { "epoch": 1.02, "learning_rate": 3.2994973118801315e-05, "loss": 0.0, "step": 52586 }, { "epoch": 1.02, "learning_rate": 3.2994326158544084e-05, "loss": 0.0, "step": 52588 }, { "epoch": 1.02, "learning_rate": 3.2993679198286854e-05, "loss": 0.0, "step": 52590 }, { "epoch": 1.02, "learning_rate": 3.299303223802962e-05, "loss": 0.0, "step": 52592 }, { "epoch": 1.02, "learning_rate": 3.2992385277772385e-05, "loss": 0.0001, "step": 52594 }, { "epoch": 1.02, "learning_rate": 3.299173831751516e-05, "loss": 0.0003, "step": 52596 }, { "epoch": 1.02, "learning_rate": 3.299109135725792e-05, "loss": 0.0, "step": 52598 }, { "epoch": 1.02, "learning_rate": 3.299044439700069e-05, "loss": 0.0001, "step": 52600 }, { "epoch": 1.02, "learning_rate": 3.298979743674346e-05, "loss": 0.0, "step": 52602 }, { "epoch": 1.02, "learning_rate": 3.298915047648623e-05, "loss": 0.0, "step": 52604 }, { "epoch": 1.02, "learning_rate": 3.2988503516229e-05, "loss": 0.0, "step": 52606 }, { "epoch": 1.02, "learning_rate": 3.298785655597177e-05, "loss": 0.0, "step": 52608 }, { "epoch": 1.02, "learning_rate": 3.298720959571454e-05, "loss": 0.0, "step": 52610 }, { "epoch": 1.02, "learning_rate": 3.298656263545731e-05, "loss": 0.0, "step": 52612 }, { "epoch": 1.02, "learning_rate": 3.2985915675200076e-05, "loss": 0.0, "step": 52614 }, { "epoch": 1.02, "learning_rate": 3.298526871494284e-05, "loss": 0.0016, "step": 52616 }, { "epoch": 1.02, "learning_rate": 3.2984621754685614e-05, "loss": 0.0, "step": 52618 }, { "epoch": 1.02, "learning_rate": 3.2983974794428376e-05, "loss": 0.0, "step": 52620 }, { "epoch": 1.02, "learning_rate": 3.2983327834171145e-05, "loss": 0.0, "step": 52622 }, { "epoch": 1.02, "learning_rate": 3.298268087391392e-05, "loss": 0.0001, "step": 52624 }, { "epoch": 1.02, "learning_rate": 3.2982033913656683e-05, "loss": 0.0004, "step": 52626 }, { "epoch": 1.02, "learning_rate": 3.298138695339946e-05, "loss": 0.0, "step": 52628 }, { "epoch": 1.02, "learning_rate": 3.298073999314222e-05, "loss": 0.0, "step": 52630 }, { "epoch": 1.02, "learning_rate": 3.298009303288499e-05, "loss": 0.0001, "step": 52632 }, { "epoch": 1.02, "learning_rate": 3.297944607262776e-05, "loss": 0.0, "step": 52634 }, { "epoch": 1.02, "learning_rate": 3.297879911237053e-05, "loss": 0.0, "step": 52636 }, { "epoch": 1.02, "learning_rate": 3.29781521521133e-05, "loss": 0.0, "step": 52638 }, { "epoch": 1.02, "learning_rate": 3.297750519185607e-05, "loss": 0.0, "step": 52640 }, { "epoch": 1.02, "learning_rate": 3.2976858231598836e-05, "loss": 0.0025, "step": 52642 }, { "epoch": 1.02, "learning_rate": 3.29762112713416e-05, "loss": 0.0, "step": 52644 }, { "epoch": 1.02, "learning_rate": 3.2975564311084374e-05, "loss": 0.0001, "step": 52646 }, { "epoch": 1.02, "learning_rate": 3.297491735082714e-05, "loss": 0.0005, "step": 52648 }, { "epoch": 1.02, "learning_rate": 3.297427039056991e-05, "loss": 0.0, "step": 52650 }, { "epoch": 1.02, "learning_rate": 3.2973623430312675e-05, "loss": 0.0, "step": 52652 }, { "epoch": 1.02, "learning_rate": 3.2972976470055444e-05, "loss": 0.0, "step": 52654 }, { "epoch": 1.02, "learning_rate": 3.297232950979822e-05, "loss": 0.0019, "step": 52656 }, { "epoch": 1.02, "learning_rate": 3.297168254954098e-05, "loss": 0.0, "step": 52658 }, { "epoch": 1.02, "learning_rate": 3.297103558928375e-05, "loss": 0.0003, "step": 52660 }, { "epoch": 1.02, "learning_rate": 3.297038862902652e-05, "loss": 0.0006, "step": 52662 }, { "epoch": 1.02, "learning_rate": 3.296974166876929e-05, "loss": 0.0006, "step": 52664 }, { "epoch": 1.02, "learning_rate": 3.296909470851205e-05, "loss": 0.0, "step": 52666 }, { "epoch": 1.02, "learning_rate": 3.296844774825483e-05, "loss": 0.0, "step": 52668 }, { "epoch": 1.02, "learning_rate": 3.2967800787997597e-05, "loss": 0.0001, "step": 52670 }, { "epoch": 1.02, "learning_rate": 3.2967153827740366e-05, "loss": 0.0, "step": 52672 }, { "epoch": 1.02, "learning_rate": 3.2966506867483135e-05, "loss": 0.0001, "step": 52674 }, { "epoch": 1.02, "learning_rate": 3.29658599072259e-05, "loss": 0.0011, "step": 52676 }, { "epoch": 1.02, "learning_rate": 3.296521294696867e-05, "loss": 0.0, "step": 52678 }, { "epoch": 1.02, "learning_rate": 3.2964565986711435e-05, "loss": 0.0, "step": 52680 }, { "epoch": 1.02, "learning_rate": 3.2963919026454204e-05, "loss": 0.0, "step": 52682 }, { "epoch": 1.02, "learning_rate": 3.296327206619697e-05, "loss": 0.0, "step": 52684 }, { "epoch": 1.02, "learning_rate": 3.296262510593974e-05, "loss": 0.0, "step": 52686 }, { "epoch": 1.02, "learning_rate": 3.296197814568252e-05, "loss": 0.0001, "step": 52688 }, { "epoch": 1.02, "learning_rate": 3.296133118542528e-05, "loss": 0.0, "step": 52690 }, { "epoch": 1.02, "learning_rate": 3.296068422516805e-05, "loss": 0.0, "step": 52692 }, { "epoch": 1.02, "learning_rate": 3.296003726491082e-05, "loss": 0.0037, "step": 52694 }, { "epoch": 1.02, "learning_rate": 3.295939030465359e-05, "loss": 0.0001, "step": 52696 }, { "epoch": 1.02, "learning_rate": 3.295874334439635e-05, "loss": 0.0001, "step": 52698 }, { "epoch": 1.02, "learning_rate": 3.2958096384139126e-05, "loss": 0.0, "step": 52700 }, { "epoch": 1.02, "learning_rate": 3.2957449423881895e-05, "loss": 0.0, "step": 52702 }, { "epoch": 1.02, "learning_rate": 3.295680246362466e-05, "loss": 0.0, "step": 52704 }, { "epoch": 1.02, "learning_rate": 3.295615550336743e-05, "loss": 0.0, "step": 52706 }, { "epoch": 1.02, "learning_rate": 3.2955508543110196e-05, "loss": 0.0, "step": 52708 }, { "epoch": 1.02, "learning_rate": 3.295486158285297e-05, "loss": 0.0001, "step": 52710 }, { "epoch": 1.02, "learning_rate": 3.2954214622595734e-05, "loss": 0.0001, "step": 52712 }, { "epoch": 1.02, "learning_rate": 3.29535676623385e-05, "loss": 0.0, "step": 52714 }, { "epoch": 1.02, "learning_rate": 3.295292070208127e-05, "loss": 0.0, "step": 52716 }, { "epoch": 1.02, "learning_rate": 3.295227374182404e-05, "loss": 0.0012, "step": 52718 }, { "epoch": 1.02, "learning_rate": 3.295162678156681e-05, "loss": 0.0001, "step": 52720 }, { "epoch": 1.02, "learning_rate": 3.295097982130958e-05, "loss": 0.0, "step": 52722 }, { "epoch": 1.02, "learning_rate": 3.295033286105235e-05, "loss": 0.0007, "step": 52724 }, { "epoch": 1.02, "learning_rate": 3.294968590079511e-05, "loss": 0.0007, "step": 52726 }, { "epoch": 1.02, "learning_rate": 3.2949038940537886e-05, "loss": 0.01, "step": 52728 }, { "epoch": 1.02, "learning_rate": 3.294839198028065e-05, "loss": 0.0001, "step": 52730 }, { "epoch": 1.02, "learning_rate": 3.2947745020023425e-05, "loss": 0.0, "step": 52732 }, { "epoch": 1.02, "learning_rate": 3.2947098059766194e-05, "loss": 0.0002, "step": 52734 }, { "epoch": 1.02, "learning_rate": 3.2946451099508956e-05, "loss": 0.0, "step": 52736 }, { "epoch": 1.02, "learning_rate": 3.294580413925173e-05, "loss": 0.0001, "step": 52738 }, { "epoch": 1.02, "learning_rate": 3.2945157178994494e-05, "loss": 0.0, "step": 52740 }, { "epoch": 1.02, "learning_rate": 3.294451021873726e-05, "loss": 0.0, "step": 52742 }, { "epoch": 1.02, "learning_rate": 3.294386325848003e-05, "loss": 0.0001, "step": 52744 }, { "epoch": 1.02, "learning_rate": 3.29432162982228e-05, "loss": 0.0002, "step": 52746 }, { "epoch": 1.02, "learning_rate": 3.294256933796557e-05, "loss": 0.0, "step": 52748 }, { "epoch": 1.02, "learning_rate": 3.294192237770834e-05, "loss": 0.0017, "step": 52750 }, { "epoch": 1.02, "learning_rate": 3.294127541745111e-05, "loss": 0.0, "step": 52752 }, { "epoch": 1.02, "learning_rate": 3.294062845719388e-05, "loss": 0.0, "step": 52754 }, { "epoch": 1.02, "learning_rate": 3.293998149693665e-05, "loss": 0.0, "step": 52756 }, { "epoch": 1.02, "learning_rate": 3.293933453667941e-05, "loss": 0.0, "step": 52758 }, { "epoch": 1.02, "learning_rate": 3.2938687576422185e-05, "loss": 0.0001, "step": 52760 }, { "epoch": 1.02, "learning_rate": 3.293804061616495e-05, "loss": 0.0, "step": 52762 }, { "epoch": 1.02, "learning_rate": 3.2937393655907716e-05, "loss": 0.0, "step": 52764 }, { "epoch": 1.02, "learning_rate": 3.2936746695650485e-05, "loss": 0.0, "step": 52766 }, { "epoch": 1.02, "learning_rate": 3.2936099735393255e-05, "loss": 0.0, "step": 52768 }, { "epoch": 1.02, "learning_rate": 3.293545277513603e-05, "loss": 0.0009, "step": 52770 }, { "epoch": 1.02, "learning_rate": 3.293480581487879e-05, "loss": 0.0, "step": 52772 }, { "epoch": 1.02, "learning_rate": 3.293415885462156e-05, "loss": 0.0001, "step": 52774 }, { "epoch": 1.02, "learning_rate": 3.293351189436433e-05, "loss": 0.0, "step": 52776 }, { "epoch": 1.02, "learning_rate": 3.29328649341071e-05, "loss": 0.0, "step": 52778 }, { "epoch": 1.02, "learning_rate": 3.293221797384987e-05, "loss": 0.0, "step": 52780 }, { "epoch": 1.02, "learning_rate": 3.293157101359264e-05, "loss": 0.0, "step": 52782 }, { "epoch": 1.02, "learning_rate": 3.293092405333541e-05, "loss": 0.0044, "step": 52784 }, { "epoch": 1.02, "learning_rate": 3.293027709307817e-05, "loss": 0.0002, "step": 52786 }, { "epoch": 1.02, "learning_rate": 3.2929630132820945e-05, "loss": 0.0, "step": 52788 }, { "epoch": 1.02, "learning_rate": 3.292898317256371e-05, "loss": 0.0, "step": 52790 }, { "epoch": 1.02, "learning_rate": 3.2928336212306484e-05, "loss": 0.0, "step": 52792 }, { "epoch": 1.02, "learning_rate": 3.2927689252049246e-05, "loss": 0.0179, "step": 52794 }, { "epoch": 1.02, "learning_rate": 3.2927042291792015e-05, "loss": 0.0, "step": 52796 }, { "epoch": 1.02, "learning_rate": 3.2926395331534784e-05, "loss": 0.0, "step": 52798 }, { "epoch": 1.02, "learning_rate": 3.292574837127755e-05, "loss": 0.0083, "step": 52800 }, { "epoch": 1.02, "learning_rate": 3.292510141102032e-05, "loss": 0.0, "step": 52802 }, { "epoch": 1.02, "learning_rate": 3.292445445076309e-05, "loss": 0.0, "step": 52804 }, { "epoch": 1.02, "learning_rate": 3.292380749050586e-05, "loss": 0.0, "step": 52806 }, { "epoch": 1.02, "learning_rate": 3.292316053024862e-05, "loss": 0.0, "step": 52808 }, { "epoch": 1.02, "learning_rate": 3.29225135699914e-05, "loss": 0.0, "step": 52810 }, { "epoch": 1.03, "learning_rate": 3.292186660973417e-05, "loss": 0.0, "step": 52812 }, { "epoch": 1.03, "learning_rate": 3.292121964947694e-05, "loss": 0.0, "step": 52814 }, { "epoch": 1.03, "learning_rate": 3.2920572689219706e-05, "loss": 0.0, "step": 52816 }, { "epoch": 1.03, "learning_rate": 3.291992572896247e-05, "loss": 0.0, "step": 52818 }, { "epoch": 1.03, "learning_rate": 3.2919278768705244e-05, "loss": 0.0, "step": 52820 }, { "epoch": 1.03, "learning_rate": 3.2918631808448006e-05, "loss": 0.0, "step": 52822 }, { "epoch": 1.03, "learning_rate": 3.2917984848190775e-05, "loss": 0.0001, "step": 52824 }, { "epoch": 1.03, "learning_rate": 3.2917337887933544e-05, "loss": 0.0048, "step": 52826 }, { "epoch": 1.03, "learning_rate": 3.2916690927676314e-05, "loss": 0.0007, "step": 52828 }, { "epoch": 1.03, "learning_rate": 3.291604396741908e-05, "loss": 0.0, "step": 52830 }, { "epoch": 1.03, "learning_rate": 3.291539700716185e-05, "loss": 0.0086, "step": 52832 }, { "epoch": 1.03, "learning_rate": 3.291475004690462e-05, "loss": 0.0049, "step": 52834 }, { "epoch": 1.03, "learning_rate": 3.291410308664739e-05, "loss": 0.0, "step": 52836 }, { "epoch": 1.03, "learning_rate": 3.291345612639016e-05, "loss": 0.0, "step": 52838 }, { "epoch": 1.03, "learning_rate": 3.291280916613292e-05, "loss": 0.0004, "step": 52840 }, { "epoch": 1.03, "learning_rate": 3.29121622058757e-05, "loss": 0.0, "step": 52842 }, { "epoch": 1.03, "learning_rate": 3.291151524561846e-05, "loss": 0.0, "step": 52844 }, { "epoch": 1.03, "learning_rate": 3.291086828536123e-05, "loss": 0.0, "step": 52846 }, { "epoch": 1.03, "learning_rate": 3.2910221325104004e-05, "loss": 0.0, "step": 52848 }, { "epoch": 1.03, "learning_rate": 3.290957436484677e-05, "loss": 0.0001, "step": 52850 }, { "epoch": 1.03, "learning_rate": 3.290892740458954e-05, "loss": 0.0, "step": 52852 }, { "epoch": 1.03, "learning_rate": 3.2908280444332305e-05, "loss": 0.0, "step": 52854 }, { "epoch": 1.03, "learning_rate": 3.2907633484075074e-05, "loss": 0.0, "step": 52856 }, { "epoch": 1.03, "learning_rate": 3.290698652381784e-05, "loss": 0.0, "step": 52858 }, { "epoch": 1.03, "learning_rate": 3.290633956356061e-05, "loss": 0.0, "step": 52860 }, { "epoch": 1.03, "learning_rate": 3.290569260330338e-05, "loss": 0.0, "step": 52862 }, { "epoch": 1.03, "learning_rate": 3.290504564304615e-05, "loss": 0.0, "step": 52864 }, { "epoch": 1.03, "learning_rate": 3.290439868278892e-05, "loss": 0.0, "step": 52866 }, { "epoch": 1.03, "learning_rate": 3.290375172253168e-05, "loss": 0.0, "step": 52868 }, { "epoch": 1.03, "learning_rate": 3.290310476227446e-05, "loss": 0.0, "step": 52870 }, { "epoch": 1.03, "learning_rate": 3.290245780201722e-05, "loss": 0.0001, "step": 52872 }, { "epoch": 1.03, "learning_rate": 3.2901810841759996e-05, "loss": 0.0002, "step": 52874 }, { "epoch": 1.03, "learning_rate": 3.290116388150276e-05, "loss": 0.0005, "step": 52876 }, { "epoch": 1.03, "learning_rate": 3.290051692124553e-05, "loss": 0.0, "step": 52878 }, { "epoch": 1.03, "learning_rate": 3.28998699609883e-05, "loss": 0.0, "step": 52880 }, { "epoch": 1.03, "learning_rate": 3.2899223000731065e-05, "loss": 0.0, "step": 52882 }, { "epoch": 1.03, "learning_rate": 3.2898576040473834e-05, "loss": 0.0, "step": 52884 }, { "epoch": 1.03, "learning_rate": 3.2897929080216603e-05, "loss": 0.0, "step": 52886 }, { "epoch": 1.03, "learning_rate": 3.289728211995937e-05, "loss": 0.0, "step": 52888 }, { "epoch": 1.03, "learning_rate": 3.289663515970214e-05, "loss": 0.0, "step": 52890 }, { "epoch": 1.03, "learning_rate": 3.289598819944491e-05, "loss": 0.0, "step": 52892 }, { "epoch": 1.03, "learning_rate": 3.289534123918768e-05, "loss": 0.0005, "step": 52894 }, { "epoch": 1.03, "learning_rate": 3.289469427893045e-05, "loss": 0.0001, "step": 52896 }, { "epoch": 1.03, "learning_rate": 3.289404731867322e-05, "loss": 0.0, "step": 52898 }, { "epoch": 1.03, "learning_rate": 3.289340035841598e-05, "loss": 0.0, "step": 52900 }, { "epoch": 1.03, "learning_rate": 3.2892753398158756e-05, "loss": 0.0, "step": 52902 }, { "epoch": 1.03, "learning_rate": 3.289210643790152e-05, "loss": 0.0, "step": 52904 }, { "epoch": 1.03, "learning_rate": 3.289145947764429e-05, "loss": 0.0002, "step": 52906 }, { "epoch": 1.03, "learning_rate": 3.2890812517387057e-05, "loss": 0.0, "step": 52908 }, { "epoch": 1.03, "learning_rate": 3.2890165557129826e-05, "loss": 0.0001, "step": 52910 }, { "epoch": 1.03, "learning_rate": 3.28895185968726e-05, "loss": 0.0044, "step": 52912 }, { "epoch": 1.03, "learning_rate": 3.2888871636615364e-05, "loss": 0.0, "step": 52914 }, { "epoch": 1.03, "learning_rate": 3.288822467635813e-05, "loss": 0.0, "step": 52916 }, { "epoch": 1.03, "learning_rate": 3.28875777161009e-05, "loss": 0.0001, "step": 52918 }, { "epoch": 1.03, "learning_rate": 3.288693075584367e-05, "loss": 0.0001, "step": 52920 }, { "epoch": 1.03, "learning_rate": 3.2886283795586433e-05, "loss": 0.0, "step": 52922 }, { "epoch": 1.03, "learning_rate": 3.288563683532921e-05, "loss": 0.0001, "step": 52924 }, { "epoch": 1.03, "learning_rate": 3.288498987507198e-05, "loss": 0.0, "step": 52926 }, { "epoch": 1.03, "learning_rate": 3.288434291481474e-05, "loss": 0.0, "step": 52928 }, { "epoch": 1.03, "learning_rate": 3.2883695954557517e-05, "loss": 0.0, "step": 52930 }, { "epoch": 1.03, "learning_rate": 3.288304899430028e-05, "loss": 0.0, "step": 52932 }, { "epoch": 1.03, "learning_rate": 3.2882402034043055e-05, "loss": 0.0, "step": 52934 }, { "epoch": 1.03, "learning_rate": 3.288175507378582e-05, "loss": 0.0001, "step": 52936 }, { "epoch": 1.03, "learning_rate": 3.2881108113528586e-05, "loss": 0.0, "step": 52938 }, { "epoch": 1.03, "learning_rate": 3.2880461153271355e-05, "loss": 0.002, "step": 52940 }, { "epoch": 1.03, "learning_rate": 3.2879814193014124e-05, "loss": 0.0064, "step": 52942 }, { "epoch": 1.03, "learning_rate": 3.287916723275689e-05, "loss": 0.0, "step": 52944 }, { "epoch": 1.03, "learning_rate": 3.287852027249966e-05, "loss": 0.0, "step": 52946 }, { "epoch": 1.03, "learning_rate": 3.287787331224243e-05, "loss": 0.0, "step": 52948 }, { "epoch": 1.03, "learning_rate": 3.2877226351985194e-05, "loss": 0.0, "step": 52950 }, { "epoch": 1.03, "learning_rate": 3.287657939172797e-05, "loss": 0.0, "step": 52952 }, { "epoch": 1.03, "learning_rate": 3.287593243147073e-05, "loss": 0.0, "step": 52954 }, { "epoch": 1.03, "learning_rate": 3.287528547121351e-05, "loss": 0.0002, "step": 52956 }, { "epoch": 1.03, "learning_rate": 3.287463851095628e-05, "loss": 0.0, "step": 52958 }, { "epoch": 1.03, "learning_rate": 3.287399155069904e-05, "loss": 0.0003, "step": 52960 }, { "epoch": 1.03, "learning_rate": 3.2873344590441815e-05, "loss": 0.0001, "step": 52962 }, { "epoch": 1.03, "learning_rate": 3.287269763018458e-05, "loss": 0.0, "step": 52964 }, { "epoch": 1.03, "learning_rate": 3.2872050669927346e-05, "loss": 0.0, "step": 52966 }, { "epoch": 1.03, "learning_rate": 3.2871403709670116e-05, "loss": 0.0, "step": 52968 }, { "epoch": 1.03, "learning_rate": 3.2870756749412885e-05, "loss": 0.0, "step": 52970 }, { "epoch": 1.03, "learning_rate": 3.2870109789155654e-05, "loss": 0.0022, "step": 52972 }, { "epoch": 1.03, "learning_rate": 3.286946282889842e-05, "loss": 0.0024, "step": 52974 }, { "epoch": 1.03, "learning_rate": 3.286881586864119e-05, "loss": 0.0002, "step": 52976 }, { "epoch": 1.03, "learning_rate": 3.286816890838396e-05, "loss": 0.0, "step": 52978 }, { "epoch": 1.03, "learning_rate": 3.286752194812673e-05, "loss": 0.0002, "step": 52980 }, { "epoch": 1.03, "learning_rate": 3.286687498786949e-05, "loss": 0.0007, "step": 52982 }, { "epoch": 1.03, "learning_rate": 3.286622802761227e-05, "loss": 0.0006, "step": 52984 }, { "epoch": 1.03, "learning_rate": 3.286558106735503e-05, "loss": 0.0001, "step": 52986 }, { "epoch": 1.03, "learning_rate": 3.28649341070978e-05, "loss": 0.0001, "step": 52988 }, { "epoch": 1.03, "learning_rate": 3.2864287146840576e-05, "loss": 0.0, "step": 52990 }, { "epoch": 1.03, "learning_rate": 3.286364018658334e-05, "loss": 0.0015, "step": 52992 }, { "epoch": 1.03, "learning_rate": 3.2862993226326114e-05, "loss": 0.0, "step": 52994 }, { "epoch": 1.03, "learning_rate": 3.2862346266068876e-05, "loss": 0.0004, "step": 52996 }, { "epoch": 1.03, "learning_rate": 3.2861699305811645e-05, "loss": 0.0, "step": 52998 }, { "epoch": 1.03, "learning_rate": 3.2861052345554414e-05, "loss": 0.0, "step": 53000 }, { "epoch": 1.03, "learning_rate": 3.286040538529718e-05, "loss": 0.0, "step": 53002 }, { "epoch": 1.03, "learning_rate": 3.285975842503995e-05, "loss": 0.0011, "step": 53004 }, { "epoch": 1.03, "learning_rate": 3.285911146478272e-05, "loss": 0.0, "step": 53006 }, { "epoch": 1.03, "learning_rate": 3.285846450452549e-05, "loss": 0.0008, "step": 53008 }, { "epoch": 1.03, "learning_rate": 3.285781754426825e-05, "loss": 0.0, "step": 53010 }, { "epoch": 1.03, "learning_rate": 3.285717058401103e-05, "loss": 0.0, "step": 53012 }, { "epoch": 1.03, "learning_rate": 3.285652362375379e-05, "loss": 0.0, "step": 53014 }, { "epoch": 1.03, "learning_rate": 3.285587666349657e-05, "loss": 0.0, "step": 53016 }, { "epoch": 1.03, "learning_rate": 3.285522970323933e-05, "loss": 0.0, "step": 53018 }, { "epoch": 1.03, "learning_rate": 3.28545827429821e-05, "loss": 0.0, "step": 53020 }, { "epoch": 1.03, "learning_rate": 3.285393578272487e-05, "loss": 0.0, "step": 53022 }, { "epoch": 1.03, "learning_rate": 3.2853288822467636e-05, "loss": 0.0001, "step": 53024 }, { "epoch": 1.03, "learning_rate": 3.2852641862210405e-05, "loss": 0.0, "step": 53026 }, { "epoch": 1.03, "learning_rate": 3.2851994901953175e-05, "loss": 0.0, "step": 53028 }, { "epoch": 1.03, "learning_rate": 3.2851347941695944e-05, "loss": 0.0, "step": 53030 }, { "epoch": 1.03, "learning_rate": 3.285070098143871e-05, "loss": 0.0, "step": 53032 }, { "epoch": 1.03, "learning_rate": 3.285005402118148e-05, "loss": 0.0, "step": 53034 }, { "epoch": 1.03, "learning_rate": 3.284940706092425e-05, "loss": 0.0, "step": 53036 }, { "epoch": 1.03, "learning_rate": 3.284876010066702e-05, "loss": 0.0002, "step": 53038 }, { "epoch": 1.03, "learning_rate": 3.284811314040979e-05, "loss": 0.0, "step": 53040 }, { "epoch": 1.03, "learning_rate": 3.284746618015255e-05, "loss": 0.0, "step": 53042 }, { "epoch": 1.03, "learning_rate": 3.284681921989533e-05, "loss": 0.0, "step": 53044 }, { "epoch": 1.03, "learning_rate": 3.284617225963809e-05, "loss": 0.0, "step": 53046 }, { "epoch": 1.03, "learning_rate": 3.284552529938086e-05, "loss": 0.0002, "step": 53048 }, { "epoch": 1.03, "learning_rate": 3.284520181925225e-05, "loss": 0.0217, "step": 53050 }, { "epoch": 1.03, "learning_rate": 3.284455485899501e-05, "loss": 0.0, "step": 53052 }, { "epoch": 1.03, "learning_rate": 3.284390789873778e-05, "loss": 0.0001, "step": 53054 }, { "epoch": 1.03, "learning_rate": 3.284326093848055e-05, "loss": 0.0001, "step": 53056 }, { "epoch": 1.03, "learning_rate": 3.284261397822332e-05, "loss": 0.0, "step": 53058 }, { "epoch": 1.03, "learning_rate": 3.284196701796609e-05, "loss": 0.011, "step": 53060 }, { "epoch": 1.03, "learning_rate": 3.284132005770886e-05, "loss": 0.0, "step": 53062 }, { "epoch": 1.03, "learning_rate": 3.284067309745163e-05, "loss": 0.0001, "step": 53064 }, { "epoch": 1.03, "learning_rate": 3.2840026137194396e-05, "loss": 0.0, "step": 53066 }, { "epoch": 1.03, "learning_rate": 3.2839379176937165e-05, "loss": 0.0002, "step": 53068 }, { "epoch": 1.03, "learning_rate": 3.283873221667993e-05, "loss": 0.0, "step": 53070 }, { "epoch": 1.03, "learning_rate": 3.28380852564227e-05, "loss": 0.0, "step": 53072 }, { "epoch": 1.03, "learning_rate": 3.2837438296165465e-05, "loss": 0.0001, "step": 53074 }, { "epoch": 1.03, "learning_rate": 3.2836791335908234e-05, "loss": 0.0001, "step": 53076 }, { "epoch": 1.03, "learning_rate": 3.2836144375651004e-05, "loss": 0.0, "step": 53078 }, { "epoch": 1.03, "learning_rate": 3.283549741539377e-05, "loss": 0.0219, "step": 53080 }, { "epoch": 1.03, "learning_rate": 3.283485045513654e-05, "loss": 0.0, "step": 53082 }, { "epoch": 1.03, "learning_rate": 3.283420349487931e-05, "loss": 0.0176, "step": 53084 }, { "epoch": 1.03, "learning_rate": 3.283355653462208e-05, "loss": 0.0001, "step": 53086 }, { "epoch": 1.03, "learning_rate": 3.283290957436485e-05, "loss": 0.0003, "step": 53088 }, { "epoch": 1.03, "learning_rate": 3.283226261410762e-05, "loss": 0.0, "step": 53090 }, { "epoch": 1.03, "learning_rate": 3.283161565385039e-05, "loss": 0.0, "step": 53092 }, { "epoch": 1.03, "learning_rate": 3.2830968693593156e-05, "loss": 0.0001, "step": 53094 }, { "epoch": 1.03, "learning_rate": 3.2830321733335925e-05, "loss": 0.0, "step": 53096 }, { "epoch": 1.03, "learning_rate": 3.282967477307869e-05, "loss": 0.0011, "step": 53098 }, { "epoch": 1.03, "learning_rate": 3.2829027812821464e-05, "loss": 0.0, "step": 53100 }, { "epoch": 1.03, "learning_rate": 3.2828380852564226e-05, "loss": 0.0, "step": 53102 }, { "epoch": 1.03, "learning_rate": 3.282805737243561e-05, "loss": 0.0501, "step": 53104 }, { "epoch": 1.03, "learning_rate": 3.2827410412178386e-05, "loss": 0.0, "step": 53106 }, { "epoch": 1.03, "learning_rate": 3.282676345192115e-05, "loss": 0.0, "step": 53108 }, { "epoch": 1.03, "learning_rate": 3.282611649166392e-05, "loss": 0.0, "step": 53110 }, { "epoch": 1.03, "learning_rate": 3.282546953140669e-05, "loss": 0.002, "step": 53112 }, { "epoch": 1.03, "learning_rate": 3.2824822571149456e-05, "loss": 0.0, "step": 53114 }, { "epoch": 1.03, "learning_rate": 3.2824175610892225e-05, "loss": 0.0, "step": 53116 }, { "epoch": 1.03, "learning_rate": 3.2823528650634994e-05, "loss": 0.0, "step": 53118 }, { "epoch": 1.03, "learning_rate": 3.282288169037776e-05, "loss": 0.0, "step": 53120 }, { "epoch": 1.03, "learning_rate": 3.282223473012053e-05, "loss": 0.0, "step": 53122 }, { "epoch": 1.03, "learning_rate": 3.28215877698633e-05, "loss": 0.0, "step": 53124 }, { "epoch": 1.03, "learning_rate": 3.2820940809606064e-05, "loss": 0.0, "step": 53126 }, { "epoch": 1.03, "learning_rate": 3.282029384934884e-05, "loss": 0.0, "step": 53128 }, { "epoch": 1.03, "learning_rate": 3.28196468890916e-05, "loss": 0.0073, "step": 53130 }, { "epoch": 1.03, "learning_rate": 3.281899992883437e-05, "loss": 0.0001, "step": 53132 }, { "epoch": 1.03, "learning_rate": 3.281835296857714e-05, "loss": 0.0018, "step": 53134 }, { "epoch": 1.03, "learning_rate": 3.281770600831991e-05, "loss": 0.0001, "step": 53136 }, { "epoch": 1.03, "learning_rate": 3.2817059048062685e-05, "loss": 0.0004, "step": 53138 }, { "epoch": 1.03, "learning_rate": 3.281641208780545e-05, "loss": 0.0, "step": 53140 }, { "epoch": 1.03, "learning_rate": 3.2815765127548216e-05, "loss": 0.0, "step": 53142 }, { "epoch": 1.03, "learning_rate": 3.2815118167290985e-05, "loss": 0.0, "step": 53144 }, { "epoch": 1.03, "learning_rate": 3.2814471207033754e-05, "loss": 0.0248, "step": 53146 }, { "epoch": 1.03, "learning_rate": 3.281382424677652e-05, "loss": 0.0001, "step": 53148 }, { "epoch": 1.03, "learning_rate": 3.281317728651929e-05, "loss": 0.0, "step": 53150 }, { "epoch": 1.03, "learning_rate": 3.281253032626206e-05, "loss": 0.0, "step": 53152 }, { "epoch": 1.03, "learning_rate": 3.2811883366004824e-05, "loss": 0.0001, "step": 53154 }, { "epoch": 1.03, "learning_rate": 3.28112364057476e-05, "loss": 0.0, "step": 53156 }, { "epoch": 1.03, "learning_rate": 3.281058944549036e-05, "loss": 0.0, "step": 53158 }, { "epoch": 1.03, "learning_rate": 3.280994248523314e-05, "loss": 0.0, "step": 53160 }, { "epoch": 1.03, "learning_rate": 3.28092955249759e-05, "loss": 0.0113, "step": 53162 }, { "epoch": 1.03, "learning_rate": 3.280864856471867e-05, "loss": 0.0001, "step": 53164 }, { "epoch": 1.03, "learning_rate": 3.280800160446144e-05, "loss": 0.0, "step": 53166 }, { "epoch": 1.03, "learning_rate": 3.280735464420421e-05, "loss": 0.0, "step": 53168 }, { "epoch": 1.03, "learning_rate": 3.2806707683946977e-05, "loss": 0.0, "step": 53170 }, { "epoch": 1.03, "learning_rate": 3.2806060723689746e-05, "loss": 0.0, "step": 53172 }, { "epoch": 1.03, "learning_rate": 3.2805413763432515e-05, "loss": 0.0, "step": 53174 }, { "epoch": 1.03, "learning_rate": 3.2804766803175284e-05, "loss": 0.0, "step": 53176 }, { "epoch": 1.03, "learning_rate": 3.280411984291805e-05, "loss": 0.0, "step": 53178 }, { "epoch": 1.03, "learning_rate": 3.2803472882660815e-05, "loss": 0.0, "step": 53180 }, { "epoch": 1.03, "learning_rate": 3.280282592240359e-05, "loss": 0.0307, "step": 53182 }, { "epoch": 1.03, "learning_rate": 3.280217896214636e-05, "loss": 0.0, "step": 53184 }, { "epoch": 1.03, "learning_rate": 3.280153200188912e-05, "loss": 0.0, "step": 53186 }, { "epoch": 1.03, "learning_rate": 3.28008850416319e-05, "loss": 0.0, "step": 53188 }, { "epoch": 1.03, "learning_rate": 3.280023808137466e-05, "loss": 0.0001, "step": 53190 }, { "epoch": 1.03, "learning_rate": 3.279959112111743e-05, "loss": 0.0029, "step": 53192 }, { "epoch": 1.03, "learning_rate": 3.27989441608602e-05, "loss": 0.0, "step": 53194 }, { "epoch": 1.03, "learning_rate": 3.279829720060297e-05, "loss": 0.0002, "step": 53196 }, { "epoch": 1.03, "learning_rate": 3.279765024034574e-05, "loss": 0.0, "step": 53198 }, { "epoch": 1.03, "learning_rate": 3.2797003280088506e-05, "loss": 0.0, "step": 53200 }, { "epoch": 1.03, "learning_rate": 3.2796356319831275e-05, "loss": 0.0, "step": 53202 }, { "epoch": 1.03, "learning_rate": 3.2795709359574044e-05, "loss": 0.009, "step": 53204 }, { "epoch": 1.03, "learning_rate": 3.279506239931681e-05, "loss": 0.0, "step": 53206 }, { "epoch": 1.03, "learning_rate": 3.2794415439059576e-05, "loss": 0.0, "step": 53208 }, { "epoch": 1.03, "learning_rate": 3.279376847880235e-05, "loss": 0.0, "step": 53210 }, { "epoch": 1.03, "learning_rate": 3.2793121518545114e-05, "loss": 0.0001, "step": 53212 }, { "epoch": 1.03, "learning_rate": 3.279247455828788e-05, "loss": 0.0001, "step": 53214 }, { "epoch": 1.03, "learning_rate": 3.279182759803066e-05, "loss": 0.0004, "step": 53216 }, { "epoch": 1.03, "learning_rate": 3.279118063777342e-05, "loss": 0.0, "step": 53218 }, { "epoch": 1.03, "learning_rate": 3.27905336775162e-05, "loss": 0.0, "step": 53220 }, { "epoch": 1.03, "learning_rate": 3.278988671725896e-05, "loss": 0.0, "step": 53222 }, { "epoch": 1.03, "learning_rate": 3.278923975700173e-05, "loss": 0.0092, "step": 53224 }, { "epoch": 1.03, "learning_rate": 3.27885927967445e-05, "loss": 0.0018, "step": 53226 }, { "epoch": 1.03, "learning_rate": 3.2787945836487266e-05, "loss": 0.0, "step": 53228 }, { "epoch": 1.03, "learning_rate": 3.2787298876230036e-05, "loss": 0.0, "step": 53230 }, { "epoch": 1.03, "learning_rate": 3.2786651915972805e-05, "loss": 0.0001, "step": 53232 }, { "epoch": 1.03, "learning_rate": 3.2786004955715574e-05, "loss": 0.0, "step": 53234 }, { "epoch": 1.03, "learning_rate": 3.2785357995458336e-05, "loss": 0.0, "step": 53236 }, { "epoch": 1.03, "learning_rate": 3.278471103520111e-05, "loss": 0.0, "step": 53238 }, { "epoch": 1.03, "learning_rate": 3.2784064074943874e-05, "loss": 0.0001, "step": 53240 }, { "epoch": 1.03, "learning_rate": 3.278341711468665e-05, "loss": 0.0, "step": 53242 }, { "epoch": 1.03, "learning_rate": 3.278277015442941e-05, "loss": 0.0005, "step": 53244 }, { "epoch": 1.03, "learning_rate": 3.278212319417218e-05, "loss": 0.0, "step": 53246 }, { "epoch": 1.03, "learning_rate": 3.278147623391495e-05, "loss": 0.0, "step": 53248 }, { "epoch": 1.03, "learning_rate": 3.278082927365772e-05, "loss": 0.0002, "step": 53250 }, { "epoch": 1.03, "learning_rate": 3.278018231340049e-05, "loss": 0.0005, "step": 53252 }, { "epoch": 1.03, "learning_rate": 3.277953535314326e-05, "loss": 0.0033, "step": 53254 }, { "epoch": 1.03, "learning_rate": 3.277888839288603e-05, "loss": 0.0001, "step": 53256 }, { "epoch": 1.03, "learning_rate": 3.2778241432628796e-05, "loss": 0.0, "step": 53258 }, { "epoch": 1.03, "learning_rate": 3.2777594472371565e-05, "loss": 0.0, "step": 53260 }, { "epoch": 1.03, "learning_rate": 3.2776947512114334e-05, "loss": 0.0, "step": 53262 }, { "epoch": 1.03, "learning_rate": 3.27763005518571e-05, "loss": 0.0001, "step": 53264 }, { "epoch": 1.03, "learning_rate": 3.277565359159987e-05, "loss": 0.0, "step": 53266 }, { "epoch": 1.03, "learning_rate": 3.2775006631342635e-05, "loss": 0.0388, "step": 53268 }, { "epoch": 1.03, "learning_rate": 3.277435967108541e-05, "loss": 0.0001, "step": 53270 }, { "epoch": 1.03, "learning_rate": 3.277371271082817e-05, "loss": 0.0, "step": 53272 }, { "epoch": 1.03, "learning_rate": 3.277306575057094e-05, "loss": 0.0, "step": 53274 }, { "epoch": 1.03, "learning_rate": 3.277241879031371e-05, "loss": 0.0, "step": 53276 }, { "epoch": 1.03, "learning_rate": 3.277177183005648e-05, "loss": 0.0008, "step": 53278 }, { "epoch": 1.03, "learning_rate": 3.277112486979925e-05, "loss": 0.0, "step": 53280 }, { "epoch": 1.03, "learning_rate": 3.277047790954202e-05, "loss": 0.0, "step": 53282 }, { "epoch": 1.03, "learning_rate": 3.276983094928479e-05, "loss": 0.0, "step": 53284 }, { "epoch": 1.03, "learning_rate": 3.2769183989027556e-05, "loss": 0.0032, "step": 53286 }, { "epoch": 1.03, "learning_rate": 3.2768537028770325e-05, "loss": 0.0001, "step": 53288 }, { "epoch": 1.03, "learning_rate": 3.276789006851309e-05, "loss": 0.0, "step": 53290 }, { "epoch": 1.03, "learning_rate": 3.2767243108255864e-05, "loss": 0.0, "step": 53292 }, { "epoch": 1.03, "learning_rate": 3.276659614799863e-05, "loss": 0.0001, "step": 53294 }, { "epoch": 1.03, "learning_rate": 3.2765949187741395e-05, "loss": 0.0, "step": 53296 }, { "epoch": 1.03, "learning_rate": 3.276530222748417e-05, "loss": 0.0, "step": 53298 }, { "epoch": 1.03, "learning_rate": 3.276465526722693e-05, "loss": 0.0004, "step": 53300 }, { "epoch": 1.03, "learning_rate": 3.276400830696971e-05, "loss": 0.0, "step": 53302 }, { "epoch": 1.03, "learning_rate": 3.276336134671247e-05, "loss": 0.0, "step": 53304 }, { "epoch": 1.03, "learning_rate": 3.276271438645524e-05, "loss": 0.0, "step": 53306 }, { "epoch": 1.03, "learning_rate": 3.276206742619801e-05, "loss": 0.0, "step": 53308 }, { "epoch": 1.03, "learning_rate": 3.276142046594078e-05, "loss": 0.0, "step": 53310 }, { "epoch": 1.03, "learning_rate": 3.276077350568355e-05, "loss": 0.0, "step": 53312 }, { "epoch": 1.03, "learning_rate": 3.276012654542632e-05, "loss": 0.0, "step": 53314 }, { "epoch": 1.03, "learning_rate": 3.2759479585169086e-05, "loss": 0.0, "step": 53316 }, { "epoch": 1.03, "learning_rate": 3.2758832624911855e-05, "loss": 0.0, "step": 53318 }, { "epoch": 1.03, "learning_rate": 3.2758185664654624e-05, "loss": 0.0, "step": 53320 }, { "epoch": 1.03, "learning_rate": 3.2757538704397386e-05, "loss": 0.0064, "step": 53322 }, { "epoch": 1.03, "learning_rate": 3.275689174414016e-05, "loss": 0.0, "step": 53324 }, { "epoch": 1.03, "learning_rate": 3.2756244783882925e-05, "loss": 0.0021, "step": 53326 }, { "epoch": 1.04, "learning_rate": 3.2755597823625694e-05, "loss": 0.0, "step": 53328 }, { "epoch": 1.04, "learning_rate": 3.275495086336847e-05, "loss": 0.0108, "step": 53330 }, { "epoch": 1.04, "learning_rate": 3.275430390311123e-05, "loss": 0.0, "step": 53332 }, { "epoch": 1.04, "learning_rate": 3.2753656942854e-05, "loss": 0.0, "step": 53334 }, { "epoch": 1.04, "learning_rate": 3.275300998259677e-05, "loss": 0.0012, "step": 53336 }, { "epoch": 1.04, "learning_rate": 3.275236302233954e-05, "loss": 0.0, "step": 53338 }, { "epoch": 1.04, "learning_rate": 3.275171606208231e-05, "loss": 0.0077, "step": 53340 }, { "epoch": 1.04, "learning_rate": 3.275106910182508e-05, "loss": 0.0, "step": 53342 }, { "epoch": 1.04, "learning_rate": 3.2750422141567846e-05, "loss": 0.0, "step": 53344 }, { "epoch": 1.04, "learning_rate": 3.2749775181310615e-05, "loss": 0.0, "step": 53346 }, { "epoch": 1.04, "learning_rate": 3.2749128221053384e-05, "loss": 0.042, "step": 53348 }, { "epoch": 1.04, "learning_rate": 3.274848126079615e-05, "loss": 0.0, "step": 53350 }, { "epoch": 1.04, "learning_rate": 3.274783430053892e-05, "loss": 0.0021, "step": 53352 }, { "epoch": 1.04, "learning_rate": 3.2747187340281685e-05, "loss": 0.0, "step": 53354 }, { "epoch": 1.04, "learning_rate": 3.2746540380024454e-05, "loss": 0.0008, "step": 53356 }, { "epoch": 1.04, "learning_rate": 3.274589341976722e-05, "loss": 0.0, "step": 53358 }, { "epoch": 1.04, "learning_rate": 3.274524645950999e-05, "loss": 0.0, "step": 53360 }, { "epoch": 1.04, "learning_rate": 3.274459949925277e-05, "loss": 0.0, "step": 53362 }, { "epoch": 1.04, "learning_rate": 3.274395253899553e-05, "loss": 0.0, "step": 53364 }, { "epoch": 1.04, "learning_rate": 3.27433055787383e-05, "loss": 0.0, "step": 53366 }, { "epoch": 1.04, "learning_rate": 3.274265861848107e-05, "loss": 0.0, "step": 53368 }, { "epoch": 1.04, "learning_rate": 3.274201165822384e-05, "loss": 0.0, "step": 53370 }, { "epoch": 1.04, "learning_rate": 3.274136469796661e-05, "loss": 0.0, "step": 53372 }, { "epoch": 1.04, "learning_rate": 3.2740717737709376e-05, "loss": 0.0, "step": 53374 }, { "epoch": 1.04, "learning_rate": 3.2740070777452145e-05, "loss": 0.0003, "step": 53376 }, { "epoch": 1.04, "learning_rate": 3.273942381719491e-05, "loss": 0.0001, "step": 53378 }, { "epoch": 1.04, "learning_rate": 3.273877685693768e-05, "loss": 0.0002, "step": 53380 }, { "epoch": 1.04, "learning_rate": 3.2738129896680445e-05, "loss": 0.0, "step": 53382 }, { "epoch": 1.04, "learning_rate": 3.273748293642322e-05, "loss": 0.0, "step": 53384 }, { "epoch": 1.04, "learning_rate": 3.2736835976165983e-05, "loss": 0.0, "step": 53386 }, { "epoch": 1.04, "learning_rate": 3.273618901590875e-05, "loss": 0.0, "step": 53388 }, { "epoch": 1.04, "learning_rate": 3.273554205565152e-05, "loss": 0.0, "step": 53390 }, { "epoch": 1.04, "learning_rate": 3.273489509539429e-05, "loss": 0.0, "step": 53392 }, { "epoch": 1.04, "learning_rate": 3.273424813513706e-05, "loss": 0.0002, "step": 53394 }, { "epoch": 1.04, "learning_rate": 3.273360117487983e-05, "loss": 0.0, "step": 53396 }, { "epoch": 1.04, "learning_rate": 3.27329542146226e-05, "loss": 0.0041, "step": 53398 }, { "epoch": 1.04, "learning_rate": 3.273230725436537e-05, "loss": 0.0, "step": 53400 }, { "epoch": 1.04, "learning_rate": 3.2731660294108136e-05, "loss": 0.0, "step": 53402 }, { "epoch": 1.04, "learning_rate": 3.27310133338509e-05, "loss": 0.0002, "step": 53404 }, { "epoch": 1.04, "learning_rate": 3.2730366373593674e-05, "loss": 0.0, "step": 53406 }, { "epoch": 1.04, "learning_rate": 3.2729719413336443e-05, "loss": 0.0346, "step": 53408 }, { "epoch": 1.04, "learning_rate": 3.2729072453079206e-05, "loss": 0.0001, "step": 53410 }, { "epoch": 1.04, "learning_rate": 3.272842549282198e-05, "loss": 0.0, "step": 53412 }, { "epoch": 1.04, "learning_rate": 3.2727778532564744e-05, "loss": 0.0, "step": 53414 }, { "epoch": 1.04, "learning_rate": 3.272713157230751e-05, "loss": 0.0, "step": 53416 }, { "epoch": 1.04, "learning_rate": 3.272648461205028e-05, "loss": 0.0042, "step": 53418 }, { "epoch": 1.04, "learning_rate": 3.272583765179305e-05, "loss": 0.0, "step": 53420 }, { "epoch": 1.04, "learning_rate": 3.272519069153582e-05, "loss": 0.0, "step": 53422 }, { "epoch": 1.04, "learning_rate": 3.272454373127859e-05, "loss": 0.0, "step": 53424 }, { "epoch": 1.04, "learning_rate": 3.272389677102136e-05, "loss": 0.0, "step": 53426 }, { "epoch": 1.04, "learning_rate": 3.272324981076413e-05, "loss": 0.0, "step": 53428 }, { "epoch": 1.04, "learning_rate": 3.2722602850506897e-05, "loss": 0.0, "step": 53430 }, { "epoch": 1.04, "learning_rate": 3.272195589024966e-05, "loss": 0.0, "step": 53432 }, { "epoch": 1.04, "learning_rate": 3.2721308929992435e-05, "loss": 0.0, "step": 53434 }, { "epoch": 1.04, "learning_rate": 3.27206619697352e-05, "loss": 0.0, "step": 53436 }, { "epoch": 1.04, "learning_rate": 3.2720015009477966e-05, "loss": 0.0002, "step": 53438 }, { "epoch": 1.04, "learning_rate": 3.271936804922074e-05, "loss": 0.0004, "step": 53440 }, { "epoch": 1.04, "learning_rate": 3.2718721088963504e-05, "loss": 0.0, "step": 53442 }, { "epoch": 1.04, "learning_rate": 3.271807412870628e-05, "loss": 0.0001, "step": 53444 }, { "epoch": 1.04, "learning_rate": 3.271742716844904e-05, "loss": 0.0, "step": 53446 }, { "epoch": 1.04, "learning_rate": 3.271678020819181e-05, "loss": 0.0, "step": 53448 }, { "epoch": 1.04, "learning_rate": 3.271613324793458e-05, "loss": 0.0, "step": 53450 }, { "epoch": 1.04, "learning_rate": 3.271548628767735e-05, "loss": 0.0, "step": 53452 }, { "epoch": 1.04, "learning_rate": 3.271483932742012e-05, "loss": 0.0, "step": 53454 }, { "epoch": 1.04, "learning_rate": 3.271419236716289e-05, "loss": 0.0022, "step": 53456 }, { "epoch": 1.04, "learning_rate": 3.271354540690566e-05, "loss": 0.0001, "step": 53458 }, { "epoch": 1.04, "learning_rate": 3.271289844664842e-05, "loss": 0.0001, "step": 53460 }, { "epoch": 1.04, "learning_rate": 3.2712251486391195e-05, "loss": 0.0013, "step": 53462 }, { "epoch": 1.04, "learning_rate": 3.271160452613396e-05, "loss": 0.0001, "step": 53464 }, { "epoch": 1.04, "learning_rate": 3.271095756587673e-05, "loss": 0.0, "step": 53466 }, { "epoch": 1.04, "learning_rate": 3.2710310605619496e-05, "loss": 0.0, "step": 53468 }, { "epoch": 1.04, "learning_rate": 3.2709663645362265e-05, "loss": 0.0, "step": 53470 }, { "epoch": 1.04, "learning_rate": 3.270901668510504e-05, "loss": 0.0, "step": 53472 }, { "epoch": 1.04, "learning_rate": 3.27083697248478e-05, "loss": 0.0, "step": 53474 }, { "epoch": 1.04, "learning_rate": 3.270772276459057e-05, "loss": 0.0, "step": 53476 }, { "epoch": 1.04, "learning_rate": 3.270707580433334e-05, "loss": 0.0046, "step": 53478 }, { "epoch": 1.04, "learning_rate": 3.270642884407611e-05, "loss": 0.0, "step": 53480 }, { "epoch": 1.04, "learning_rate": 3.270578188381888e-05, "loss": 0.0, "step": 53482 }, { "epoch": 1.04, "learning_rate": 3.270513492356165e-05, "loss": 0.0003, "step": 53484 }, { "epoch": 1.04, "learning_rate": 3.270448796330442e-05, "loss": 0.0, "step": 53486 }, { "epoch": 1.04, "learning_rate": 3.2703841003047186e-05, "loss": 0.0, "step": 53488 }, { "epoch": 1.04, "learning_rate": 3.2703194042789956e-05, "loss": 0.001, "step": 53490 }, { "epoch": 1.04, "learning_rate": 3.270254708253272e-05, "loss": 0.0217, "step": 53492 }, { "epoch": 1.04, "learning_rate": 3.2701900122275494e-05, "loss": 0.0002, "step": 53494 }, { "epoch": 1.04, "learning_rate": 3.2701253162018256e-05, "loss": 0.0052, "step": 53496 }, { "epoch": 1.04, "learning_rate": 3.2700606201761025e-05, "loss": 0.0001, "step": 53498 }, { "epoch": 1.04, "learning_rate": 3.2699959241503794e-05, "loss": 0.0, "step": 53500 }, { "epoch": 1.04, "learning_rate": 3.269931228124656e-05, "loss": 0.0082, "step": 53502 }, { "epoch": 1.04, "learning_rate": 3.269866532098933e-05, "loss": 0.0005, "step": 53504 }, { "epoch": 1.04, "learning_rate": 3.26980183607321e-05, "loss": 0.0001, "step": 53506 }, { "epoch": 1.04, "learning_rate": 3.269737140047487e-05, "loss": 0.0002, "step": 53508 }, { "epoch": 1.04, "learning_rate": 3.269672444021764e-05, "loss": 0.0001, "step": 53510 }, { "epoch": 1.04, "learning_rate": 3.269607747996041e-05, "loss": 0.0, "step": 53512 }, { "epoch": 1.04, "learning_rate": 3.269543051970317e-05, "loss": 0.0001, "step": 53514 }, { "epoch": 1.04, "learning_rate": 3.269478355944595e-05, "loss": 0.0, "step": 53516 }, { "epoch": 1.04, "learning_rate": 3.2694136599188716e-05, "loss": 0.0056, "step": 53518 }, { "epoch": 1.04, "learning_rate": 3.269348963893148e-05, "loss": 0.0001, "step": 53520 }, { "epoch": 1.04, "learning_rate": 3.2692842678674254e-05, "loss": 0.0, "step": 53522 }, { "epoch": 1.04, "learning_rate": 3.2692195718417016e-05, "loss": 0.0, "step": 53524 }, { "epoch": 1.04, "learning_rate": 3.269154875815979e-05, "loss": 0.0002, "step": 53526 }, { "epoch": 1.04, "learning_rate": 3.2690901797902555e-05, "loss": 0.0, "step": 53528 }, { "epoch": 1.04, "learning_rate": 3.2690254837645324e-05, "loss": 0.0025, "step": 53530 }, { "epoch": 1.04, "learning_rate": 3.268960787738809e-05, "loss": 0.0, "step": 53532 }, { "epoch": 1.04, "learning_rate": 3.268896091713086e-05, "loss": 0.0, "step": 53534 }, { "epoch": 1.04, "learning_rate": 3.268831395687363e-05, "loss": 0.0, "step": 53536 }, { "epoch": 1.04, "learning_rate": 3.26876669966164e-05, "loss": 0.0, "step": 53538 }, { "epoch": 1.04, "learning_rate": 3.268702003635917e-05, "loss": 0.0, "step": 53540 }, { "epoch": 1.04, "learning_rate": 3.268637307610194e-05, "loss": 0.0, "step": 53542 }, { "epoch": 1.04, "learning_rate": 3.268572611584471e-05, "loss": 0.0, "step": 53544 }, { "epoch": 1.04, "learning_rate": 3.268507915558747e-05, "loss": 0.0, "step": 53546 }, { "epoch": 1.04, "learning_rate": 3.2684432195330245e-05, "loss": 0.0, "step": 53548 }, { "epoch": 1.04, "learning_rate": 3.2683785235073015e-05, "loss": 0.0, "step": 53550 }, { "epoch": 1.04, "learning_rate": 3.268313827481578e-05, "loss": 0.0037, "step": 53552 }, { "epoch": 1.04, "learning_rate": 3.268249131455855e-05, "loss": 0.0, "step": 53554 }, { "epoch": 1.04, "learning_rate": 3.2681844354301315e-05, "loss": 0.0, "step": 53556 }, { "epoch": 1.04, "learning_rate": 3.2681197394044084e-05, "loss": 0.0, "step": 53558 }, { "epoch": 1.04, "learning_rate": 3.268055043378685e-05, "loss": 0.0, "step": 53560 }, { "epoch": 1.04, "learning_rate": 3.267990347352962e-05, "loss": 0.0, "step": 53562 }, { "epoch": 1.04, "learning_rate": 3.267925651327239e-05, "loss": 0.0, "step": 53564 }, { "epoch": 1.04, "learning_rate": 3.267860955301516e-05, "loss": 0.0018, "step": 53566 }, { "epoch": 1.04, "learning_rate": 3.267796259275793e-05, "loss": 0.0, "step": 53568 }, { "epoch": 1.04, "learning_rate": 3.26773156325007e-05, "loss": 0.0, "step": 53570 }, { "epoch": 1.04, "learning_rate": 3.267666867224347e-05, "loss": 0.0368, "step": 53572 }, { "epoch": 1.04, "learning_rate": 3.267602171198623e-05, "loss": 0.0, "step": 53574 }, { "epoch": 1.04, "learning_rate": 3.2675374751729006e-05, "loss": 0.0, "step": 53576 }, { "epoch": 1.04, "learning_rate": 3.267472779147177e-05, "loss": 0.0106, "step": 53578 }, { "epoch": 1.04, "learning_rate": 3.267408083121454e-05, "loss": 0.0001, "step": 53580 }, { "epoch": 1.04, "learning_rate": 3.2673433870957306e-05, "loss": 0.0002, "step": 53582 }, { "epoch": 1.04, "learning_rate": 3.2672786910700075e-05, "loss": 0.0003, "step": 53584 }, { "epoch": 1.04, "learning_rate": 3.267213995044285e-05, "loss": 0.0, "step": 53586 }, { "epoch": 1.04, "learning_rate": 3.2671492990185614e-05, "loss": 0.0, "step": 53588 }, { "epoch": 1.04, "learning_rate": 3.267084602992838e-05, "loss": 0.0094, "step": 53590 }, { "epoch": 1.04, "learning_rate": 3.267019906967115e-05, "loss": 0.0, "step": 53592 }, { "epoch": 1.04, "learning_rate": 3.266955210941392e-05, "loss": 0.0, "step": 53594 }, { "epoch": 1.04, "learning_rate": 3.266890514915669e-05, "loss": 0.0, "step": 53596 }, { "epoch": 1.04, "learning_rate": 3.266825818889946e-05, "loss": 0.0001, "step": 53598 }, { "epoch": 1.04, "learning_rate": 3.266761122864223e-05, "loss": 0.0, "step": 53600 }, { "epoch": 1.04, "learning_rate": 3.266696426838499e-05, "loss": 0.0, "step": 53602 }, { "epoch": 1.04, "learning_rate": 3.2666317308127766e-05, "loss": 0.0, "step": 53604 }, { "epoch": 1.04, "learning_rate": 3.266567034787053e-05, "loss": 0.0, "step": 53606 }, { "epoch": 1.04, "learning_rate": 3.2665023387613304e-05, "loss": 0.0001, "step": 53608 }, { "epoch": 1.04, "learning_rate": 3.266437642735607e-05, "loss": 0.0, "step": 53610 }, { "epoch": 1.04, "learning_rate": 3.2663729467098836e-05, "loss": 0.0, "step": 53612 }, { "epoch": 1.04, "learning_rate": 3.2663082506841605e-05, "loss": 0.0001, "step": 53614 }, { "epoch": 1.04, "learning_rate": 3.2662435546584374e-05, "loss": 0.0002, "step": 53616 }, { "epoch": 1.04, "learning_rate": 3.266178858632714e-05, "loss": 0.0021, "step": 53618 }, { "epoch": 1.04, "learning_rate": 3.266114162606991e-05, "loss": 0.0, "step": 53620 }, { "epoch": 1.04, "learning_rate": 3.266049466581268e-05, "loss": 0.0, "step": 53622 }, { "epoch": 1.04, "learning_rate": 3.265984770555545e-05, "loss": 0.0, "step": 53624 }, { "epoch": 1.04, "learning_rate": 3.265920074529822e-05, "loss": 0.0, "step": 53626 }, { "epoch": 1.04, "learning_rate": 3.265855378504098e-05, "loss": 0.0, "step": 53628 }, { "epoch": 1.04, "learning_rate": 3.265790682478376e-05, "loss": 0.0, "step": 53630 }, { "epoch": 1.04, "learning_rate": 3.265725986452653e-05, "loss": 0.0001, "step": 53632 }, { "epoch": 1.04, "learning_rate": 3.265661290426929e-05, "loss": 0.0003, "step": 53634 }, { "epoch": 1.04, "learning_rate": 3.2655965944012065e-05, "loss": 0.0001, "step": 53636 }, { "epoch": 1.04, "learning_rate": 3.265531898375483e-05, "loss": 0.0001, "step": 53638 }, { "epoch": 1.04, "learning_rate": 3.2654672023497596e-05, "loss": 0.0, "step": 53640 }, { "epoch": 1.04, "learning_rate": 3.2654025063240365e-05, "loss": 0.0, "step": 53642 }, { "epoch": 1.04, "learning_rate": 3.2653378102983134e-05, "loss": 0.0, "step": 53644 }, { "epoch": 1.04, "learning_rate": 3.2652731142725903e-05, "loss": 0.0001, "step": 53646 }, { "epoch": 1.04, "learning_rate": 3.265208418246867e-05, "loss": 0.0, "step": 53648 }, { "epoch": 1.04, "learning_rate": 3.265143722221144e-05, "loss": 0.0007, "step": 53650 }, { "epoch": 1.04, "learning_rate": 3.265079026195421e-05, "loss": 0.0, "step": 53652 }, { "epoch": 1.04, "learning_rate": 3.265014330169698e-05, "loss": 0.0015, "step": 53654 }, { "epoch": 1.04, "learning_rate": 3.264949634143974e-05, "loss": 0.0, "step": 53656 }, { "epoch": 1.04, "learning_rate": 3.264884938118252e-05, "loss": 0.0, "step": 53658 }, { "epoch": 1.04, "learning_rate": 3.264820242092528e-05, "loss": 0.0, "step": 53660 }, { "epoch": 1.04, "learning_rate": 3.264755546066805e-05, "loss": 0.0, "step": 53662 }, { "epoch": 1.04, "learning_rate": 3.2646908500410825e-05, "loss": 0.0, "step": 53664 }, { "epoch": 1.04, "learning_rate": 3.264626154015359e-05, "loss": 0.0, "step": 53666 }, { "epoch": 1.04, "learning_rate": 3.2645614579896363e-05, "loss": 0.0001, "step": 53668 }, { "epoch": 1.04, "learning_rate": 3.2644967619639126e-05, "loss": 0.0001, "step": 53670 }, { "epoch": 1.04, "learning_rate": 3.2644320659381895e-05, "loss": 0.0, "step": 53672 }, { "epoch": 1.04, "learning_rate": 3.2643673699124664e-05, "loss": 0.0, "step": 53674 }, { "epoch": 1.04, "learning_rate": 3.264302673886743e-05, "loss": 0.0002, "step": 53676 }, { "epoch": 1.04, "learning_rate": 3.26423797786102e-05, "loss": 0.0002, "step": 53678 }, { "epoch": 1.04, "learning_rate": 3.264173281835297e-05, "loss": 0.0, "step": 53680 }, { "epoch": 1.04, "learning_rate": 3.264108585809574e-05, "loss": 0.0, "step": 53682 }, { "epoch": 1.04, "learning_rate": 3.264043889783851e-05, "loss": 0.0, "step": 53684 }, { "epoch": 1.04, "learning_rate": 3.263979193758128e-05, "loss": 0.0, "step": 53686 }, { "epoch": 1.04, "learning_rate": 3.263914497732404e-05, "loss": 0.0001, "step": 53688 }, { "epoch": 1.04, "learning_rate": 3.2638498017066817e-05, "loss": 0.0, "step": 53690 }, { "epoch": 1.04, "learning_rate": 3.263785105680958e-05, "loss": 0.0003, "step": 53692 }, { "epoch": 1.04, "learning_rate": 3.263720409655235e-05, "loss": 0.0, "step": 53694 }, { "epoch": 1.04, "learning_rate": 3.2636557136295124e-05, "loss": 0.0, "step": 53696 }, { "epoch": 1.04, "learning_rate": 3.2635910176037886e-05, "loss": 0.0, "step": 53698 }, { "epoch": 1.04, "learning_rate": 3.2635263215780655e-05, "loss": 0.0, "step": 53700 }, { "epoch": 1.04, "learning_rate": 3.2634616255523424e-05, "loss": 0.0001, "step": 53702 }, { "epoch": 1.04, "learning_rate": 3.263396929526619e-05, "loss": 0.01, "step": 53704 }, { "epoch": 1.04, "learning_rate": 3.263332233500896e-05, "loss": 0.0, "step": 53706 }, { "epoch": 1.04, "learning_rate": 3.263267537475173e-05, "loss": 0.0, "step": 53708 }, { "epoch": 1.04, "learning_rate": 3.26320284144945e-05, "loss": 0.0004, "step": 53710 }, { "epoch": 1.04, "learning_rate": 3.263138145423727e-05, "loss": 0.0, "step": 53712 }, { "epoch": 1.04, "learning_rate": 3.263073449398004e-05, "loss": 0.0, "step": 53714 }, { "epoch": 1.04, "learning_rate": 3.26300875337228e-05, "loss": 0.0, "step": 53716 }, { "epoch": 1.04, "learning_rate": 3.262944057346558e-05, "loss": 0.0, "step": 53718 }, { "epoch": 1.04, "learning_rate": 3.262879361320834e-05, "loss": 0.0, "step": 53720 }, { "epoch": 1.04, "learning_rate": 3.262814665295111e-05, "loss": 0.0005, "step": 53722 }, { "epoch": 1.04, "learning_rate": 3.262749969269388e-05, "loss": 0.0162, "step": 53724 }, { "epoch": 1.04, "learning_rate": 3.2626852732436647e-05, "loss": 0.0, "step": 53726 }, { "epoch": 1.04, "learning_rate": 3.2626205772179416e-05, "loss": 0.0, "step": 53728 }, { "epoch": 1.04, "learning_rate": 3.2625558811922185e-05, "loss": 0.003, "step": 53730 }, { "epoch": 1.04, "learning_rate": 3.2624911851664954e-05, "loss": 0.0, "step": 53732 }, { "epoch": 1.04, "learning_rate": 3.262426489140772e-05, "loss": 0.0002, "step": 53734 }, { "epoch": 1.04, "learning_rate": 3.262361793115049e-05, "loss": 0.0001, "step": 53736 }, { "epoch": 1.04, "learning_rate": 3.2622970970893254e-05, "loss": 0.0001, "step": 53738 }, { "epoch": 1.04, "learning_rate": 3.262232401063603e-05, "loss": 0.0, "step": 53740 }, { "epoch": 1.04, "learning_rate": 3.26216770503788e-05, "loss": 0.0, "step": 53742 }, { "epoch": 1.04, "learning_rate": 3.262103009012156e-05, "loss": 0.0, "step": 53744 }, { "epoch": 1.04, "learning_rate": 3.262038312986434e-05, "loss": 0.0, "step": 53746 }, { "epoch": 1.04, "learning_rate": 3.26197361696071e-05, "loss": 0.0, "step": 53748 }, { "epoch": 1.04, "learning_rate": 3.2619089209349876e-05, "loss": 0.0, "step": 53750 }, { "epoch": 1.04, "learning_rate": 3.261844224909264e-05, "loss": 0.0, "step": 53752 }, { "epoch": 1.04, "learning_rate": 3.261779528883541e-05, "loss": 0.0, "step": 53754 }, { "epoch": 1.04, "learning_rate": 3.2617148328578176e-05, "loss": 0.0, "step": 53756 }, { "epoch": 1.04, "learning_rate": 3.2616501368320945e-05, "loss": 0.0011, "step": 53758 }, { "epoch": 1.04, "learning_rate": 3.2615854408063714e-05, "loss": 0.0242, "step": 53760 }, { "epoch": 1.04, "learning_rate": 3.261520744780648e-05, "loss": 0.0, "step": 53762 }, { "epoch": 1.04, "learning_rate": 3.261456048754925e-05, "loss": 0.008, "step": 53764 }, { "epoch": 1.04, "learning_rate": 3.261391352729202e-05, "loss": 0.0, "step": 53766 }, { "epoch": 1.04, "learning_rate": 3.261326656703479e-05, "loss": 0.0001, "step": 53768 }, { "epoch": 1.04, "learning_rate": 3.261261960677755e-05, "loss": 0.0, "step": 53770 }, { "epoch": 1.04, "learning_rate": 3.261197264652033e-05, "loss": 0.0015, "step": 53772 }, { "epoch": 1.04, "learning_rate": 3.26113256862631e-05, "loss": 0.0083, "step": 53774 }, { "epoch": 1.04, "learning_rate": 3.261067872600586e-05, "loss": 0.0, "step": 53776 }, { "epoch": 1.04, "learning_rate": 3.2610031765748636e-05, "loss": 0.0, "step": 53778 }, { "epoch": 1.04, "learning_rate": 3.26093848054914e-05, "loss": 0.0001, "step": 53780 }, { "epoch": 1.04, "learning_rate": 3.260873784523417e-05, "loss": 0.0, "step": 53782 }, { "epoch": 1.04, "learning_rate": 3.2608090884976936e-05, "loss": 0.0009, "step": 53784 }, { "epoch": 1.04, "learning_rate": 3.2607443924719705e-05, "loss": 0.0003, "step": 53786 }, { "epoch": 1.04, "learning_rate": 3.2606796964462475e-05, "loss": 0.0012, "step": 53788 }, { "epoch": 1.04, "learning_rate": 3.2606150004205244e-05, "loss": 0.0, "step": 53790 }, { "epoch": 1.04, "learning_rate": 3.260550304394801e-05, "loss": 0.0, "step": 53792 }, { "epoch": 1.04, "learning_rate": 3.260485608369078e-05, "loss": 0.0, "step": 53794 }, { "epoch": 1.04, "learning_rate": 3.260420912343355e-05, "loss": 0.0, "step": 53796 }, { "epoch": 1.04, "learning_rate": 3.260356216317631e-05, "loss": 0.0, "step": 53798 }, { "epoch": 1.04, "learning_rate": 3.260291520291909e-05, "loss": 0.0009, "step": 53800 }, { "epoch": 1.04, "learning_rate": 3.260226824266185e-05, "loss": 0.0, "step": 53802 }, { "epoch": 1.04, "learning_rate": 3.260162128240462e-05, "loss": 0.0, "step": 53804 }, { "epoch": 1.04, "learning_rate": 3.260097432214739e-05, "loss": 0.0001, "step": 53806 }, { "epoch": 1.04, "learning_rate": 3.260032736189016e-05, "loss": 0.0, "step": 53808 }, { "epoch": 1.04, "learning_rate": 3.2599680401632935e-05, "loss": 0.0, "step": 53810 }, { "epoch": 1.04, "learning_rate": 3.25990334413757e-05, "loss": 0.0, "step": 53812 }, { "epoch": 1.04, "learning_rate": 3.2598386481118466e-05, "loss": 0.0086, "step": 53814 }, { "epoch": 1.04, "learning_rate": 3.2597739520861235e-05, "loss": 0.0003, "step": 53816 }, { "epoch": 1.04, "learning_rate": 3.2597092560604004e-05, "loss": 0.0002, "step": 53818 }, { "epoch": 1.04, "learning_rate": 3.259644560034677e-05, "loss": 0.0001, "step": 53820 }, { "epoch": 1.04, "learning_rate": 3.259579864008954e-05, "loss": 0.0, "step": 53822 }, { "epoch": 1.04, "learning_rate": 3.259515167983231e-05, "loss": 0.0382, "step": 53824 }, { "epoch": 1.04, "learning_rate": 3.259450471957508e-05, "loss": 0.0, "step": 53826 }, { "epoch": 1.04, "learning_rate": 3.259385775931785e-05, "loss": 0.0, "step": 53828 }, { "epoch": 1.04, "learning_rate": 3.259321079906061e-05, "loss": 0.0, "step": 53830 }, { "epoch": 1.04, "learning_rate": 3.259256383880339e-05, "loss": 0.0, "step": 53832 }, { "epoch": 1.04, "learning_rate": 3.259191687854615e-05, "loss": 0.0, "step": 53834 }, { "epoch": 1.04, "learning_rate": 3.259126991828892e-05, "loss": 0.001, "step": 53836 }, { "epoch": 1.04, "learning_rate": 3.259062295803169e-05, "loss": 0.0005, "step": 53838 }, { "epoch": 1.04, "learning_rate": 3.258997599777446e-05, "loss": 0.0068, "step": 53840 }, { "epoch": 1.05, "learning_rate": 3.2589329037517226e-05, "loss": 0.0, "step": 53842 }, { "epoch": 1.05, "learning_rate": 3.2588682077259995e-05, "loss": 0.0, "step": 53844 }, { "epoch": 1.05, "learning_rate": 3.2588035117002764e-05, "loss": 0.0, "step": 53846 }, { "epoch": 1.05, "learning_rate": 3.2587388156745534e-05, "loss": 0.0001, "step": 53848 }, { "epoch": 1.05, "learning_rate": 3.25867411964883e-05, "loss": 0.0, "step": 53850 }, { "epoch": 1.05, "learning_rate": 3.258609423623107e-05, "loss": 0.0, "step": 53852 }, { "epoch": 1.05, "learning_rate": 3.258544727597384e-05, "loss": 0.0, "step": 53854 }, { "epoch": 1.05, "learning_rate": 3.258480031571661e-05, "loss": 0.0311, "step": 53856 }, { "epoch": 1.05, "learning_rate": 3.258415335545937e-05, "loss": 0.0002, "step": 53858 }, { "epoch": 1.05, "learning_rate": 3.258350639520215e-05, "loss": 0.0, "step": 53860 }, { "epoch": 1.05, "learning_rate": 3.258285943494491e-05, "loss": 0.0, "step": 53862 }, { "epoch": 1.05, "learning_rate": 3.258221247468768e-05, "loss": 0.0, "step": 53864 }, { "epoch": 1.05, "learning_rate": 3.258156551443045e-05, "loss": 0.0, "step": 53866 }, { "epoch": 1.05, "learning_rate": 3.258091855417322e-05, "loss": 0.0018, "step": 53868 }, { "epoch": 1.05, "learning_rate": 3.258027159391599e-05, "loss": 0.0085, "step": 53870 }, { "epoch": 1.05, "learning_rate": 3.2579624633658756e-05, "loss": 0.0, "step": 53872 }, { "epoch": 1.05, "learning_rate": 3.2578977673401525e-05, "loss": 0.0011, "step": 53874 }, { "epoch": 1.05, "learning_rate": 3.2578330713144294e-05, "loss": 0.0, "step": 53876 }, { "epoch": 1.05, "learning_rate": 3.257768375288706e-05, "loss": 0.0, "step": 53878 }, { "epoch": 1.05, "learning_rate": 3.2577036792629825e-05, "loss": 0.0217, "step": 53880 }, { "epoch": 1.05, "learning_rate": 3.25763898323726e-05, "loss": 0.0001, "step": 53882 }, { "epoch": 1.05, "learning_rate": 3.2575742872115364e-05, "loss": 0.0, "step": 53884 }, { "epoch": 1.05, "learning_rate": 3.257509591185813e-05, "loss": 0.0003, "step": 53886 }, { "epoch": 1.05, "learning_rate": 3.257444895160091e-05, "loss": 0.0, "step": 53888 }, { "epoch": 1.05, "learning_rate": 3.257380199134367e-05, "loss": 0.0, "step": 53890 }, { "epoch": 1.05, "learning_rate": 3.257315503108645e-05, "loss": 0.0001, "step": 53892 }, { "epoch": 1.05, "learning_rate": 3.257250807082921e-05, "loss": 0.0022, "step": 53894 }, { "epoch": 1.05, "learning_rate": 3.257186111057198e-05, "loss": 0.0, "step": 53896 }, { "epoch": 1.05, "learning_rate": 3.257121415031475e-05, "loss": 0.0, "step": 53898 }, { "epoch": 1.05, "learning_rate": 3.2570567190057516e-05, "loss": 0.0001, "step": 53900 }, { "epoch": 1.05, "learning_rate": 3.2569920229800285e-05, "loss": 0.0011, "step": 53902 }, { "epoch": 1.05, "learning_rate": 3.2569273269543054e-05, "loss": 0.0, "step": 53904 }, { "epoch": 1.05, "learning_rate": 3.2568626309285823e-05, "loss": 0.0, "step": 53906 }, { "epoch": 1.05, "learning_rate": 3.256797934902859e-05, "loss": 0.0, "step": 53908 }, { "epoch": 1.05, "learning_rate": 3.256733238877136e-05, "loss": 0.0, "step": 53910 }, { "epoch": 1.05, "learning_rate": 3.2566685428514124e-05, "loss": 0.0, "step": 53912 }, { "epoch": 1.05, "learning_rate": 3.25660384682569e-05, "loss": 0.007, "step": 53914 }, { "epoch": 1.05, "learning_rate": 3.256539150799966e-05, "loss": 0.0002, "step": 53916 }, { "epoch": 1.05, "learning_rate": 3.256474454774243e-05, "loss": 0.0027, "step": 53918 }, { "epoch": 1.05, "learning_rate": 3.256409758748521e-05, "loss": 0.0212, "step": 53920 }, { "epoch": 1.05, "learning_rate": 3.256345062722797e-05, "loss": 0.0014, "step": 53922 }, { "epoch": 1.05, "learning_rate": 3.256280366697074e-05, "loss": 0.0, "step": 53924 }, { "epoch": 1.05, "learning_rate": 3.256215670671351e-05, "loss": 0.0001, "step": 53926 }, { "epoch": 1.05, "learning_rate": 3.2561509746456277e-05, "loss": 0.0, "step": 53928 }, { "epoch": 1.05, "learning_rate": 3.2560862786199046e-05, "loss": 0.0, "step": 53930 }, { "epoch": 1.05, "learning_rate": 3.2560215825941815e-05, "loss": 0.0, "step": 53932 }, { "epoch": 1.05, "learning_rate": 3.2559568865684584e-05, "loss": 0.0, "step": 53934 }, { "epoch": 1.05, "learning_rate": 3.255892190542735e-05, "loss": 0.0, "step": 53936 }, { "epoch": 1.05, "learning_rate": 3.255827494517012e-05, "loss": 0.0001, "step": 53938 }, { "epoch": 1.05, "learning_rate": 3.2557627984912884e-05, "loss": 0.0001, "step": 53940 }, { "epoch": 1.05, "learning_rate": 3.255698102465566e-05, "loss": 0.0001, "step": 53942 }, { "epoch": 1.05, "learning_rate": 3.255633406439842e-05, "loss": 0.0001, "step": 53944 }, { "epoch": 1.05, "learning_rate": 3.255568710414119e-05, "loss": 0.0002, "step": 53946 }, { "epoch": 1.05, "learning_rate": 3.255504014388396e-05, "loss": 0.0, "step": 53948 }, { "epoch": 1.05, "learning_rate": 3.255439318362673e-05, "loss": 0.0, "step": 53950 }, { "epoch": 1.05, "learning_rate": 3.2553746223369506e-05, "loss": 0.0, "step": 53952 }, { "epoch": 1.05, "learning_rate": 3.255309926311227e-05, "loss": 0.0, "step": 53954 }, { "epoch": 1.05, "learning_rate": 3.255245230285504e-05, "loss": 0.0, "step": 53956 }, { "epoch": 1.05, "learning_rate": 3.2551805342597806e-05, "loss": 0.0062, "step": 53958 }, { "epoch": 1.05, "learning_rate": 3.2551158382340575e-05, "loss": 0.0, "step": 53960 }, { "epoch": 1.05, "learning_rate": 3.255051142208334e-05, "loss": 0.0, "step": 53962 }, { "epoch": 1.05, "learning_rate": 3.254986446182611e-05, "loss": 0.0019, "step": 53964 }, { "epoch": 1.05, "learning_rate": 3.254921750156888e-05, "loss": 0.0, "step": 53966 }, { "epoch": 1.05, "learning_rate": 3.254857054131165e-05, "loss": 0.0, "step": 53968 }, { "epoch": 1.05, "learning_rate": 3.254792358105442e-05, "loss": 0.0, "step": 53970 }, { "epoch": 1.05, "learning_rate": 3.254727662079718e-05, "loss": 0.0001, "step": 53972 }, { "epoch": 1.05, "learning_rate": 3.254662966053996e-05, "loss": 0.0085, "step": 53974 }, { "epoch": 1.05, "learning_rate": 3.254598270028272e-05, "loss": 0.0, "step": 53976 }, { "epoch": 1.05, "learning_rate": 3.254533574002549e-05, "loss": 0.0, "step": 53978 }, { "epoch": 1.05, "learning_rate": 3.254468877976826e-05, "loss": 0.0, "step": 53980 }, { "epoch": 1.05, "learning_rate": 3.254404181951103e-05, "loss": 0.0009, "step": 53982 }, { "epoch": 1.05, "learning_rate": 3.25433948592538e-05, "loss": 0.0, "step": 53984 }, { "epoch": 1.05, "learning_rate": 3.2542747898996566e-05, "loss": 0.0092, "step": 53986 }, { "epoch": 1.05, "learning_rate": 3.2542100938739336e-05, "loss": 0.0, "step": 53988 }, { "epoch": 1.05, "learning_rate": 3.2541453978482105e-05, "loss": 0.0, "step": 53990 }, { "epoch": 1.05, "learning_rate": 3.2540807018224874e-05, "loss": 0.0, "step": 53992 }, { "epoch": 1.05, "learning_rate": 3.2540160057967636e-05, "loss": 0.0001, "step": 53994 }, { "epoch": 1.05, "learning_rate": 3.253951309771041e-05, "loss": 0.0, "step": 53996 }, { "epoch": 1.05, "learning_rate": 3.253886613745318e-05, "loss": 0.0, "step": 53998 }, { "epoch": 1.05, "learning_rate": 3.253821917719594e-05, "loss": 0.0, "step": 54000 }, { "epoch": 1.05, "learning_rate": 3.253757221693872e-05, "loss": 0.0, "step": 54002 }, { "epoch": 1.05, "learning_rate": 3.253692525668148e-05, "loss": 0.0, "step": 54004 }, { "epoch": 1.05, "learning_rate": 3.253627829642425e-05, "loss": 0.0, "step": 54006 }, { "epoch": 1.05, "learning_rate": 3.253563133616702e-05, "loss": 0.0, "step": 54008 }, { "epoch": 1.05, "learning_rate": 3.253498437590979e-05, "loss": 0.0, "step": 54010 }, { "epoch": 1.05, "learning_rate": 3.253433741565256e-05, "loss": 0.0, "step": 54012 }, { "epoch": 1.05, "learning_rate": 3.253369045539533e-05, "loss": 0.0, "step": 54014 }, { "epoch": 1.05, "learning_rate": 3.2533043495138096e-05, "loss": 0.0, "step": 54016 }, { "epoch": 1.05, "learning_rate": 3.2532396534880865e-05, "loss": 0.0, "step": 54018 }, { "epoch": 1.05, "learning_rate": 3.2531749574623634e-05, "loss": 0.0, "step": 54020 }, { "epoch": 1.05, "learning_rate": 3.2531102614366396e-05, "loss": 0.0, "step": 54022 }, { "epoch": 1.05, "learning_rate": 3.253045565410917e-05, "loss": 0.0, "step": 54024 }, { "epoch": 1.05, "learning_rate": 3.2529808693851935e-05, "loss": 0.0, "step": 54026 }, { "epoch": 1.05, "learning_rate": 3.2529161733594704e-05, "loss": 0.0, "step": 54028 }, { "epoch": 1.05, "learning_rate": 3.252851477333748e-05, "loss": 0.0, "step": 54030 }, { "epoch": 1.05, "learning_rate": 3.252786781308024e-05, "loss": 0.0, "step": 54032 }, { "epoch": 1.05, "learning_rate": 3.252722085282302e-05, "loss": 0.0, "step": 54034 }, { "epoch": 1.05, "learning_rate": 3.252657389256578e-05, "loss": 0.0001, "step": 54036 }, { "epoch": 1.05, "learning_rate": 3.252592693230855e-05, "loss": 0.0, "step": 54038 }, { "epoch": 1.05, "learning_rate": 3.252527997205132e-05, "loss": 0.0, "step": 54040 }, { "epoch": 1.05, "learning_rate": 3.252463301179409e-05, "loss": 0.0, "step": 54042 }, { "epoch": 1.05, "learning_rate": 3.2523986051536856e-05, "loss": 0.0, "step": 54044 }, { "epoch": 1.05, "learning_rate": 3.2523339091279625e-05, "loss": 0.0, "step": 54046 }, { "epoch": 1.05, "learning_rate": 3.2522692131022395e-05, "loss": 0.0, "step": 54048 }, { "epoch": 1.05, "learning_rate": 3.2522045170765164e-05, "loss": 0.0001, "step": 54050 }, { "epoch": 1.05, "learning_rate": 3.252139821050793e-05, "loss": 0.0, "step": 54052 }, { "epoch": 1.05, "learning_rate": 3.2520751250250695e-05, "loss": 0.0, "step": 54054 }, { "epoch": 1.05, "learning_rate": 3.252010428999347e-05, "loss": 0.0006, "step": 54056 }, { "epoch": 1.05, "learning_rate": 3.251945732973623e-05, "loss": 0.0, "step": 54058 }, { "epoch": 1.05, "learning_rate": 3.2518810369479e-05, "loss": 0.0047, "step": 54060 }, { "epoch": 1.05, "learning_rate": 3.251816340922177e-05, "loss": 0.0, "step": 54062 }, { "epoch": 1.05, "learning_rate": 3.251751644896454e-05, "loss": 0.0, "step": 54064 }, { "epoch": 1.05, "learning_rate": 3.251686948870731e-05, "loss": 0.0, "step": 54066 }, { "epoch": 1.05, "learning_rate": 3.251622252845008e-05, "loss": 0.0, "step": 54068 }, { "epoch": 1.05, "learning_rate": 3.251557556819285e-05, "loss": 0.0, "step": 54070 }, { "epoch": 1.05, "learning_rate": 3.251492860793562e-05, "loss": 0.0002, "step": 54072 }, { "epoch": 1.05, "learning_rate": 3.2514281647678386e-05, "loss": 0.0004, "step": 54074 }, { "epoch": 1.05, "learning_rate": 3.2513634687421155e-05, "loss": 0.0, "step": 54076 }, { "epoch": 1.05, "learning_rate": 3.2512987727163924e-05, "loss": 0.0, "step": 54078 }, { "epoch": 1.05, "learning_rate": 3.251234076690669e-05, "loss": 0.0018, "step": 54080 }, { "epoch": 1.05, "learning_rate": 3.2511693806649455e-05, "loss": 0.0, "step": 54082 }, { "epoch": 1.05, "learning_rate": 3.251104684639223e-05, "loss": 0.0001, "step": 54084 }, { "epoch": 1.05, "learning_rate": 3.2510399886134994e-05, "loss": 0.0002, "step": 54086 }, { "epoch": 1.05, "learning_rate": 3.250975292587776e-05, "loss": 0.002, "step": 54088 }, { "epoch": 1.05, "learning_rate": 3.250910596562053e-05, "loss": 0.0, "step": 54090 }, { "epoch": 1.05, "learning_rate": 3.25084590053633e-05, "loss": 0.0079, "step": 54092 }, { "epoch": 1.05, "learning_rate": 3.250781204510607e-05, "loss": 0.0, "step": 54094 }, { "epoch": 1.05, "learning_rate": 3.250716508484884e-05, "loss": 0.0, "step": 54096 }, { "epoch": 1.05, "learning_rate": 3.250651812459161e-05, "loss": 0.0, "step": 54098 }, { "epoch": 1.05, "learning_rate": 3.250587116433438e-05, "loss": 0.0001, "step": 54100 }, { "epoch": 1.05, "learning_rate": 3.2505224204077146e-05, "loss": 0.0, "step": 54102 }, { "epoch": 1.05, "learning_rate": 3.250457724381991e-05, "loss": 0.0009, "step": 54104 }, { "epoch": 1.05, "learning_rate": 3.2503930283562684e-05, "loss": 0.0001, "step": 54106 }, { "epoch": 1.05, "learning_rate": 3.250328332330545e-05, "loss": 0.0031, "step": 54108 }, { "epoch": 1.05, "learning_rate": 3.250263636304822e-05, "loss": 0.0, "step": 54110 }, { "epoch": 1.05, "learning_rate": 3.250198940279099e-05, "loss": 0.0001, "step": 54112 }, { "epoch": 1.05, "learning_rate": 3.2501342442533754e-05, "loss": 0.0, "step": 54114 }, { "epoch": 1.05, "learning_rate": 3.250069548227653e-05, "loss": 0.0094, "step": 54116 }, { "epoch": 1.05, "learning_rate": 3.250004852201929e-05, "loss": 0.0021, "step": 54118 }, { "epoch": 1.05, "learning_rate": 3.249940156176206e-05, "loss": 0.0, "step": 54120 }, { "epoch": 1.05, "learning_rate": 3.249875460150483e-05, "loss": 0.0, "step": 54122 }, { "epoch": 1.05, "learning_rate": 3.24981076412476e-05, "loss": 0.0, "step": 54124 }, { "epoch": 1.05, "learning_rate": 3.249746068099037e-05, "loss": 0.0, "step": 54126 }, { "epoch": 1.05, "learning_rate": 3.249681372073314e-05, "loss": 0.0, "step": 54128 }, { "epoch": 1.05, "learning_rate": 3.249616676047591e-05, "loss": 0.0, "step": 54130 }, { "epoch": 1.05, "learning_rate": 3.2495519800218676e-05, "loss": 0.0, "step": 54132 }, { "epoch": 1.05, "learning_rate": 3.2494872839961445e-05, "loss": 0.0, "step": 54134 }, { "epoch": 1.05, "learning_rate": 3.249422587970421e-05, "loss": 0.0001, "step": 54136 }, { "epoch": 1.05, "learning_rate": 3.249357891944698e-05, "loss": 0.0, "step": 54138 }, { "epoch": 1.05, "learning_rate": 3.2492931959189745e-05, "loss": 0.0054, "step": 54140 }, { "epoch": 1.05, "learning_rate": 3.2492284998932514e-05, "loss": 0.0, "step": 54142 }, { "epoch": 1.05, "learning_rate": 3.249163803867529e-05, "loss": 0.0, "step": 54144 }, { "epoch": 1.05, "learning_rate": 3.249099107841805e-05, "loss": 0.0, "step": 54146 }, { "epoch": 1.05, "learning_rate": 3.249034411816082e-05, "loss": 0.0, "step": 54148 }, { "epoch": 1.05, "learning_rate": 3.248969715790359e-05, "loss": 0.0, "step": 54150 }, { "epoch": 1.05, "learning_rate": 3.248905019764636e-05, "loss": 0.0, "step": 54152 }, { "epoch": 1.05, "learning_rate": 3.248840323738913e-05, "loss": 0.0, "step": 54154 }, { "epoch": 1.05, "learning_rate": 3.24877562771319e-05, "loss": 0.0084, "step": 54156 }, { "epoch": 1.05, "learning_rate": 3.248710931687467e-05, "loss": 0.0, "step": 54158 }, { "epoch": 1.05, "learning_rate": 3.2486462356617436e-05, "loss": 0.0001, "step": 54160 }, { "epoch": 1.05, "learning_rate": 3.2485815396360205e-05, "loss": 0.0, "step": 54162 }, { "epoch": 1.05, "learning_rate": 3.248516843610297e-05, "loss": 0.0, "step": 54164 }, { "epoch": 1.05, "learning_rate": 3.2484521475845743e-05, "loss": 0.0, "step": 54166 }, { "epoch": 1.05, "learning_rate": 3.2483874515588506e-05, "loss": 0.0, "step": 54168 }, { "epoch": 1.05, "learning_rate": 3.2483227555331275e-05, "loss": 0.0, "step": 54170 }, { "epoch": 1.05, "learning_rate": 3.2482580595074044e-05, "loss": 0.0, "step": 54172 }, { "epoch": 1.05, "learning_rate": 3.248193363481681e-05, "loss": 0.0001, "step": 54174 }, { "epoch": 1.05, "learning_rate": 3.248128667455959e-05, "loss": 0.0, "step": 54176 }, { "epoch": 1.05, "learning_rate": 3.248063971430235e-05, "loss": 0.0, "step": 54178 }, { "epoch": 1.05, "learning_rate": 3.247999275404512e-05, "loss": 0.0, "step": 54180 }, { "epoch": 1.05, "learning_rate": 3.247934579378789e-05, "loss": 0.0003, "step": 54182 }, { "epoch": 1.05, "learning_rate": 3.247869883353066e-05, "loss": 0.0006, "step": 54184 }, { "epoch": 1.05, "learning_rate": 3.247805187327342e-05, "loss": 0.0, "step": 54186 }, { "epoch": 1.05, "learning_rate": 3.2477404913016197e-05, "loss": 0.0, "step": 54188 }, { "epoch": 1.05, "learning_rate": 3.2476757952758966e-05, "loss": 0.0, "step": 54190 }, { "epoch": 1.05, "learning_rate": 3.2476110992501735e-05, "loss": 0.0, "step": 54192 }, { "epoch": 1.05, "learning_rate": 3.2475464032244504e-05, "loss": 0.0, "step": 54194 }, { "epoch": 1.05, "learning_rate": 3.2474817071987266e-05, "loss": 0.0, "step": 54196 }, { "epoch": 1.05, "learning_rate": 3.247417011173004e-05, "loss": 0.0, "step": 54198 }, { "epoch": 1.05, "learning_rate": 3.2473523151472804e-05, "loss": 0.0002, "step": 54200 }, { "epoch": 1.05, "learning_rate": 3.2472876191215573e-05, "loss": 0.0, "step": 54202 }, { "epoch": 1.05, "learning_rate": 3.247222923095834e-05, "loss": 0.0, "step": 54204 }, { "epoch": 1.05, "learning_rate": 3.247158227070111e-05, "loss": 0.0, "step": 54206 }, { "epoch": 1.05, "learning_rate": 3.247093531044388e-05, "loss": 0.0, "step": 54208 }, { "epoch": 1.05, "learning_rate": 3.247028835018665e-05, "loss": 0.0, "step": 54210 }, { "epoch": 1.05, "learning_rate": 3.246964138992942e-05, "loss": 0.0, "step": 54212 }, { "epoch": 1.05, "learning_rate": 3.246899442967219e-05, "loss": 0.0, "step": 54214 }, { "epoch": 1.05, "learning_rate": 3.246834746941496e-05, "loss": 0.0, "step": 54216 }, { "epoch": 1.05, "learning_rate": 3.246770050915772e-05, "loss": 0.0031, "step": 54218 }, { "epoch": 1.05, "learning_rate": 3.2467053548900495e-05, "loss": 0.0, "step": 54220 }, { "epoch": 1.05, "learning_rate": 3.2466406588643264e-05, "loss": 0.0, "step": 54222 }, { "epoch": 1.05, "learning_rate": 3.2465759628386027e-05, "loss": 0.0, "step": 54224 }, { "epoch": 1.05, "learning_rate": 3.24651126681288e-05, "loss": 0.0, "step": 54226 }, { "epoch": 1.05, "learning_rate": 3.2464465707871565e-05, "loss": 0.0, "step": 54228 }, { "epoch": 1.05, "learning_rate": 3.2463818747614334e-05, "loss": 0.0037, "step": 54230 }, { "epoch": 1.05, "learning_rate": 3.24631717873571e-05, "loss": 0.0, "step": 54232 }, { "epoch": 1.05, "learning_rate": 3.246252482709987e-05, "loss": 0.0005, "step": 54234 }, { "epoch": 1.05, "learning_rate": 3.246187786684264e-05, "loss": 0.0001, "step": 54236 }, { "epoch": 1.05, "learning_rate": 3.246123090658541e-05, "loss": 0.0002, "step": 54238 }, { "epoch": 1.05, "learning_rate": 3.246058394632818e-05, "loss": 0.0, "step": 54240 }, { "epoch": 1.05, "learning_rate": 3.245993698607095e-05, "loss": 0.0001, "step": 54242 }, { "epoch": 1.05, "learning_rate": 3.245929002581372e-05, "loss": 0.003, "step": 54244 }, { "epoch": 1.05, "learning_rate": 3.245864306555648e-05, "loss": 0.0002, "step": 54246 }, { "epoch": 1.05, "learning_rate": 3.2457996105299256e-05, "loss": 0.0, "step": 54248 }, { "epoch": 1.05, "learning_rate": 3.245734914504202e-05, "loss": 0.0017, "step": 54250 }, { "epoch": 1.05, "learning_rate": 3.2456702184784794e-05, "loss": 0.0001, "step": 54252 }, { "epoch": 1.05, "learning_rate": 3.245605522452756e-05, "loss": 0.0, "step": 54254 }, { "epoch": 1.05, "learning_rate": 3.2455408264270325e-05, "loss": 0.0014, "step": 54256 }, { "epoch": 1.05, "learning_rate": 3.24547613040131e-05, "loss": 0.0041, "step": 54258 }, { "epoch": 1.05, "learning_rate": 3.245411434375586e-05, "loss": 0.0, "step": 54260 }, { "epoch": 1.05, "learning_rate": 3.245346738349863e-05, "loss": 0.0, "step": 54262 }, { "epoch": 1.05, "learning_rate": 3.24528204232414e-05, "loss": 0.0, "step": 54264 }, { "epoch": 1.05, "learning_rate": 3.245217346298417e-05, "loss": 0.0, "step": 54266 }, { "epoch": 1.05, "learning_rate": 3.245152650272694e-05, "loss": 0.0, "step": 54268 }, { "epoch": 1.05, "learning_rate": 3.245087954246971e-05, "loss": 0.0, "step": 54270 }, { "epoch": 1.05, "learning_rate": 3.245023258221248e-05, "loss": 0.0079, "step": 54272 }, { "epoch": 1.05, "learning_rate": 3.244958562195525e-05, "loss": 0.0, "step": 54274 }, { "epoch": 1.05, "learning_rate": 3.2448938661698016e-05, "loss": 0.0, "step": 54276 }, { "epoch": 1.05, "learning_rate": 3.244829170144078e-05, "loss": 0.0088, "step": 54278 }, { "epoch": 1.05, "learning_rate": 3.2447644741183554e-05, "loss": 0.0002, "step": 54280 }, { "epoch": 1.05, "learning_rate": 3.2446997780926316e-05, "loss": 0.0, "step": 54282 }, { "epoch": 1.05, "learning_rate": 3.2446350820669086e-05, "loss": 0.0, "step": 54284 }, { "epoch": 1.05, "learning_rate": 3.2445703860411855e-05, "loss": 0.0, "step": 54286 }, { "epoch": 1.05, "learning_rate": 3.2445056900154624e-05, "loss": 0.0015, "step": 54288 }, { "epoch": 1.05, "learning_rate": 3.244440993989739e-05, "loss": 0.0, "step": 54290 }, { "epoch": 1.05, "learning_rate": 3.244376297964016e-05, "loss": 0.0, "step": 54292 }, { "epoch": 1.05, "learning_rate": 3.244311601938293e-05, "loss": 0.0248, "step": 54294 }, { "epoch": 1.05, "learning_rate": 3.24424690591257e-05, "loss": 0.0, "step": 54296 }, { "epoch": 1.05, "learning_rate": 3.244182209886847e-05, "loss": 0.0006, "step": 54298 }, { "epoch": 1.05, "learning_rate": 3.244117513861124e-05, "loss": 0.0, "step": 54300 }, { "epoch": 1.05, "learning_rate": 3.244052817835401e-05, "loss": 0.0013, "step": 54302 }, { "epoch": 1.05, "learning_rate": 3.2439881218096776e-05, "loss": 0.0, "step": 54304 }, { "epoch": 1.05, "learning_rate": 3.243923425783954e-05, "loss": 0.0, "step": 54306 }, { "epoch": 1.05, "learning_rate": 3.2438587297582315e-05, "loss": 0.0008, "step": 54308 }, { "epoch": 1.05, "learning_rate": 3.243794033732508e-05, "loss": 0.0, "step": 54310 }, { "epoch": 1.05, "learning_rate": 3.2437293377067846e-05, "loss": 0.0, "step": 54312 }, { "epoch": 1.05, "learning_rate": 3.2436646416810615e-05, "loss": 0.0007, "step": 54314 }, { "epoch": 1.05, "learning_rate": 3.2435999456553384e-05, "loss": 0.0, "step": 54316 }, { "epoch": 1.05, "learning_rate": 3.243535249629615e-05, "loss": 0.0, "step": 54318 }, { "epoch": 1.05, "learning_rate": 3.243470553603892e-05, "loss": 0.0, "step": 54320 }, { "epoch": 1.05, "learning_rate": 3.243405857578169e-05, "loss": 0.0001, "step": 54322 }, { "epoch": 1.05, "learning_rate": 3.243341161552446e-05, "loss": 0.0, "step": 54324 }, { "epoch": 1.05, "learning_rate": 3.243276465526723e-05, "loss": 0.0, "step": 54326 }, { "epoch": 1.05, "learning_rate": 3.243211769500999e-05, "loss": 0.0, "step": 54328 }, { "epoch": 1.05, "learning_rate": 3.243147073475277e-05, "loss": 0.0004, "step": 54330 }, { "epoch": 1.05, "learning_rate": 3.243082377449554e-05, "loss": 0.0, "step": 54332 }, { "epoch": 1.05, "learning_rate": 3.2430176814238306e-05, "loss": 0.0012, "step": 54334 }, { "epoch": 1.05, "learning_rate": 3.2429529853981075e-05, "loss": 0.0, "step": 54336 }, { "epoch": 1.05, "learning_rate": 3.242888289372384e-05, "loss": 0.0, "step": 54338 }, { "epoch": 1.05, "learning_rate": 3.242823593346661e-05, "loss": 0.0038, "step": 54340 }, { "epoch": 1.05, "learning_rate": 3.2427588973209375e-05, "loss": 0.0009, "step": 54342 }, { "epoch": 1.05, "learning_rate": 3.2426942012952145e-05, "loss": 0.0, "step": 54344 }, { "epoch": 1.05, "learning_rate": 3.2426295052694914e-05, "loss": 0.0, "step": 54346 }, { "epoch": 1.05, "learning_rate": 3.242564809243768e-05, "loss": 0.0001, "step": 54348 }, { "epoch": 1.05, "learning_rate": 3.242500113218045e-05, "loss": 0.0, "step": 54350 }, { "epoch": 1.05, "learning_rate": 3.242435417192322e-05, "loss": 0.0, "step": 54352 }, { "epoch": 1.05, "learning_rate": 3.242370721166599e-05, "loss": 0.0002, "step": 54354 }, { "epoch": 1.05, "learning_rate": 3.242306025140876e-05, "loss": 0.0, "step": 54356 }, { "epoch": 1.06, "learning_rate": 3.242241329115153e-05, "loss": 0.0, "step": 54358 }, { "epoch": 1.06, "learning_rate": 3.242176633089429e-05, "loss": 0.0, "step": 54360 }, { "epoch": 1.06, "learning_rate": 3.2421119370637066e-05, "loss": 0.0, "step": 54362 }, { "epoch": 1.06, "learning_rate": 3.242047241037983e-05, "loss": 0.0, "step": 54364 }, { "epoch": 1.06, "learning_rate": 3.24198254501226e-05, "loss": 0.0, "step": 54366 }, { "epoch": 1.06, "learning_rate": 3.2419178489865374e-05, "loss": 0.0, "step": 54368 }, { "epoch": 1.06, "learning_rate": 3.2418531529608136e-05, "loss": 0.0, "step": 54370 }, { "epoch": 1.06, "learning_rate": 3.2417884569350905e-05, "loss": 0.0, "step": 54372 }, { "epoch": 1.06, "learning_rate": 3.2417237609093674e-05, "loss": 0.0, "step": 54374 }, { "epoch": 1.06, "learning_rate": 3.241659064883644e-05, "loss": 0.008, "step": 54376 }, { "epoch": 1.06, "learning_rate": 3.241594368857921e-05, "loss": 0.0, "step": 54378 }, { "epoch": 1.06, "learning_rate": 3.241529672832198e-05, "loss": 0.0006, "step": 54380 }, { "epoch": 1.06, "learning_rate": 3.241464976806475e-05, "loss": 0.0, "step": 54382 }, { "epoch": 1.06, "learning_rate": 3.241400280780752e-05, "loss": 0.0001, "step": 54384 }, { "epoch": 1.06, "learning_rate": 3.241335584755029e-05, "loss": 0.0029, "step": 54386 }, { "epoch": 1.06, "learning_rate": 3.241270888729305e-05, "loss": 0.0001, "step": 54388 }, { "epoch": 1.06, "learning_rate": 3.241206192703583e-05, "loss": 0.0001, "step": 54390 }, { "epoch": 1.06, "learning_rate": 3.241141496677859e-05, "loss": 0.0, "step": 54392 }, { "epoch": 1.06, "learning_rate": 3.2410768006521365e-05, "loss": 0.0001, "step": 54394 }, { "epoch": 1.06, "learning_rate": 3.241012104626413e-05, "loss": 0.0, "step": 54396 }, { "epoch": 1.06, "learning_rate": 3.2409474086006896e-05, "loss": 0.0, "step": 54398 }, { "epoch": 1.06, "learning_rate": 3.240882712574967e-05, "loss": 0.0, "step": 54400 }, { "epoch": 1.06, "learning_rate": 3.2408180165492434e-05, "loss": 0.001, "step": 54402 }, { "epoch": 1.06, "learning_rate": 3.2407533205235203e-05, "loss": 0.0222, "step": 54404 }, { "epoch": 1.06, "learning_rate": 3.240688624497797e-05, "loss": 0.0, "step": 54406 }, { "epoch": 1.06, "learning_rate": 3.240623928472074e-05, "loss": 0.0001, "step": 54408 }, { "epoch": 1.06, "learning_rate": 3.2405592324463504e-05, "loss": 0.0, "step": 54410 }, { "epoch": 1.06, "learning_rate": 3.240494536420628e-05, "loss": 0.0, "step": 54412 }, { "epoch": 1.06, "learning_rate": 3.240429840394905e-05, "loss": 0.0069, "step": 54414 }, { "epoch": 1.06, "learning_rate": 3.240365144369182e-05, "loss": 0.0002, "step": 54416 }, { "epoch": 1.06, "learning_rate": 3.240300448343459e-05, "loss": 0.0, "step": 54418 }, { "epoch": 1.06, "learning_rate": 3.240235752317735e-05, "loss": 0.0, "step": 54420 }, { "epoch": 1.06, "learning_rate": 3.2401710562920125e-05, "loss": 0.0175, "step": 54422 }, { "epoch": 1.06, "learning_rate": 3.240106360266289e-05, "loss": 0.0001, "step": 54424 }, { "epoch": 1.06, "learning_rate": 3.240041664240566e-05, "loss": 0.0, "step": 54426 }, { "epoch": 1.06, "learning_rate": 3.2399769682148426e-05, "loss": 0.0, "step": 54428 }, { "epoch": 1.06, "learning_rate": 3.2399122721891195e-05, "loss": 0.0002, "step": 54430 }, { "epoch": 1.06, "learning_rate": 3.2398475761633964e-05, "loss": 0.0001, "step": 54432 }, { "epoch": 1.06, "learning_rate": 3.239782880137673e-05, "loss": 0.0, "step": 54434 }, { "epoch": 1.06, "learning_rate": 3.23971818411195e-05, "loss": 0.0002, "step": 54436 }, { "epoch": 1.06, "learning_rate": 3.239653488086227e-05, "loss": 0.0, "step": 54438 }, { "epoch": 1.06, "learning_rate": 3.239588792060504e-05, "loss": 0.0, "step": 54440 }, { "epoch": 1.06, "learning_rate": 3.23952409603478e-05, "loss": 0.0, "step": 54442 }, { "epoch": 1.06, "learning_rate": 3.239459400009058e-05, "loss": 0.0004, "step": 54444 }, { "epoch": 1.06, "learning_rate": 3.239394703983335e-05, "loss": 0.0, "step": 54446 }, { "epoch": 1.06, "learning_rate": 3.239330007957611e-05, "loss": 0.0001, "step": 54448 }, { "epoch": 1.06, "learning_rate": 3.2392653119318886e-05, "loss": 0.0001, "step": 54450 }, { "epoch": 1.06, "learning_rate": 3.239200615906165e-05, "loss": 0.0017, "step": 54452 }, { "epoch": 1.06, "learning_rate": 3.239135919880442e-05, "loss": 0.0, "step": 54454 }, { "epoch": 1.06, "learning_rate": 3.2390712238547186e-05, "loss": 0.0, "step": 54456 }, { "epoch": 1.06, "learning_rate": 3.2390065278289955e-05, "loss": 0.0, "step": 54458 }, { "epoch": 1.06, "learning_rate": 3.2389418318032724e-05, "loss": 0.0002, "step": 54460 }, { "epoch": 1.06, "learning_rate": 3.238877135777549e-05, "loss": 0.0, "step": 54462 }, { "epoch": 1.06, "learning_rate": 3.238812439751826e-05, "loss": 0.0, "step": 54464 }, { "epoch": 1.06, "learning_rate": 3.238747743726103e-05, "loss": 0.0, "step": 54466 }, { "epoch": 1.06, "learning_rate": 3.23868304770038e-05, "loss": 0.0, "step": 54468 }, { "epoch": 1.06, "learning_rate": 3.238618351674656e-05, "loss": 0.0, "step": 54470 }, { "epoch": 1.06, "learning_rate": 3.238553655648934e-05, "loss": 0.0004, "step": 54472 }, { "epoch": 1.06, "learning_rate": 3.23848895962321e-05, "loss": 0.0, "step": 54474 }, { "epoch": 1.06, "learning_rate": 3.238424263597488e-05, "loss": 0.0, "step": 54476 }, { "epoch": 1.06, "learning_rate": 3.2383595675717646e-05, "loss": 0.0, "step": 54478 }, { "epoch": 1.06, "learning_rate": 3.238294871546041e-05, "loss": 0.0001, "step": 54480 }, { "epoch": 1.06, "learning_rate": 3.2382301755203184e-05, "loss": 0.0013, "step": 54482 }, { "epoch": 1.06, "learning_rate": 3.2381654794945947e-05, "loss": 0.0004, "step": 54484 }, { "epoch": 1.06, "learning_rate": 3.2381007834688716e-05, "loss": 0.0, "step": 54486 }, { "epoch": 1.06, "learning_rate": 3.2380360874431485e-05, "loss": 0.0, "step": 54488 }, { "epoch": 1.06, "learning_rate": 3.2379713914174254e-05, "loss": 0.0008, "step": 54490 }, { "epoch": 1.06, "learning_rate": 3.237906695391702e-05, "loss": 0.0005, "step": 54492 }, { "epoch": 1.06, "learning_rate": 3.237841999365979e-05, "loss": 0.0103, "step": 54494 }, { "epoch": 1.06, "learning_rate": 3.237777303340256e-05, "loss": 0.0, "step": 54496 }, { "epoch": 1.06, "learning_rate": 3.237712607314533e-05, "loss": 0.0, "step": 54498 }, { "epoch": 1.06, "learning_rate": 3.23764791128881e-05, "loss": 0.0001, "step": 54500 }, { "epoch": 1.06, "learning_rate": 3.237583215263086e-05, "loss": 0.0, "step": 54502 }, { "epoch": 1.06, "learning_rate": 3.237518519237364e-05, "loss": 0.0, "step": 54504 }, { "epoch": 1.06, "learning_rate": 3.23745382321164e-05, "loss": 0.0, "step": 54506 }, { "epoch": 1.06, "learning_rate": 3.237389127185917e-05, "loss": 0.0, "step": 54508 }, { "epoch": 1.06, "learning_rate": 3.2373244311601945e-05, "loss": 0.0, "step": 54510 }, { "epoch": 1.06, "learning_rate": 3.237259735134471e-05, "loss": 0.0, "step": 54512 }, { "epoch": 1.06, "learning_rate": 3.2371950391087476e-05, "loss": 0.0, "step": 54514 }, { "epoch": 1.06, "learning_rate": 3.2371303430830245e-05, "loss": 0.0, "step": 54516 }, { "epoch": 1.06, "learning_rate": 3.2370656470573014e-05, "loss": 0.0, "step": 54518 }, { "epoch": 1.06, "learning_rate": 3.237000951031578e-05, "loss": 0.0, "step": 54520 }, { "epoch": 1.06, "learning_rate": 3.236936255005855e-05, "loss": 0.0001, "step": 54522 }, { "epoch": 1.06, "learning_rate": 3.236871558980132e-05, "loss": 0.0089, "step": 54524 }, { "epoch": 1.06, "learning_rate": 3.236806862954409e-05, "loss": 0.0, "step": 54526 }, { "epoch": 1.06, "learning_rate": 3.236742166928686e-05, "loss": 0.0, "step": 54528 }, { "epoch": 1.06, "learning_rate": 3.236677470902962e-05, "loss": 0.0, "step": 54530 }, { "epoch": 1.06, "learning_rate": 3.23661277487724e-05, "loss": 0.0, "step": 54532 }, { "epoch": 1.06, "learning_rate": 3.236548078851516e-05, "loss": 0.0003, "step": 54534 }, { "epoch": 1.06, "learning_rate": 3.2364833828257936e-05, "loss": 0.0002, "step": 54536 }, { "epoch": 1.06, "learning_rate": 3.23641868680007e-05, "loss": 0.0, "step": 54538 }, { "epoch": 1.06, "learning_rate": 3.236353990774347e-05, "loss": 0.0, "step": 54540 }, { "epoch": 1.06, "learning_rate": 3.2362892947486236e-05, "loss": 0.0, "step": 54542 }, { "epoch": 1.06, "learning_rate": 3.2362245987229006e-05, "loss": 0.0, "step": 54544 }, { "epoch": 1.06, "learning_rate": 3.2361599026971775e-05, "loss": 0.0, "step": 54546 }, { "epoch": 1.06, "learning_rate": 3.2360952066714544e-05, "loss": 0.0, "step": 54548 }, { "epoch": 1.06, "learning_rate": 3.236030510645731e-05, "loss": 0.0, "step": 54550 }, { "epoch": 1.06, "learning_rate": 3.2359658146200075e-05, "loss": 0.0002, "step": 54552 }, { "epoch": 1.06, "learning_rate": 3.235901118594285e-05, "loss": 0.0001, "step": 54554 }, { "epoch": 1.06, "learning_rate": 3.235836422568562e-05, "loss": 0.0001, "step": 54556 }, { "epoch": 1.06, "learning_rate": 3.235771726542839e-05, "loss": 0.0, "step": 54558 }, { "epoch": 1.06, "learning_rate": 3.235707030517116e-05, "loss": 0.0003, "step": 54560 }, { "epoch": 1.06, "learning_rate": 3.235642334491392e-05, "loss": 0.0, "step": 54562 }, { "epoch": 1.06, "learning_rate": 3.2355776384656696e-05, "loss": 0.0001, "step": 54564 }, { "epoch": 1.06, "learning_rate": 3.235512942439946e-05, "loss": 0.0, "step": 54566 }, { "epoch": 1.06, "learning_rate": 3.235448246414223e-05, "loss": 0.0, "step": 54568 }, { "epoch": 1.06, "learning_rate": 3.2353835503885e-05, "loss": 0.0002, "step": 54570 }, { "epoch": 1.06, "learning_rate": 3.2353188543627766e-05, "loss": 0.0001, "step": 54572 }, { "epoch": 1.06, "learning_rate": 3.2352541583370535e-05, "loss": 0.0001, "step": 54574 }, { "epoch": 1.06, "learning_rate": 3.2351894623113304e-05, "loss": 0.0, "step": 54576 }, { "epoch": 1.06, "learning_rate": 3.235124766285607e-05, "loss": 0.0, "step": 54578 }, { "epoch": 1.06, "learning_rate": 3.235060070259884e-05, "loss": 0.0001, "step": 54580 }, { "epoch": 1.06, "learning_rate": 3.234995374234161e-05, "loss": 0.0, "step": 54582 }, { "epoch": 1.06, "learning_rate": 3.2349306782084374e-05, "loss": 0.0, "step": 54584 }, { "epoch": 1.06, "learning_rate": 3.234865982182715e-05, "loss": 0.0027, "step": 54586 }, { "epoch": 1.06, "learning_rate": 3.234801286156991e-05, "loss": 0.0, "step": 54588 }, { "epoch": 1.06, "learning_rate": 3.234736590131268e-05, "loss": 0.0029, "step": 54590 }, { "epoch": 1.06, "learning_rate": 3.234671894105546e-05, "loss": 0.0012, "step": 54592 }, { "epoch": 1.06, "learning_rate": 3.234607198079822e-05, "loss": 0.0049, "step": 54594 }, { "epoch": 1.06, "learning_rate": 3.234542502054099e-05, "loss": 0.0001, "step": 54596 }, { "epoch": 1.06, "learning_rate": 3.234477806028376e-05, "loss": 0.0001, "step": 54598 }, { "epoch": 1.06, "learning_rate": 3.2344131100026526e-05, "loss": 0.0005, "step": 54600 }, { "epoch": 1.06, "learning_rate": 3.2343484139769295e-05, "loss": 0.0, "step": 54602 }, { "epoch": 1.06, "learning_rate": 3.2342837179512064e-05, "loss": 0.0, "step": 54604 }, { "epoch": 1.06, "learning_rate": 3.2342190219254834e-05, "loss": 0.0001, "step": 54606 }, { "epoch": 1.06, "learning_rate": 3.23415432589976e-05, "loss": 0.0, "step": 54608 }, { "epoch": 1.06, "learning_rate": 3.234089629874037e-05, "loss": 0.0, "step": 54610 }, { "epoch": 1.06, "learning_rate": 3.2340249338483134e-05, "loss": 0.0001, "step": 54612 }, { "epoch": 1.06, "learning_rate": 3.233960237822591e-05, "loss": 0.0033, "step": 54614 }, { "epoch": 1.06, "learning_rate": 3.233895541796867e-05, "loss": 0.0, "step": 54616 }, { "epoch": 1.06, "learning_rate": 3.233830845771145e-05, "loss": 0.0, "step": 54618 }, { "epoch": 1.06, "learning_rate": 3.233766149745421e-05, "loss": 0.0, "step": 54620 }, { "epoch": 1.06, "learning_rate": 3.233701453719698e-05, "loss": 0.0009, "step": 54622 }, { "epoch": 1.06, "learning_rate": 3.2336367576939755e-05, "loss": 0.0, "step": 54624 }, { "epoch": 1.06, "learning_rate": 3.233572061668252e-05, "loss": 0.0, "step": 54626 }, { "epoch": 1.06, "learning_rate": 3.233507365642529e-05, "loss": 0.0, "step": 54628 }, { "epoch": 1.06, "learning_rate": 3.2334426696168056e-05, "loss": 0.0, "step": 54630 }, { "epoch": 1.06, "learning_rate": 3.2333779735910825e-05, "loss": 0.0, "step": 54632 }, { "epoch": 1.06, "learning_rate": 3.2333132775653594e-05, "loss": 0.0, "step": 54634 }, { "epoch": 1.06, "learning_rate": 3.233248581539636e-05, "loss": 0.0, "step": 54636 }, { "epoch": 1.06, "learning_rate": 3.233183885513913e-05, "loss": 0.0001, "step": 54638 }, { "epoch": 1.06, "learning_rate": 3.23311918948819e-05, "loss": 0.0006, "step": 54640 }, { "epoch": 1.06, "learning_rate": 3.233054493462467e-05, "loss": 0.0001, "step": 54642 }, { "epoch": 1.06, "learning_rate": 3.232989797436743e-05, "loss": 0.0, "step": 54644 }, { "epoch": 1.06, "learning_rate": 3.232925101411021e-05, "loss": 0.0, "step": 54646 }, { "epoch": 1.06, "learning_rate": 3.232860405385297e-05, "loss": 0.0, "step": 54648 }, { "epoch": 1.06, "learning_rate": 3.232795709359574e-05, "loss": 0.0, "step": 54650 }, { "epoch": 1.06, "learning_rate": 3.232731013333851e-05, "loss": 0.0, "step": 54652 }, { "epoch": 1.06, "learning_rate": 3.232666317308128e-05, "loss": 0.0, "step": 54654 }, { "epoch": 1.06, "learning_rate": 3.232601621282405e-05, "loss": 0.0, "step": 54656 }, { "epoch": 1.06, "learning_rate": 3.2325369252566816e-05, "loss": 0.0, "step": 54658 }, { "epoch": 1.06, "learning_rate": 3.2324722292309585e-05, "loss": 0.0, "step": 54660 }, { "epoch": 1.06, "learning_rate": 3.2324075332052354e-05, "loss": 0.0, "step": 54662 }, { "epoch": 1.06, "learning_rate": 3.2323428371795123e-05, "loss": 0.0, "step": 54664 }, { "epoch": 1.06, "learning_rate": 3.2322781411537886e-05, "loss": 0.0001, "step": 54666 }, { "epoch": 1.06, "learning_rate": 3.232213445128066e-05, "loss": 0.0, "step": 54668 }, { "epoch": 1.06, "learning_rate": 3.232148749102343e-05, "loss": 0.0, "step": 54670 }, { "epoch": 1.06, "learning_rate": 3.232084053076619e-05, "loss": 0.0, "step": 54672 }, { "epoch": 1.06, "learning_rate": 3.232019357050897e-05, "loss": 0.0002, "step": 54674 }, { "epoch": 1.06, "learning_rate": 3.231954661025173e-05, "loss": 0.0, "step": 54676 }, { "epoch": 1.06, "learning_rate": 3.231889964999451e-05, "loss": 0.0, "step": 54678 }, { "epoch": 1.06, "learning_rate": 3.231825268973727e-05, "loss": 0.0, "step": 54680 }, { "epoch": 1.06, "learning_rate": 3.231760572948004e-05, "loss": 0.0, "step": 54682 }, { "epoch": 1.06, "learning_rate": 3.231695876922281e-05, "loss": 0.0, "step": 54684 }, { "epoch": 1.06, "learning_rate": 3.231631180896558e-05, "loss": 0.0, "step": 54686 }, { "epoch": 1.06, "learning_rate": 3.2315664848708346e-05, "loss": 0.0, "step": 54688 }, { "epoch": 1.06, "learning_rate": 3.2315017888451115e-05, "loss": 0.0, "step": 54690 }, { "epoch": 1.06, "learning_rate": 3.2314370928193884e-05, "loss": 0.0, "step": 54692 }, { "epoch": 1.06, "learning_rate": 3.2313723967936646e-05, "loss": 0.0, "step": 54694 }, { "epoch": 1.06, "learning_rate": 3.231307700767942e-05, "loss": 0.0, "step": 54696 }, { "epoch": 1.06, "learning_rate": 3.2312430047422184e-05, "loss": 0.0, "step": 54698 }, { "epoch": 1.06, "learning_rate": 3.231178308716496e-05, "loss": 0.0018, "step": 54700 }, { "epoch": 1.06, "learning_rate": 3.231113612690773e-05, "loss": 0.0, "step": 54702 }, { "epoch": 1.06, "learning_rate": 3.231048916665049e-05, "loss": 0.0, "step": 54704 }, { "epoch": 1.06, "learning_rate": 3.230984220639327e-05, "loss": 0.0, "step": 54706 }, { "epoch": 1.06, "learning_rate": 3.230919524613603e-05, "loss": 0.0, "step": 54708 }, { "epoch": 1.06, "learning_rate": 3.23085482858788e-05, "loss": 0.0, "step": 54710 }, { "epoch": 1.06, "learning_rate": 3.230790132562157e-05, "loss": 0.0001, "step": 54712 }, { "epoch": 1.06, "learning_rate": 3.230725436536434e-05, "loss": 0.0, "step": 54714 }, { "epoch": 1.06, "learning_rate": 3.2306607405107106e-05, "loss": 0.0001, "step": 54716 }, { "epoch": 1.06, "learning_rate": 3.2305960444849875e-05, "loss": 0.0001, "step": 54718 }, { "epoch": 1.06, "learning_rate": 3.2305313484592644e-05, "loss": 0.0001, "step": 54720 }, { "epoch": 1.06, "learning_rate": 3.230466652433541e-05, "loss": 0.0, "step": 54722 }, { "epoch": 1.06, "learning_rate": 3.230401956407818e-05, "loss": 0.0, "step": 54724 }, { "epoch": 1.06, "learning_rate": 3.2303372603820945e-05, "loss": 0.0004, "step": 54726 }, { "epoch": 1.06, "learning_rate": 3.230272564356372e-05, "loss": 0.0, "step": 54728 }, { "epoch": 1.06, "learning_rate": 3.230207868330648e-05, "loss": 0.0024, "step": 54730 }, { "epoch": 1.06, "learning_rate": 3.230143172304925e-05, "loss": 0.0001, "step": 54732 }, { "epoch": 1.06, "learning_rate": 3.230078476279203e-05, "loss": 0.0, "step": 54734 }, { "epoch": 1.06, "learning_rate": 3.230013780253479e-05, "loss": 0.0, "step": 54736 }, { "epoch": 1.06, "learning_rate": 3.229949084227756e-05, "loss": 0.0, "step": 54738 }, { "epoch": 1.06, "learning_rate": 3.229884388202033e-05, "loss": 0.0, "step": 54740 }, { "epoch": 1.06, "learning_rate": 3.22981969217631e-05, "loss": 0.0001, "step": 54742 }, { "epoch": 1.06, "learning_rate": 3.2297549961505867e-05, "loss": 0.0003, "step": 54744 }, { "epoch": 1.06, "learning_rate": 3.2296903001248636e-05, "loss": 0.0, "step": 54746 }, { "epoch": 1.06, "learning_rate": 3.2296256040991405e-05, "loss": 0.0031, "step": 54748 }, { "epoch": 1.06, "learning_rate": 3.2295609080734174e-05, "loss": 0.0, "step": 54750 }, { "epoch": 1.06, "learning_rate": 3.229496212047694e-05, "loss": 0.0001, "step": 54752 }, { "epoch": 1.06, "learning_rate": 3.2294315160219705e-05, "loss": 0.0, "step": 54754 }, { "epoch": 1.06, "learning_rate": 3.229366819996248e-05, "loss": 0.0, "step": 54756 }, { "epoch": 1.06, "learning_rate": 3.229302123970524e-05, "loss": 0.0, "step": 54758 }, { "epoch": 1.06, "learning_rate": 3.229237427944802e-05, "loss": 0.0, "step": 54760 }, { "epoch": 1.06, "learning_rate": 3.229172731919078e-05, "loss": 0.0033, "step": 54762 }, { "epoch": 1.06, "learning_rate": 3.229108035893355e-05, "loss": 0.0, "step": 54764 }, { "epoch": 1.06, "learning_rate": 3.229043339867632e-05, "loss": 0.0, "step": 54766 }, { "epoch": 1.06, "learning_rate": 3.228978643841909e-05, "loss": 0.0, "step": 54768 }, { "epoch": 1.06, "learning_rate": 3.228913947816186e-05, "loss": 0.0, "step": 54770 }, { "epoch": 1.06, "learning_rate": 3.228849251790463e-05, "loss": 0.0, "step": 54772 }, { "epoch": 1.06, "learning_rate": 3.2287845557647396e-05, "loss": 0.002, "step": 54774 }, { "epoch": 1.06, "learning_rate": 3.228719859739016e-05, "loss": 0.0, "step": 54776 }, { "epoch": 1.06, "learning_rate": 3.2286551637132934e-05, "loss": 0.0, "step": 54778 }, { "epoch": 1.06, "learning_rate": 3.22859046768757e-05, "loss": 0.0, "step": 54780 }, { "epoch": 1.06, "learning_rate": 3.228525771661847e-05, "loss": 0.0001, "step": 54782 }, { "epoch": 1.06, "learning_rate": 3.228461075636124e-05, "loss": 0.0001, "step": 54784 }, { "epoch": 1.06, "learning_rate": 3.2283963796104004e-05, "loss": 0.0, "step": 54786 }, { "epoch": 1.06, "learning_rate": 3.228331683584678e-05, "loss": 0.0, "step": 54788 }, { "epoch": 1.06, "learning_rate": 3.228266987558954e-05, "loss": 0.0001, "step": 54790 }, { "epoch": 1.06, "learning_rate": 3.228202291533231e-05, "loss": 0.0022, "step": 54792 }, { "epoch": 1.06, "learning_rate": 3.228137595507508e-05, "loss": 0.0, "step": 54794 }, { "epoch": 1.06, "learning_rate": 3.228072899481785e-05, "loss": 0.0004, "step": 54796 }, { "epoch": 1.06, "learning_rate": 3.228008203456062e-05, "loss": 0.0001, "step": 54798 }, { "epoch": 1.06, "learning_rate": 3.227943507430339e-05, "loss": 0.0007, "step": 54800 }, { "epoch": 1.06, "learning_rate": 3.2278788114046156e-05, "loss": 0.0001, "step": 54802 }, { "epoch": 1.06, "learning_rate": 3.2278141153788925e-05, "loss": 0.0002, "step": 54804 }, { "epoch": 1.06, "learning_rate": 3.2277494193531695e-05, "loss": 0.0, "step": 54806 }, { "epoch": 1.06, "learning_rate": 3.227684723327446e-05, "loss": 0.0, "step": 54808 }, { "epoch": 1.06, "learning_rate": 3.227620027301723e-05, "loss": 0.0, "step": 54810 }, { "epoch": 1.06, "learning_rate": 3.227555331276e-05, "loss": 0.0, "step": 54812 }, { "epoch": 1.06, "learning_rate": 3.2274906352502764e-05, "loss": 0.0, "step": 54814 }, { "epoch": 1.06, "learning_rate": 3.227425939224554e-05, "loss": 0.0, "step": 54816 }, { "epoch": 1.06, "learning_rate": 3.22736124319883e-05, "loss": 0.0, "step": 54818 }, { "epoch": 1.06, "learning_rate": 3.227296547173108e-05, "loss": 0.0006, "step": 54820 }, { "epoch": 1.06, "learning_rate": 3.227231851147384e-05, "loss": 0.0, "step": 54822 }, { "epoch": 1.06, "learning_rate": 3.227167155121661e-05, "loss": 0.0, "step": 54824 }, { "epoch": 1.06, "learning_rate": 3.227102459095938e-05, "loss": 0.0001, "step": 54826 }, { "epoch": 1.06, "learning_rate": 3.227037763070215e-05, "loss": 0.0, "step": 54828 }, { "epoch": 1.06, "learning_rate": 3.226973067044492e-05, "loss": 0.0, "step": 54830 }, { "epoch": 1.06, "learning_rate": 3.2269083710187686e-05, "loss": 0.0001, "step": 54832 }, { "epoch": 1.06, "learning_rate": 3.2268436749930455e-05, "loss": 0.0, "step": 54834 }, { "epoch": 1.06, "learning_rate": 3.226778978967322e-05, "loss": 0.0, "step": 54836 }, { "epoch": 1.06, "learning_rate": 3.226714282941599e-05, "loss": 0.0, "step": 54838 }, { "epoch": 1.06, "learning_rate": 3.2266495869158755e-05, "loss": 0.0, "step": 54840 }, { "epoch": 1.06, "learning_rate": 3.226584890890153e-05, "loss": 0.0001, "step": 54842 }, { "epoch": 1.06, "learning_rate": 3.2265201948644294e-05, "loss": 0.0, "step": 54844 }, { "epoch": 1.06, "learning_rate": 3.226455498838706e-05, "loss": 0.0018, "step": 54846 }, { "epoch": 1.06, "learning_rate": 3.226390802812984e-05, "loss": 0.0, "step": 54848 }, { "epoch": 1.06, "learning_rate": 3.22632610678726e-05, "loss": 0.0006, "step": 54850 }, { "epoch": 1.06, "learning_rate": 3.226261410761537e-05, "loss": 0.0076, "step": 54852 }, { "epoch": 1.06, "learning_rate": 3.226196714735814e-05, "loss": 0.0, "step": 54854 }, { "epoch": 1.06, "learning_rate": 3.226132018710091e-05, "loss": 0.0, "step": 54856 }, { "epoch": 1.06, "learning_rate": 3.226067322684368e-05, "loss": 0.0002, "step": 54858 }, { "epoch": 1.06, "learning_rate": 3.2260026266586446e-05, "loss": 0.0, "step": 54860 }, { "epoch": 1.06, "learning_rate": 3.2259379306329215e-05, "loss": 0.0001, "step": 54862 }, { "epoch": 1.06, "learning_rate": 3.2258732346071984e-05, "loss": 0.0, "step": 54864 }, { "epoch": 1.06, "learning_rate": 3.2258085385814754e-05, "loss": 0.0, "step": 54866 }, { "epoch": 1.06, "learning_rate": 3.2257438425557516e-05, "loss": 0.0, "step": 54868 }, { "epoch": 1.06, "learning_rate": 3.225679146530029e-05, "loss": 0.0, "step": 54870 }, { "epoch": 1.07, "learning_rate": 3.2256144505043054e-05, "loss": 0.0018, "step": 54872 }, { "epoch": 1.07, "learning_rate": 3.225549754478582e-05, "loss": 0.0002, "step": 54874 }, { "epoch": 1.07, "learning_rate": 3.225485058452859e-05, "loss": 0.0, "step": 54876 }, { "epoch": 1.07, "learning_rate": 3.225420362427136e-05, "loss": 0.0004, "step": 54878 }, { "epoch": 1.07, "learning_rate": 3.225355666401413e-05, "loss": 0.0071, "step": 54880 }, { "epoch": 1.07, "learning_rate": 3.22529097037569e-05, "loss": 0.0, "step": 54882 }, { "epoch": 1.07, "learning_rate": 3.225226274349967e-05, "loss": 0.0126, "step": 54884 }, { "epoch": 1.07, "learning_rate": 3.225161578324244e-05, "loss": 0.0, "step": 54886 }, { "epoch": 1.07, "learning_rate": 3.225096882298521e-05, "loss": 0.0014, "step": 54888 }, { "epoch": 1.07, "learning_rate": 3.2250321862727976e-05, "loss": 0.0, "step": 54890 }, { "epoch": 1.07, "learning_rate": 3.2249674902470745e-05, "loss": 0.0, "step": 54892 }, { "epoch": 1.07, "learning_rate": 3.2249027942213514e-05, "loss": 0.0, "step": 54894 }, { "epoch": 1.07, "learning_rate": 3.2248380981956276e-05, "loss": 0.0, "step": 54896 }, { "epoch": 1.07, "learning_rate": 3.224773402169905e-05, "loss": 0.0, "step": 54898 }, { "epoch": 1.07, "learning_rate": 3.2247087061441814e-05, "loss": 0.0, "step": 54900 }, { "epoch": 1.07, "learning_rate": 3.224644010118459e-05, "loss": 0.0, "step": 54902 }, { "epoch": 1.07, "learning_rate": 3.224579314092735e-05, "loss": 0.0, "step": 54904 }, { "epoch": 1.07, "learning_rate": 3.224514618067012e-05, "loss": 0.0, "step": 54906 }, { "epoch": 1.07, "learning_rate": 3.224449922041289e-05, "loss": 0.0, "step": 54908 }, { "epoch": 1.07, "learning_rate": 3.224385226015566e-05, "loss": 0.0001, "step": 54910 }, { "epoch": 1.07, "learning_rate": 3.224320529989843e-05, "loss": 0.0, "step": 54912 }, { "epoch": 1.07, "learning_rate": 3.22425583396412e-05, "loss": 0.0, "step": 54914 }, { "epoch": 1.07, "learning_rate": 3.224191137938397e-05, "loss": 0.0, "step": 54916 }, { "epoch": 1.07, "learning_rate": 3.224126441912673e-05, "loss": 0.0032, "step": 54918 }, { "epoch": 1.07, "learning_rate": 3.2240617458869505e-05, "loss": 0.0, "step": 54920 }, { "epoch": 1.07, "learning_rate": 3.223997049861227e-05, "loss": 0.0, "step": 54922 }, { "epoch": 1.07, "learning_rate": 3.2239323538355043e-05, "loss": 0.0, "step": 54924 }, { "epoch": 1.07, "learning_rate": 3.223867657809781e-05, "loss": 0.0, "step": 54926 }, { "epoch": 1.07, "learning_rate": 3.2238029617840575e-05, "loss": 0.0001, "step": 54928 }, { "epoch": 1.07, "learning_rate": 3.223738265758335e-05, "loss": 0.0, "step": 54930 }, { "epoch": 1.07, "learning_rate": 3.223673569732611e-05, "loss": 0.0, "step": 54932 }, { "epoch": 1.07, "learning_rate": 3.223608873706888e-05, "loss": 0.0001, "step": 54934 }, { "epoch": 1.07, "learning_rate": 3.223544177681165e-05, "loss": 0.0055, "step": 54936 }, { "epoch": 1.07, "learning_rate": 3.223479481655442e-05, "loss": 0.0, "step": 54938 }, { "epoch": 1.07, "learning_rate": 3.223414785629719e-05, "loss": 0.0001, "step": 54940 }, { "epoch": 1.07, "learning_rate": 3.223350089603996e-05, "loss": 0.0, "step": 54942 }, { "epoch": 1.07, "learning_rate": 3.223285393578273e-05, "loss": 0.0, "step": 54944 }, { "epoch": 1.07, "learning_rate": 3.2232206975525497e-05, "loss": 0.0003, "step": 54946 }, { "epoch": 1.07, "learning_rate": 3.2231560015268266e-05, "loss": 0.0001, "step": 54948 }, { "epoch": 1.07, "learning_rate": 3.223091305501103e-05, "loss": 0.0002, "step": 54950 }, { "epoch": 1.07, "learning_rate": 3.2230266094753804e-05, "loss": 0.0, "step": 54952 }, { "epoch": 1.07, "learning_rate": 3.2229619134496566e-05, "loss": 0.0, "step": 54954 }, { "epoch": 1.07, "learning_rate": 3.2228972174239335e-05, "loss": 0.0, "step": 54956 }, { "epoch": 1.07, "learning_rate": 3.222832521398211e-05, "loss": 0.0, "step": 54958 }, { "epoch": 1.07, "learning_rate": 3.2227678253724873e-05, "loss": 0.0, "step": 54960 }, { "epoch": 1.07, "learning_rate": 3.222703129346765e-05, "loss": 0.0, "step": 54962 }, { "epoch": 1.07, "learning_rate": 3.222638433321041e-05, "loss": 0.0001, "step": 54964 }, { "epoch": 1.07, "learning_rate": 3.222573737295318e-05, "loss": 0.0, "step": 54966 }, { "epoch": 1.07, "learning_rate": 3.222509041269595e-05, "loss": 0.0, "step": 54968 }, { "epoch": 1.07, "learning_rate": 3.222444345243872e-05, "loss": 0.0, "step": 54970 }, { "epoch": 1.07, "learning_rate": 3.222379649218149e-05, "loss": 0.0, "step": 54972 }, { "epoch": 1.07, "learning_rate": 3.222314953192426e-05, "loss": 0.0, "step": 54974 }, { "epoch": 1.07, "learning_rate": 3.2222502571667026e-05, "loss": 0.0016, "step": 54976 }, { "epoch": 1.07, "learning_rate": 3.222185561140979e-05, "loss": 0.013, "step": 54978 }, { "epoch": 1.07, "learning_rate": 3.2221208651152564e-05, "loss": 0.0014, "step": 54980 }, { "epoch": 1.07, "learning_rate": 3.2220561690895327e-05, "loss": 0.0006, "step": 54982 }, { "epoch": 1.07, "learning_rate": 3.22199147306381e-05, "loss": 0.0, "step": 54984 }, { "epoch": 1.07, "learning_rate": 3.2219267770380865e-05, "loss": 0.0066, "step": 54986 }, { "epoch": 1.07, "learning_rate": 3.2218620810123634e-05, "loss": 0.0, "step": 54988 }, { "epoch": 1.07, "learning_rate": 3.221797384986641e-05, "loss": 0.0, "step": 54990 }, { "epoch": 1.07, "learning_rate": 3.221732688960917e-05, "loss": 0.0, "step": 54992 }, { "epoch": 1.07, "learning_rate": 3.221667992935194e-05, "loss": 0.0, "step": 54994 }, { "epoch": 1.07, "learning_rate": 3.221603296909471e-05, "loss": 0.001, "step": 54996 }, { "epoch": 1.07, "learning_rate": 3.221538600883748e-05, "loss": 0.0002, "step": 54998 }, { "epoch": 1.07, "learning_rate": 3.221473904858024e-05, "loss": 0.0, "step": 55000 }, { "epoch": 1.07, "learning_rate": 3.221409208832302e-05, "loss": 0.0, "step": 55002 }, { "epoch": 1.07, "learning_rate": 3.2213445128065786e-05, "loss": 0.0, "step": 55004 }, { "epoch": 1.07, "learning_rate": 3.2212798167808556e-05, "loss": 0.0, "step": 55006 }, { "epoch": 1.07, "learning_rate": 3.2212151207551325e-05, "loss": 0.0, "step": 55008 }, { "epoch": 1.07, "learning_rate": 3.221150424729409e-05, "loss": 0.0, "step": 55010 }, { "epoch": 1.07, "learning_rate": 3.221085728703686e-05, "loss": 0.0, "step": 55012 }, { "epoch": 1.07, "learning_rate": 3.2210210326779625e-05, "loss": 0.0046, "step": 55014 }, { "epoch": 1.07, "learning_rate": 3.2209563366522394e-05, "loss": 0.0, "step": 55016 }, { "epoch": 1.07, "learning_rate": 3.220891640626516e-05, "loss": 0.0001, "step": 55018 }, { "epoch": 1.07, "learning_rate": 3.220826944600793e-05, "loss": 0.0, "step": 55020 }, { "epoch": 1.07, "learning_rate": 3.22076224857507e-05, "loss": 0.0, "step": 55022 }, { "epoch": 1.07, "learning_rate": 3.220697552549347e-05, "loss": 0.0, "step": 55024 }, { "epoch": 1.07, "learning_rate": 3.220632856523624e-05, "loss": 0.0002, "step": 55026 }, { "epoch": 1.07, "learning_rate": 3.220568160497901e-05, "loss": 0.0, "step": 55028 }, { "epoch": 1.07, "learning_rate": 3.220503464472178e-05, "loss": 0.0, "step": 55030 }, { "epoch": 1.07, "learning_rate": 3.220438768446454e-05, "loss": 0.0, "step": 55032 }, { "epoch": 1.07, "learning_rate": 3.2203740724207316e-05, "loss": 0.0, "step": 55034 }, { "epoch": 1.07, "learning_rate": 3.2203093763950085e-05, "loss": 0.0, "step": 55036 }, { "epoch": 1.07, "learning_rate": 3.220244680369285e-05, "loss": 0.0, "step": 55038 }, { "epoch": 1.07, "learning_rate": 3.220179984343562e-05, "loss": 0.0, "step": 55040 }, { "epoch": 1.07, "learning_rate": 3.2201152883178386e-05, "loss": 0.0, "step": 55042 }, { "epoch": 1.07, "learning_rate": 3.220050592292116e-05, "loss": 0.0014, "step": 55044 }, { "epoch": 1.07, "learning_rate": 3.2199858962663924e-05, "loss": 0.0, "step": 55046 }, { "epoch": 1.07, "learning_rate": 3.219921200240669e-05, "loss": 0.003, "step": 55048 }, { "epoch": 1.07, "learning_rate": 3.219856504214946e-05, "loss": 0.0, "step": 55050 }, { "epoch": 1.07, "learning_rate": 3.219791808189223e-05, "loss": 0.0, "step": 55052 }, { "epoch": 1.07, "learning_rate": 3.2197271121635e-05, "loss": 0.0004, "step": 55054 }, { "epoch": 1.07, "learning_rate": 3.219662416137777e-05, "loss": 0.0, "step": 55056 }, { "epoch": 1.07, "learning_rate": 3.219597720112054e-05, "loss": 0.0002, "step": 55058 }, { "epoch": 1.07, "learning_rate": 3.21953302408633e-05, "loss": 0.0, "step": 55060 }, { "epoch": 1.07, "learning_rate": 3.2194683280606076e-05, "loss": 0.0004, "step": 55062 }, { "epoch": 1.07, "learning_rate": 3.219403632034884e-05, "loss": 0.0, "step": 55064 }, { "epoch": 1.07, "learning_rate": 3.2193389360091615e-05, "loss": 0.0008, "step": 55066 }, { "epoch": 1.07, "learning_rate": 3.219274239983438e-05, "loss": 0.0, "step": 55068 }, { "epoch": 1.07, "learning_rate": 3.2192095439577146e-05, "loss": 0.0, "step": 55070 }, { "epoch": 1.07, "learning_rate": 3.219144847931992e-05, "loss": 0.0, "step": 55072 }, { "epoch": 1.07, "learning_rate": 3.2190801519062684e-05, "loss": 0.0, "step": 55074 }, { "epoch": 1.07, "learning_rate": 3.219015455880545e-05, "loss": 0.0, "step": 55076 }, { "epoch": 1.07, "learning_rate": 3.218950759854822e-05, "loss": 0.0, "step": 55078 }, { "epoch": 1.07, "learning_rate": 3.218886063829099e-05, "loss": 0.0028, "step": 55080 }, { "epoch": 1.07, "learning_rate": 3.218821367803376e-05, "loss": 0.0, "step": 55082 }, { "epoch": 1.07, "learning_rate": 3.218756671777653e-05, "loss": 0.0, "step": 55084 }, { "epoch": 1.07, "learning_rate": 3.21869197575193e-05, "loss": 0.0, "step": 55086 }, { "epoch": 1.07, "learning_rate": 3.218627279726207e-05, "loss": 0.0, "step": 55088 }, { "epoch": 1.07, "learning_rate": 3.218562583700484e-05, "loss": 0.0, "step": 55090 }, { "epoch": 1.07, "learning_rate": 3.21849788767476e-05, "loss": 0.0, "step": 55092 }, { "epoch": 1.07, "learning_rate": 3.2184331916490375e-05, "loss": 0.0, "step": 55094 }, { "epoch": 1.07, "learning_rate": 3.218368495623314e-05, "loss": 0.0, "step": 55096 }, { "epoch": 1.07, "learning_rate": 3.2183037995975906e-05, "loss": 0.0001, "step": 55098 }, { "epoch": 1.07, "learning_rate": 3.2182391035718675e-05, "loss": 0.0, "step": 55100 }, { "epoch": 1.07, "learning_rate": 3.2181744075461445e-05, "loss": 0.0001, "step": 55102 }, { "epoch": 1.07, "learning_rate": 3.2181097115204214e-05, "loss": 0.0, "step": 55104 }, { "epoch": 1.07, "learning_rate": 3.218045015494698e-05, "loss": 0.0, "step": 55106 }, { "epoch": 1.07, "learning_rate": 3.217980319468975e-05, "loss": 0.0, "step": 55108 }, { "epoch": 1.07, "learning_rate": 3.217915623443252e-05, "loss": 0.0001, "step": 55110 }, { "epoch": 1.07, "learning_rate": 3.217850927417529e-05, "loss": 0.0021, "step": 55112 }, { "epoch": 1.07, "learning_rate": 3.217786231391806e-05, "loss": 0.0, "step": 55114 }, { "epoch": 1.07, "learning_rate": 3.217721535366083e-05, "loss": 0.0, "step": 55116 }, { "epoch": 1.07, "learning_rate": 3.21765683934036e-05, "loss": 0.0001, "step": 55118 }, { "epoch": 1.07, "learning_rate": 3.217592143314636e-05, "loss": 0.0001, "step": 55120 }, { "epoch": 1.07, "learning_rate": 3.2175274472889135e-05, "loss": 0.0, "step": 55122 }, { "epoch": 1.07, "learning_rate": 3.21746275126319e-05, "loss": 0.0, "step": 55124 }, { "epoch": 1.07, "learning_rate": 3.2173980552374674e-05, "loss": 0.0, "step": 55126 }, { "epoch": 1.07, "learning_rate": 3.2173333592117436e-05, "loss": 0.0, "step": 55128 }, { "epoch": 1.07, "learning_rate": 3.2172686631860205e-05, "loss": 0.0, "step": 55130 }, { "epoch": 1.07, "learning_rate": 3.2172039671602974e-05, "loss": 0.008, "step": 55132 }, { "epoch": 1.07, "learning_rate": 3.217139271134574e-05, "loss": 0.0, "step": 55134 }, { "epoch": 1.07, "learning_rate": 3.217074575108851e-05, "loss": 0.0, "step": 55136 }, { "epoch": 1.07, "learning_rate": 3.217009879083128e-05, "loss": 0.0, "step": 55138 }, { "epoch": 1.07, "learning_rate": 3.216945183057405e-05, "loss": 0.0118, "step": 55140 }, { "epoch": 1.07, "learning_rate": 3.216880487031681e-05, "loss": 0.0044, "step": 55142 }, { "epoch": 1.07, "learning_rate": 3.216815791005959e-05, "loss": 0.0, "step": 55144 }, { "epoch": 1.07, "learning_rate": 3.216751094980235e-05, "loss": 0.0, "step": 55146 }, { "epoch": 1.07, "learning_rate": 3.216686398954513e-05, "loss": 0.0, "step": 55148 }, { "epoch": 1.07, "learning_rate": 3.2166217029287896e-05, "loss": 0.0029, "step": 55150 }, { "epoch": 1.07, "learning_rate": 3.216557006903066e-05, "loss": 0.0, "step": 55152 }, { "epoch": 1.07, "learning_rate": 3.2164923108773434e-05, "loss": 0.0, "step": 55154 }, { "epoch": 1.07, "learning_rate": 3.2164276148516196e-05, "loss": 0.0, "step": 55156 }, { "epoch": 1.07, "learning_rate": 3.2163629188258965e-05, "loss": 0.0, "step": 55158 }, { "epoch": 1.07, "learning_rate": 3.2162982228001734e-05, "loss": 0.0, "step": 55160 }, { "epoch": 1.07, "learning_rate": 3.2162335267744504e-05, "loss": 0.0002, "step": 55162 }, { "epoch": 1.07, "learning_rate": 3.216168830748727e-05, "loss": 0.0001, "step": 55164 }, { "epoch": 1.07, "learning_rate": 3.216104134723004e-05, "loss": 0.017, "step": 55166 }, { "epoch": 1.07, "learning_rate": 3.216039438697281e-05, "loss": 0.0, "step": 55168 }, { "epoch": 1.07, "learning_rate": 3.215974742671558e-05, "loss": 0.0073, "step": 55170 }, { "epoch": 1.07, "learning_rate": 3.215910046645835e-05, "loss": 0.0034, "step": 55172 }, { "epoch": 1.07, "learning_rate": 3.215845350620111e-05, "loss": 0.0, "step": 55174 }, { "epoch": 1.07, "learning_rate": 3.215780654594389e-05, "loss": 0.0004, "step": 55176 }, { "epoch": 1.07, "learning_rate": 3.215715958568665e-05, "loss": 0.0004, "step": 55178 }, { "epoch": 1.07, "learning_rate": 3.215651262542942e-05, "loss": 0.0043, "step": 55180 }, { "epoch": 1.07, "learning_rate": 3.2155865665172194e-05, "loss": 0.0, "step": 55182 }, { "epoch": 1.07, "learning_rate": 3.215521870491496e-05, "loss": 0.0, "step": 55184 }, { "epoch": 1.07, "learning_rate": 3.215457174465773e-05, "loss": 0.0, "step": 55186 }, { "epoch": 1.07, "learning_rate": 3.2153924784400495e-05, "loss": 0.0, "step": 55188 }, { "epoch": 1.07, "learning_rate": 3.2153277824143264e-05, "loss": 0.0, "step": 55190 }, { "epoch": 1.07, "learning_rate": 3.215263086388603e-05, "loss": 0.0, "step": 55192 }, { "epoch": 1.07, "learning_rate": 3.21519839036288e-05, "loss": 0.0, "step": 55194 }, { "epoch": 1.07, "learning_rate": 3.215133694337157e-05, "loss": 0.0015, "step": 55196 }, { "epoch": 1.07, "learning_rate": 3.215068998311434e-05, "loss": 0.0, "step": 55198 }, { "epoch": 1.07, "learning_rate": 3.215004302285711e-05, "loss": 0.0, "step": 55200 }, { "epoch": 1.07, "learning_rate": 3.214939606259987e-05, "loss": 0.0007, "step": 55202 }, { "epoch": 1.07, "learning_rate": 3.214874910234265e-05, "loss": 0.0, "step": 55204 }, { "epoch": 1.07, "learning_rate": 3.214810214208541e-05, "loss": 0.0, "step": 55206 }, { "epoch": 1.07, "learning_rate": 3.2147455181828186e-05, "loss": 0.0, "step": 55208 }, { "epoch": 1.07, "learning_rate": 3.214680822157095e-05, "loss": 0.0, "step": 55210 }, { "epoch": 1.07, "learning_rate": 3.214616126131372e-05, "loss": 0.0, "step": 55212 }, { "epoch": 1.07, "learning_rate": 3.214551430105649e-05, "loss": 0.0001, "step": 55214 }, { "epoch": 1.07, "learning_rate": 3.2144867340799255e-05, "loss": 0.0003, "step": 55216 }, { "epoch": 1.07, "learning_rate": 3.2144220380542024e-05, "loss": 0.0, "step": 55218 }, { "epoch": 1.07, "learning_rate": 3.2143573420284793e-05, "loss": 0.0, "step": 55220 }, { "epoch": 1.07, "learning_rate": 3.214292646002756e-05, "loss": 0.0, "step": 55222 }, { "epoch": 1.07, "learning_rate": 3.2142279499770325e-05, "loss": 0.0, "step": 55224 }, { "epoch": 1.07, "learning_rate": 3.21416325395131e-05, "loss": 0.0, "step": 55226 }, { "epoch": 1.07, "learning_rate": 3.214098557925587e-05, "loss": 0.0, "step": 55228 }, { "epoch": 1.07, "learning_rate": 3.214033861899864e-05, "loss": 0.0, "step": 55230 }, { "epoch": 1.07, "learning_rate": 3.213969165874141e-05, "loss": 0.0, "step": 55232 }, { "epoch": 1.07, "learning_rate": 3.213904469848417e-05, "loss": 0.002, "step": 55234 }, { "epoch": 1.07, "learning_rate": 3.2138397738226946e-05, "loss": 0.0, "step": 55236 }, { "epoch": 1.07, "learning_rate": 3.213775077796971e-05, "loss": 0.0, "step": 55238 }, { "epoch": 1.07, "learning_rate": 3.213710381771248e-05, "loss": 0.0, "step": 55240 }, { "epoch": 1.07, "learning_rate": 3.2136456857455247e-05, "loss": 0.0, "step": 55242 }, { "epoch": 1.07, "learning_rate": 3.2135809897198016e-05, "loss": 0.0001, "step": 55244 }, { "epoch": 1.07, "learning_rate": 3.2135162936940785e-05, "loss": 0.0, "step": 55246 }, { "epoch": 1.07, "learning_rate": 3.2134515976683554e-05, "loss": 0.0, "step": 55248 }, { "epoch": 1.07, "learning_rate": 3.213386901642632e-05, "loss": 0.0, "step": 55250 }, { "epoch": 1.07, "learning_rate": 3.213322205616909e-05, "loss": 0.0, "step": 55252 }, { "epoch": 1.07, "learning_rate": 3.213257509591186e-05, "loss": 0.0, "step": 55254 }, { "epoch": 1.07, "learning_rate": 3.213192813565462e-05, "loss": 0.0014, "step": 55256 }, { "epoch": 1.07, "learning_rate": 3.21312811753974e-05, "loss": 0.0, "step": 55258 }, { "epoch": 1.07, "learning_rate": 3.213063421514017e-05, "loss": 0.0, "step": 55260 }, { "epoch": 1.07, "learning_rate": 3.212998725488293e-05, "loss": 0.0, "step": 55262 }, { "epoch": 1.07, "learning_rate": 3.2129340294625706e-05, "loss": 0.0, "step": 55264 }, { "epoch": 1.07, "learning_rate": 3.212869333436847e-05, "loss": 0.0002, "step": 55266 }, { "epoch": 1.07, "learning_rate": 3.2128046374111245e-05, "loss": 0.0, "step": 55268 }, { "epoch": 1.07, "learning_rate": 3.212739941385401e-05, "loss": 0.0, "step": 55270 }, { "epoch": 1.07, "learning_rate": 3.2126752453596776e-05, "loss": 0.0, "step": 55272 }, { "epoch": 1.07, "learning_rate": 3.2126105493339545e-05, "loss": 0.0, "step": 55274 }, { "epoch": 1.07, "learning_rate": 3.2125458533082314e-05, "loss": 0.0, "step": 55276 }, { "epoch": 1.07, "learning_rate": 3.212481157282508e-05, "loss": 0.0, "step": 55278 }, { "epoch": 1.07, "learning_rate": 3.212416461256785e-05, "loss": 0.0, "step": 55280 }, { "epoch": 1.07, "learning_rate": 3.212351765231062e-05, "loss": 0.0, "step": 55282 }, { "epoch": 1.07, "learning_rate": 3.2122870692053384e-05, "loss": 0.0001, "step": 55284 }, { "epoch": 1.07, "learning_rate": 3.212222373179616e-05, "loss": 0.0, "step": 55286 }, { "epoch": 1.07, "learning_rate": 3.212157677153892e-05, "loss": 0.0109, "step": 55288 }, { "epoch": 1.07, "learning_rate": 3.21209298112817e-05, "loss": 0.0, "step": 55290 }, { "epoch": 1.07, "learning_rate": 3.212028285102447e-05, "loss": 0.0, "step": 55292 }, { "epoch": 1.07, "learning_rate": 3.211963589076723e-05, "loss": 0.0019, "step": 55294 }, { "epoch": 1.07, "learning_rate": 3.2118988930510005e-05, "loss": 0.0, "step": 55296 }, { "epoch": 1.07, "learning_rate": 3.211834197025277e-05, "loss": 0.0, "step": 55298 }, { "epoch": 1.07, "learning_rate": 3.2117695009995536e-05, "loss": 0.0, "step": 55300 }, { "epoch": 1.07, "learning_rate": 3.2117048049738306e-05, "loss": 0.0007, "step": 55302 }, { "epoch": 1.07, "learning_rate": 3.2116401089481075e-05, "loss": 0.0, "step": 55304 }, { "epoch": 1.07, "learning_rate": 3.2115754129223844e-05, "loss": 0.0, "step": 55306 }, { "epoch": 1.07, "learning_rate": 3.211510716896661e-05, "loss": 0.0, "step": 55308 }, { "epoch": 1.07, "learning_rate": 3.211446020870938e-05, "loss": 0.0, "step": 55310 }, { "epoch": 1.07, "learning_rate": 3.211381324845215e-05, "loss": 0.0, "step": 55312 }, { "epoch": 1.07, "learning_rate": 3.211316628819492e-05, "loss": 0.0, "step": 55314 }, { "epoch": 1.07, "learning_rate": 3.211251932793768e-05, "loss": 0.0009, "step": 55316 }, { "epoch": 1.07, "learning_rate": 3.211187236768046e-05, "loss": 0.0007, "step": 55318 }, { "epoch": 1.07, "learning_rate": 3.211122540742322e-05, "loss": 0.0, "step": 55320 }, { "epoch": 1.07, "learning_rate": 3.211057844716599e-05, "loss": 0.0, "step": 55322 }, { "epoch": 1.07, "learning_rate": 3.210993148690876e-05, "loss": 0.0, "step": 55324 }, { "epoch": 1.07, "learning_rate": 3.210928452665153e-05, "loss": 0.0001, "step": 55326 }, { "epoch": 1.07, "learning_rate": 3.2108637566394304e-05, "loss": 0.0, "step": 55328 }, { "epoch": 1.07, "learning_rate": 3.2107990606137066e-05, "loss": 0.0001, "step": 55330 }, { "epoch": 1.07, "learning_rate": 3.2107343645879835e-05, "loss": 0.0, "step": 55332 }, { "epoch": 1.07, "learning_rate": 3.2106696685622604e-05, "loss": 0.0, "step": 55334 }, { "epoch": 1.07, "learning_rate": 3.210604972536537e-05, "loss": 0.0001, "step": 55336 }, { "epoch": 1.07, "learning_rate": 3.210540276510814e-05, "loss": 0.0001, "step": 55338 }, { "epoch": 1.07, "learning_rate": 3.210475580485091e-05, "loss": 0.0, "step": 55340 }, { "epoch": 1.07, "learning_rate": 3.210410884459368e-05, "loss": 0.0, "step": 55342 }, { "epoch": 1.07, "learning_rate": 3.210346188433644e-05, "loss": 0.0, "step": 55344 }, { "epoch": 1.07, "learning_rate": 3.210281492407922e-05, "loss": 0.0057, "step": 55346 }, { "epoch": 1.07, "learning_rate": 3.210216796382198e-05, "loss": 0.0001, "step": 55348 }, { "epoch": 1.07, "learning_rate": 3.210152100356476e-05, "loss": 0.0, "step": 55350 }, { "epoch": 1.07, "learning_rate": 3.210087404330752e-05, "loss": 0.0, "step": 55352 }, { "epoch": 1.07, "learning_rate": 3.210022708305029e-05, "loss": 0.0003, "step": 55354 }, { "epoch": 1.07, "learning_rate": 3.209958012279306e-05, "loss": 0.0, "step": 55356 }, { "epoch": 1.07, "learning_rate": 3.2098933162535826e-05, "loss": 0.0, "step": 55358 }, { "epoch": 1.07, "learning_rate": 3.2098286202278595e-05, "loss": 0.0005, "step": 55360 }, { "epoch": 1.07, "learning_rate": 3.2097639242021365e-05, "loss": 0.0003, "step": 55362 }, { "epoch": 1.07, "learning_rate": 3.2096992281764134e-05, "loss": 0.0, "step": 55364 }, { "epoch": 1.07, "learning_rate": 3.2096345321506896e-05, "loss": 0.0004, "step": 55366 }, { "epoch": 1.07, "learning_rate": 3.209569836124967e-05, "loss": 0.0, "step": 55368 }, { "epoch": 1.07, "learning_rate": 3.209505140099244e-05, "loss": 0.0, "step": 55370 }, { "epoch": 1.07, "learning_rate": 3.209440444073521e-05, "loss": 0.0014, "step": 55372 }, { "epoch": 1.07, "learning_rate": 3.209375748047798e-05, "loss": 0.0152, "step": 55374 }, { "epoch": 1.07, "learning_rate": 3.209311052022074e-05, "loss": 0.0, "step": 55376 }, { "epoch": 1.07, "learning_rate": 3.209246355996352e-05, "loss": 0.0, "step": 55378 }, { "epoch": 1.07, "learning_rate": 3.209181659970628e-05, "loss": 0.017, "step": 55380 }, { "epoch": 1.07, "learning_rate": 3.209116963944905e-05, "loss": 0.0, "step": 55382 }, { "epoch": 1.07, "learning_rate": 3.209052267919182e-05, "loss": 0.0, "step": 55384 }, { "epoch": 1.07, "learning_rate": 3.208987571893459e-05, "loss": 0.0, "step": 55386 }, { "epoch": 1.08, "learning_rate": 3.2089228758677356e-05, "loss": 0.0003, "step": 55388 }, { "epoch": 1.08, "learning_rate": 3.2088581798420125e-05, "loss": 0.0004, "step": 55390 }, { "epoch": 1.08, "learning_rate": 3.2087934838162894e-05, "loss": 0.0, "step": 55392 }, { "epoch": 1.08, "learning_rate": 3.208728787790566e-05, "loss": 0.0, "step": 55394 }, { "epoch": 1.08, "learning_rate": 3.208664091764843e-05, "loss": 0.0, "step": 55396 }, { "epoch": 1.08, "learning_rate": 3.2085993957391194e-05, "loss": 0.0001, "step": 55398 }, { "epoch": 1.08, "learning_rate": 3.208534699713397e-05, "loss": 0.0008, "step": 55400 }, { "epoch": 1.08, "learning_rate": 3.208470003687673e-05, "loss": 0.0, "step": 55402 }, { "epoch": 1.08, "learning_rate": 3.20840530766195e-05, "loss": 0.0036, "step": 55404 }, { "epoch": 1.08, "learning_rate": 3.208340611636228e-05, "loss": 0.0007, "step": 55406 }, { "epoch": 1.08, "learning_rate": 3.208275915610504e-05, "loss": 0.0, "step": 55408 }, { "epoch": 1.08, "learning_rate": 3.2082112195847816e-05, "loss": 0.0044, "step": 55410 }, { "epoch": 1.08, "learning_rate": 3.208146523559058e-05, "loss": 0.004, "step": 55412 }, { "epoch": 1.08, "learning_rate": 3.208081827533335e-05, "loss": 0.0078, "step": 55414 }, { "epoch": 1.08, "learning_rate": 3.2080171315076116e-05, "loss": 0.0, "step": 55416 }, { "epoch": 1.08, "learning_rate": 3.2079524354818885e-05, "loss": 0.0, "step": 55418 }, { "epoch": 1.08, "learning_rate": 3.2078877394561654e-05, "loss": 0.0, "step": 55420 }, { "epoch": 1.08, "learning_rate": 3.2078230434304423e-05, "loss": 0.0019, "step": 55422 }, { "epoch": 1.08, "learning_rate": 3.207758347404719e-05, "loss": 0.0, "step": 55424 }, { "epoch": 1.08, "learning_rate": 3.2076936513789955e-05, "loss": 0.0, "step": 55426 }, { "epoch": 1.08, "learning_rate": 3.207628955353273e-05, "loss": 0.0001, "step": 55428 }, { "epoch": 1.08, "learning_rate": 3.207564259327549e-05, "loss": 0.0, "step": 55430 }, { "epoch": 1.08, "learning_rate": 3.207499563301827e-05, "loss": 0.001, "step": 55432 }, { "epoch": 1.08, "learning_rate": 3.207434867276103e-05, "loss": 0.0, "step": 55434 }, { "epoch": 1.08, "learning_rate": 3.20737017125038e-05, "loss": 0.0, "step": 55436 }, { "epoch": 1.08, "learning_rate": 3.2073054752246576e-05, "loss": 0.0137, "step": 55438 }, { "epoch": 1.08, "learning_rate": 3.207240779198934e-05, "loss": 0.0, "step": 55440 }, { "epoch": 1.08, "learning_rate": 3.207176083173211e-05, "loss": 0.0, "step": 55442 }, { "epoch": 1.08, "learning_rate": 3.207111387147488e-05, "loss": 0.0, "step": 55444 }, { "epoch": 1.08, "learning_rate": 3.2070466911217646e-05, "loss": 0.0, "step": 55446 }, { "epoch": 1.08, "learning_rate": 3.206981995096041e-05, "loss": 0.0002, "step": 55448 }, { "epoch": 1.08, "learning_rate": 3.2069172990703184e-05, "loss": 0.0002, "step": 55450 }, { "epoch": 1.08, "learning_rate": 3.206852603044595e-05, "loss": 0.0, "step": 55452 }, { "epoch": 1.08, "learning_rate": 3.206787907018872e-05, "loss": 0.0, "step": 55454 }, { "epoch": 1.08, "learning_rate": 3.206723210993149e-05, "loss": 0.0, "step": 55456 }, { "epoch": 1.08, "learning_rate": 3.2066585149674253e-05, "loss": 0.0019, "step": 55458 }, { "epoch": 1.08, "learning_rate": 3.206593818941703e-05, "loss": 0.0, "step": 55460 }, { "epoch": 1.08, "learning_rate": 3.206529122915979e-05, "loss": 0.0, "step": 55462 }, { "epoch": 1.08, "learning_rate": 3.206464426890256e-05, "loss": 0.0, "step": 55464 }, { "epoch": 1.08, "learning_rate": 3.206399730864533e-05, "loss": 0.0, "step": 55466 }, { "epoch": 1.08, "learning_rate": 3.20633503483881e-05, "loss": 0.0007, "step": 55468 }, { "epoch": 1.08, "learning_rate": 3.2062703388130875e-05, "loss": 0.0, "step": 55470 }, { "epoch": 1.08, "learning_rate": 3.206205642787364e-05, "loss": 0.0, "step": 55472 }, { "epoch": 1.08, "learning_rate": 3.2061409467616406e-05, "loss": 0.0022, "step": 55474 }, { "epoch": 1.08, "learning_rate": 3.2060762507359175e-05, "loss": 0.0001, "step": 55476 }, { "epoch": 1.08, "learning_rate": 3.2060115547101944e-05, "loss": 0.0, "step": 55478 }, { "epoch": 1.08, "learning_rate": 3.2059468586844707e-05, "loss": 0.0, "step": 55480 }, { "epoch": 1.08, "learning_rate": 3.205882162658748e-05, "loss": 0.0, "step": 55482 }, { "epoch": 1.08, "learning_rate": 3.205817466633025e-05, "loss": 0.0, "step": 55484 }, { "epoch": 1.08, "learning_rate": 3.2057527706073014e-05, "loss": 0.0, "step": 55486 }, { "epoch": 1.08, "learning_rate": 3.205688074581579e-05, "loss": 0.0005, "step": 55488 }, { "epoch": 1.08, "learning_rate": 3.205623378555855e-05, "loss": 0.0, "step": 55490 }, { "epoch": 1.08, "learning_rate": 3.205558682530133e-05, "loss": 0.0, "step": 55492 }, { "epoch": 1.08, "learning_rate": 3.205493986504409e-05, "loss": 0.0, "step": 55494 }, { "epoch": 1.08, "learning_rate": 3.205429290478686e-05, "loss": 0.0, "step": 55496 }, { "epoch": 1.08, "learning_rate": 3.205364594452963e-05, "loss": 0.0, "step": 55498 }, { "epoch": 1.08, "learning_rate": 3.20529989842724e-05, "loss": 0.0, "step": 55500 }, { "epoch": 1.08, "learning_rate": 3.2052352024015167e-05, "loss": 0.0001, "step": 55502 }, { "epoch": 1.08, "learning_rate": 3.2051705063757936e-05, "loss": 0.0, "step": 55504 }, { "epoch": 1.08, "learning_rate": 3.2051058103500705e-05, "loss": 0.0, "step": 55506 }, { "epoch": 1.08, "learning_rate": 3.205041114324347e-05, "loss": 0.0, "step": 55508 }, { "epoch": 1.08, "learning_rate": 3.204976418298624e-05, "loss": 0.0, "step": 55510 }, { "epoch": 1.08, "learning_rate": 3.2049117222729005e-05, "loss": 0.0008, "step": 55512 }, { "epoch": 1.08, "learning_rate": 3.204847026247178e-05, "loss": 0.0, "step": 55514 }, { "epoch": 1.08, "learning_rate": 3.204782330221455e-05, "loss": 0.0041, "step": 55516 }, { "epoch": 1.08, "learning_rate": 3.204717634195731e-05, "loss": 0.0, "step": 55518 }, { "epoch": 1.08, "learning_rate": 3.204652938170009e-05, "loss": 0.0, "step": 55520 }, { "epoch": 1.08, "learning_rate": 3.204588242144285e-05, "loss": 0.0, "step": 55522 }, { "epoch": 1.08, "learning_rate": 3.204523546118562e-05, "loss": 0.0, "step": 55524 }, { "epoch": 1.08, "learning_rate": 3.204458850092839e-05, "loss": 0.0, "step": 55526 }, { "epoch": 1.08, "learning_rate": 3.204394154067116e-05, "loss": 0.0, "step": 55528 }, { "epoch": 1.08, "learning_rate": 3.204329458041393e-05, "loss": 0.009, "step": 55530 }, { "epoch": 1.08, "learning_rate": 3.2042647620156696e-05, "loss": 0.0164, "step": 55532 }, { "epoch": 1.08, "learning_rate": 3.2042000659899465e-05, "loss": 0.0, "step": 55534 }, { "epoch": 1.08, "learning_rate": 3.2041353699642234e-05, "loss": 0.0, "step": 55536 }, { "epoch": 1.08, "learning_rate": 3.2040706739385e-05, "loss": 0.0, "step": 55538 }, { "epoch": 1.08, "learning_rate": 3.2040059779127766e-05, "loss": 0.0, "step": 55540 }, { "epoch": 1.08, "learning_rate": 3.203941281887054e-05, "loss": 0.0017, "step": 55542 }, { "epoch": 1.08, "learning_rate": 3.2038765858613304e-05, "loss": 0.0055, "step": 55544 }, { "epoch": 1.08, "learning_rate": 3.203811889835607e-05, "loss": 0.0007, "step": 55546 }, { "epoch": 1.08, "learning_rate": 3.203747193809884e-05, "loss": 0.0163, "step": 55548 }, { "epoch": 1.08, "learning_rate": 3.203682497784161e-05, "loss": 0.0, "step": 55550 }, { "epoch": 1.08, "learning_rate": 3.203617801758439e-05, "loss": 0.0001, "step": 55552 }, { "epoch": 1.08, "learning_rate": 3.203553105732715e-05, "loss": 0.0, "step": 55554 }, { "epoch": 1.08, "learning_rate": 3.203488409706992e-05, "loss": 0.0, "step": 55556 }, { "epoch": 1.08, "learning_rate": 3.203423713681269e-05, "loss": 0.0, "step": 55558 }, { "epoch": 1.08, "learning_rate": 3.2033590176555456e-05, "loss": 0.0, "step": 55560 }, { "epoch": 1.08, "learning_rate": 3.2032943216298226e-05, "loss": 0.0, "step": 55562 }, { "epoch": 1.08, "learning_rate": 3.2032296256040995e-05, "loss": 0.0, "step": 55564 }, { "epoch": 1.08, "learning_rate": 3.2031649295783764e-05, "loss": 0.0, "step": 55566 }, { "epoch": 1.08, "learning_rate": 3.2031002335526526e-05, "loss": 0.0, "step": 55568 }, { "epoch": 1.08, "learning_rate": 3.20303553752693e-05, "loss": 0.0, "step": 55570 }, { "epoch": 1.08, "learning_rate": 3.2029708415012064e-05, "loss": 0.0, "step": 55572 }, { "epoch": 1.08, "learning_rate": 3.202906145475484e-05, "loss": 0.0, "step": 55574 }, { "epoch": 1.08, "learning_rate": 3.20284144944976e-05, "loss": 0.0, "step": 55576 }, { "epoch": 1.08, "learning_rate": 3.202776753424037e-05, "loss": 0.0, "step": 55578 }, { "epoch": 1.08, "learning_rate": 3.202712057398314e-05, "loss": 0.0, "step": 55580 }, { "epoch": 1.08, "learning_rate": 3.202647361372591e-05, "loss": 0.0, "step": 55582 }, { "epoch": 1.08, "learning_rate": 3.202582665346868e-05, "loss": 0.0, "step": 55584 }, { "epoch": 1.08, "learning_rate": 3.202517969321145e-05, "loss": 0.0055, "step": 55586 }, { "epoch": 1.08, "learning_rate": 3.202453273295422e-05, "loss": 0.0, "step": 55588 }, { "epoch": 1.08, "learning_rate": 3.202388577269698e-05, "loss": 0.0, "step": 55590 }, { "epoch": 1.08, "learning_rate": 3.2023238812439755e-05, "loss": 0.0, "step": 55592 }, { "epoch": 1.08, "learning_rate": 3.2022591852182524e-05, "loss": 0.0, "step": 55594 }, { "epoch": 1.08, "learning_rate": 3.202194489192529e-05, "loss": 0.0, "step": 55596 }, { "epoch": 1.08, "learning_rate": 3.202129793166806e-05, "loss": 0.0002, "step": 55598 }, { "epoch": 1.08, "learning_rate": 3.2020650971410825e-05, "loss": 0.0022, "step": 55600 }, { "epoch": 1.08, "learning_rate": 3.20200040111536e-05, "loss": 0.0148, "step": 55602 }, { "epoch": 1.08, "learning_rate": 3.201935705089636e-05, "loss": 0.0, "step": 55604 }, { "epoch": 1.08, "learning_rate": 3.201871009063913e-05, "loss": 0.0061, "step": 55606 }, { "epoch": 1.08, "learning_rate": 3.20180631303819e-05, "loss": 0.0, "step": 55608 }, { "epoch": 1.08, "learning_rate": 3.201741617012467e-05, "loss": 0.0013, "step": 55610 }, { "epoch": 1.08, "learning_rate": 3.201676920986744e-05, "loss": 0.0, "step": 55612 }, { "epoch": 1.08, "learning_rate": 3.201612224961021e-05, "loss": 0.0005, "step": 55614 }, { "epoch": 1.08, "learning_rate": 3.201547528935298e-05, "loss": 0.0, "step": 55616 }, { "epoch": 1.08, "learning_rate": 3.2014828329095746e-05, "loss": 0.0, "step": 55618 }, { "epoch": 1.08, "learning_rate": 3.2014181368838515e-05, "loss": 0.0007, "step": 55620 }, { "epoch": 1.08, "learning_rate": 3.201353440858128e-05, "loss": 0.0, "step": 55622 }, { "epoch": 1.08, "learning_rate": 3.2012887448324054e-05, "loss": 0.0, "step": 55624 }, { "epoch": 1.08, "learning_rate": 3.2012240488066816e-05, "loss": 0.0, "step": 55626 }, { "epoch": 1.08, "learning_rate": 3.2011593527809585e-05, "loss": 0.0038, "step": 55628 }, { "epoch": 1.08, "learning_rate": 3.201094656755236e-05, "loss": 0.008, "step": 55630 }, { "epoch": 1.08, "learning_rate": 3.201029960729512e-05, "loss": 0.0, "step": 55632 }, { "epoch": 1.08, "learning_rate": 3.20096526470379e-05, "loss": 0.0, "step": 55634 }, { "epoch": 1.08, "learning_rate": 3.200900568678066e-05, "loss": 0.0, "step": 55636 }, { "epoch": 1.08, "learning_rate": 3.200835872652343e-05, "loss": 0.0, "step": 55638 }, { "epoch": 1.08, "learning_rate": 3.20077117662662e-05, "loss": 0.0, "step": 55640 }, { "epoch": 1.08, "learning_rate": 3.200706480600897e-05, "loss": 0.0, "step": 55642 }, { "epoch": 1.08, "learning_rate": 3.200641784575174e-05, "loss": 0.0, "step": 55644 }, { "epoch": 1.08, "learning_rate": 3.200577088549451e-05, "loss": 0.0, "step": 55646 }, { "epoch": 1.08, "learning_rate": 3.2005123925237276e-05, "loss": 0.0, "step": 55648 }, { "epoch": 1.08, "learning_rate": 3.200447696498004e-05, "loss": 0.0, "step": 55650 }, { "epoch": 1.08, "learning_rate": 3.2003830004722814e-05, "loss": 0.0, "step": 55652 }, { "epoch": 1.08, "learning_rate": 3.2003183044465576e-05, "loss": 0.0001, "step": 55654 }, { "epoch": 1.08, "learning_rate": 3.200253608420835e-05, "loss": 0.0002, "step": 55656 }, { "epoch": 1.08, "learning_rate": 3.2001889123951114e-05, "loss": 0.0008, "step": 55658 }, { "epoch": 1.08, "learning_rate": 3.2001242163693884e-05, "loss": 0.0, "step": 55660 }, { "epoch": 1.08, "learning_rate": 3.200059520343666e-05, "loss": 0.0, "step": 55662 }, { "epoch": 1.08, "learning_rate": 3.199994824317942e-05, "loss": 0.0, "step": 55664 }, { "epoch": 1.08, "learning_rate": 3.199930128292219e-05, "loss": 0.005, "step": 55666 }, { "epoch": 1.08, "learning_rate": 3.199865432266496e-05, "loss": 0.0, "step": 55668 }, { "epoch": 1.08, "learning_rate": 3.199800736240773e-05, "loss": 0.0, "step": 55670 }, { "epoch": 1.08, "learning_rate": 3.19973604021505e-05, "loss": 0.0001, "step": 55672 }, { "epoch": 1.08, "learning_rate": 3.199671344189327e-05, "loss": 0.0, "step": 55674 }, { "epoch": 1.08, "learning_rate": 3.1996066481636036e-05, "loss": 0.0001, "step": 55676 }, { "epoch": 1.08, "learning_rate": 3.1995419521378805e-05, "loss": 0.0001, "step": 55678 }, { "epoch": 1.08, "learning_rate": 3.1994772561121574e-05, "loss": 0.0, "step": 55680 }, { "epoch": 1.08, "learning_rate": 3.199412560086434e-05, "loss": 0.0, "step": 55682 }, { "epoch": 1.08, "learning_rate": 3.199347864060711e-05, "loss": 0.0047, "step": 55684 }, { "epoch": 1.08, "learning_rate": 3.1992831680349875e-05, "loss": 0.0, "step": 55686 }, { "epoch": 1.08, "learning_rate": 3.1992184720092644e-05, "loss": 0.0, "step": 55688 }, { "epoch": 1.08, "learning_rate": 3.199153775983541e-05, "loss": 0.0003, "step": 55690 }, { "epoch": 1.08, "learning_rate": 3.199089079957818e-05, "loss": 0.0001, "step": 55692 }, { "epoch": 1.08, "learning_rate": 3.199024383932096e-05, "loss": 0.0025, "step": 55694 }, { "epoch": 1.08, "learning_rate": 3.198959687906372e-05, "loss": 0.0, "step": 55696 }, { "epoch": 1.08, "learning_rate": 3.198894991880649e-05, "loss": 0.0018, "step": 55698 }, { "epoch": 1.08, "learning_rate": 3.198830295854926e-05, "loss": 0.0007, "step": 55700 }, { "epoch": 1.08, "learning_rate": 3.198765599829203e-05, "loss": 0.0, "step": 55702 }, { "epoch": 1.08, "learning_rate": 3.198700903803479e-05, "loss": 0.0001, "step": 55704 }, { "epoch": 1.08, "learning_rate": 3.1986362077777566e-05, "loss": 0.0, "step": 55706 }, { "epoch": 1.08, "learning_rate": 3.1985715117520335e-05, "loss": 0.0001, "step": 55708 }, { "epoch": 1.08, "learning_rate": 3.19850681572631e-05, "loss": 0.0, "step": 55710 }, { "epoch": 1.08, "learning_rate": 3.198442119700587e-05, "loss": 0.0, "step": 55712 }, { "epoch": 1.08, "learning_rate": 3.1983774236748635e-05, "loss": 0.0, "step": 55714 }, { "epoch": 1.08, "learning_rate": 3.198312727649141e-05, "loss": 0.0, "step": 55716 }, { "epoch": 1.08, "learning_rate": 3.1982480316234173e-05, "loss": 0.0, "step": 55718 }, { "epoch": 1.08, "learning_rate": 3.198183335597694e-05, "loss": 0.0, "step": 55720 }, { "epoch": 1.08, "learning_rate": 3.198118639571971e-05, "loss": 0.0, "step": 55722 }, { "epoch": 1.08, "learning_rate": 3.198053943546248e-05, "loss": 0.0, "step": 55724 }, { "epoch": 1.08, "learning_rate": 3.197989247520525e-05, "loss": 0.0, "step": 55726 }, { "epoch": 1.08, "learning_rate": 3.197924551494802e-05, "loss": 0.0021, "step": 55728 }, { "epoch": 1.08, "learning_rate": 3.197859855469079e-05, "loss": 0.0008, "step": 55730 }, { "epoch": 1.08, "learning_rate": 3.197795159443355e-05, "loss": 0.0, "step": 55732 }, { "epoch": 1.08, "learning_rate": 3.1977304634176326e-05, "loss": 0.0001, "step": 55734 }, { "epoch": 1.08, "learning_rate": 3.197665767391909e-05, "loss": 0.0001, "step": 55736 }, { "epoch": 1.08, "learning_rate": 3.1976010713661864e-05, "loss": 0.0034, "step": 55738 }, { "epoch": 1.08, "learning_rate": 3.197536375340463e-05, "loss": 0.0073, "step": 55740 }, { "epoch": 1.08, "learning_rate": 3.1974716793147396e-05, "loss": 0.0021, "step": 55742 }, { "epoch": 1.08, "learning_rate": 3.197406983289017e-05, "loss": 0.0009, "step": 55744 }, { "epoch": 1.08, "learning_rate": 3.1973422872632934e-05, "loss": 0.0, "step": 55746 }, { "epoch": 1.08, "learning_rate": 3.19727759123757e-05, "loss": 0.0, "step": 55748 }, { "epoch": 1.08, "learning_rate": 3.197212895211847e-05, "loss": 0.0, "step": 55750 }, { "epoch": 1.08, "learning_rate": 3.197148199186124e-05, "loss": 0.001, "step": 55752 }, { "epoch": 1.08, "learning_rate": 3.197083503160401e-05, "loss": 0.0, "step": 55754 }, { "epoch": 1.08, "learning_rate": 3.197018807134678e-05, "loss": 0.0001, "step": 55756 }, { "epoch": 1.08, "learning_rate": 3.196954111108955e-05, "loss": 0.0001, "step": 55758 }, { "epoch": 1.08, "learning_rate": 3.196889415083232e-05, "loss": 0.0, "step": 55760 }, { "epoch": 1.08, "learning_rate": 3.1968247190575087e-05, "loss": 0.0001, "step": 55762 }, { "epoch": 1.08, "learning_rate": 3.196760023031785e-05, "loss": 0.0001, "step": 55764 }, { "epoch": 1.08, "learning_rate": 3.1966953270060625e-05, "loss": 0.0, "step": 55766 }, { "epoch": 1.08, "learning_rate": 3.196630630980339e-05, "loss": 0.0005, "step": 55768 }, { "epoch": 1.08, "learning_rate": 3.1965659349546156e-05, "loss": 0.0118, "step": 55770 }, { "epoch": 1.08, "learning_rate": 3.196501238928893e-05, "loss": 0.0, "step": 55772 }, { "epoch": 1.08, "learning_rate": 3.1964365429031694e-05, "loss": 0.0018, "step": 55774 }, { "epoch": 1.08, "learning_rate": 3.196371846877447e-05, "loss": 0.0056, "step": 55776 }, { "epoch": 1.08, "learning_rate": 3.196307150851723e-05, "loss": 0.0001, "step": 55778 }, { "epoch": 1.08, "learning_rate": 3.196242454826e-05, "loss": 0.0, "step": 55780 }, { "epoch": 1.08, "learning_rate": 3.196177758800277e-05, "loss": 0.0, "step": 55782 }, { "epoch": 1.08, "learning_rate": 3.196113062774554e-05, "loss": 0.0074, "step": 55784 }, { "epoch": 1.08, "learning_rate": 3.196048366748831e-05, "loss": 0.0, "step": 55786 }, { "epoch": 1.08, "learning_rate": 3.195983670723108e-05, "loss": 0.0, "step": 55788 }, { "epoch": 1.08, "learning_rate": 3.195918974697385e-05, "loss": 0.0001, "step": 55790 }, { "epoch": 1.08, "learning_rate": 3.195854278671661e-05, "loss": 0.0, "step": 55792 }, { "epoch": 1.08, "learning_rate": 3.1957895826459385e-05, "loss": 0.0084, "step": 55794 }, { "epoch": 1.08, "learning_rate": 3.195724886620215e-05, "loss": 0.0, "step": 55796 }, { "epoch": 1.08, "learning_rate": 3.195660190594492e-05, "loss": 0.0002, "step": 55798 }, { "epoch": 1.08, "learning_rate": 3.1955954945687686e-05, "loss": 0.0004, "step": 55800 }, { "epoch": 1.08, "learning_rate": 3.1955307985430455e-05, "loss": 0.0, "step": 55802 }, { "epoch": 1.08, "learning_rate": 3.1954661025173224e-05, "loss": 0.0, "step": 55804 }, { "epoch": 1.08, "learning_rate": 3.195401406491599e-05, "loss": 0.0074, "step": 55806 }, { "epoch": 1.08, "learning_rate": 3.195336710465876e-05, "loss": 0.0, "step": 55808 }, { "epoch": 1.08, "learning_rate": 3.195272014440153e-05, "loss": 0.0, "step": 55810 }, { "epoch": 1.08, "learning_rate": 3.19520731841443e-05, "loss": 0.0003, "step": 55812 }, { "epoch": 1.08, "learning_rate": 3.195142622388706e-05, "loss": 0.0002, "step": 55814 }, { "epoch": 1.08, "learning_rate": 3.195077926362984e-05, "loss": 0.0, "step": 55816 }, { "epoch": 1.08, "learning_rate": 3.195013230337261e-05, "loss": 0.0107, "step": 55818 }, { "epoch": 1.08, "learning_rate": 3.1949485343115376e-05, "loss": 0.0, "step": 55820 }, { "epoch": 1.08, "learning_rate": 3.1948838382858145e-05, "loss": 0.0, "step": 55822 }, { "epoch": 1.08, "learning_rate": 3.194819142260091e-05, "loss": 0.0002, "step": 55824 }, { "epoch": 1.08, "learning_rate": 3.1947544462343684e-05, "loss": 0.0, "step": 55826 }, { "epoch": 1.08, "learning_rate": 3.1946897502086446e-05, "loss": 0.0, "step": 55828 }, { "epoch": 1.08, "learning_rate": 3.1946250541829215e-05, "loss": 0.0, "step": 55830 }, { "epoch": 1.08, "learning_rate": 3.1945603581571984e-05, "loss": 0.0, "step": 55832 }, { "epoch": 1.08, "learning_rate": 3.194495662131475e-05, "loss": 0.0, "step": 55834 }, { "epoch": 1.08, "learning_rate": 3.194430966105752e-05, "loss": 0.0, "step": 55836 }, { "epoch": 1.08, "learning_rate": 3.194366270080029e-05, "loss": 0.0, "step": 55838 }, { "epoch": 1.08, "learning_rate": 3.194301574054306e-05, "loss": 0.0001, "step": 55840 }, { "epoch": 1.08, "learning_rate": 3.194236878028583e-05, "loss": 0.0001, "step": 55842 }, { "epoch": 1.08, "learning_rate": 3.19417218200286e-05, "loss": 0.0, "step": 55844 }, { "epoch": 1.08, "learning_rate": 3.194107485977136e-05, "loss": 0.0, "step": 55846 }, { "epoch": 1.08, "learning_rate": 3.194042789951414e-05, "loss": 0.0, "step": 55848 }, { "epoch": 1.08, "learning_rate": 3.1939780939256906e-05, "loss": 0.0034, "step": 55850 }, { "epoch": 1.08, "learning_rate": 3.193913397899967e-05, "loss": 0.0, "step": 55852 }, { "epoch": 1.08, "learning_rate": 3.1938487018742444e-05, "loss": 0.0, "step": 55854 }, { "epoch": 1.08, "learning_rate": 3.1937840058485206e-05, "loss": 0.0, "step": 55856 }, { "epoch": 1.08, "learning_rate": 3.193719309822798e-05, "loss": 0.0, "step": 55858 }, { "epoch": 1.08, "learning_rate": 3.1936546137970745e-05, "loss": 0.0, "step": 55860 }, { "epoch": 1.08, "learning_rate": 3.1935899177713514e-05, "loss": 0.0, "step": 55862 }, { "epoch": 1.08, "learning_rate": 3.193525221745628e-05, "loss": 0.0, "step": 55864 }, { "epoch": 1.08, "learning_rate": 3.193460525719905e-05, "loss": 0.0, "step": 55866 }, { "epoch": 1.08, "learning_rate": 3.193395829694182e-05, "loss": 0.0, "step": 55868 }, { "epoch": 1.08, "learning_rate": 3.193331133668459e-05, "loss": 0.0, "step": 55870 }, { "epoch": 1.08, "learning_rate": 3.193266437642736e-05, "loss": 0.0012, "step": 55872 }, { "epoch": 1.08, "learning_rate": 3.193201741617012e-05, "loss": 0.0, "step": 55874 }, { "epoch": 1.08, "learning_rate": 3.19313704559129e-05, "loss": 0.0, "step": 55876 }, { "epoch": 1.08, "learning_rate": 3.193072349565566e-05, "loss": 0.0, "step": 55878 }, { "epoch": 1.08, "learning_rate": 3.1930076535398435e-05, "loss": 0.0, "step": 55880 }, { "epoch": 1.08, "learning_rate": 3.19294295751412e-05, "loss": 0.0475, "step": 55882 }, { "epoch": 1.08, "learning_rate": 3.192878261488397e-05, "loss": 0.0002, "step": 55884 }, { "epoch": 1.08, "learning_rate": 3.192813565462674e-05, "loss": 0.0, "step": 55886 }, { "epoch": 1.08, "learning_rate": 3.1927488694369505e-05, "loss": 0.0, "step": 55888 }, { "epoch": 1.08, "learning_rate": 3.1926841734112274e-05, "loss": 0.0002, "step": 55890 }, { "epoch": 1.08, "learning_rate": 3.192619477385504e-05, "loss": 0.0002, "step": 55892 }, { "epoch": 1.08, "learning_rate": 3.192554781359781e-05, "loss": 0.0, "step": 55894 }, { "epoch": 1.08, "learning_rate": 3.192490085334058e-05, "loss": 0.0, "step": 55896 }, { "epoch": 1.08, "learning_rate": 3.192425389308335e-05, "loss": 0.0, "step": 55898 }, { "epoch": 1.08, "learning_rate": 3.192360693282612e-05, "loss": 0.0001, "step": 55900 }, { "epoch": 1.08, "learning_rate": 3.192295997256889e-05, "loss": 0.0, "step": 55902 }, { "epoch": 1.09, "learning_rate": 3.192231301231166e-05, "loss": 0.0016, "step": 55904 }, { "epoch": 1.09, "learning_rate": 3.192166605205442e-05, "loss": 0.0, "step": 55906 }, { "epoch": 1.09, "learning_rate": 3.1921019091797196e-05, "loss": 0.0, "step": 55908 }, { "epoch": 1.09, "learning_rate": 3.192037213153996e-05, "loss": 0.0, "step": 55910 }, { "epoch": 1.09, "learning_rate": 3.191972517128273e-05, "loss": 0.0, "step": 55912 }, { "epoch": 1.09, "learning_rate": 3.1919078211025496e-05, "loss": 0.0, "step": 55914 }, { "epoch": 1.09, "learning_rate": 3.1918431250768265e-05, "loss": 0.0062, "step": 55916 }, { "epoch": 1.09, "learning_rate": 3.191778429051104e-05, "loss": 0.0, "step": 55918 }, { "epoch": 1.09, "learning_rate": 3.1917137330253804e-05, "loss": 0.0, "step": 55920 }, { "epoch": 1.09, "learning_rate": 3.191649036999657e-05, "loss": 0.0006, "step": 55922 }, { "epoch": 1.09, "learning_rate": 3.191584340973934e-05, "loss": 0.0011, "step": 55924 }, { "epoch": 1.09, "learning_rate": 3.191519644948211e-05, "loss": 0.0, "step": 55926 }, { "epoch": 1.09, "learning_rate": 3.191454948922487e-05, "loss": 0.0, "step": 55928 }, { "epoch": 1.09, "learning_rate": 3.191390252896765e-05, "loss": 0.0019, "step": 55930 }, { "epoch": 1.09, "learning_rate": 3.191325556871042e-05, "loss": 0.0, "step": 55932 }, { "epoch": 1.09, "learning_rate": 3.191260860845318e-05, "loss": 0.0, "step": 55934 }, { "epoch": 1.09, "learning_rate": 3.1911961648195956e-05, "loss": 0.0001, "step": 55936 }, { "epoch": 1.09, "learning_rate": 3.191131468793872e-05, "loss": 0.0009, "step": 55938 }, { "epoch": 1.09, "learning_rate": 3.1910667727681494e-05, "loss": 0.0001, "step": 55940 }, { "epoch": 1.09, "learning_rate": 3.191002076742426e-05, "loss": 0.0, "step": 55942 }, { "epoch": 1.09, "learning_rate": 3.1909373807167026e-05, "loss": 0.0, "step": 55944 }, { "epoch": 1.09, "learning_rate": 3.1908726846909795e-05, "loss": 0.0069, "step": 55946 }, { "epoch": 1.09, "learning_rate": 3.1908079886652564e-05, "loss": 0.0, "step": 55948 }, { "epoch": 1.09, "learning_rate": 3.190743292639533e-05, "loss": 0.0, "step": 55950 }, { "epoch": 1.09, "learning_rate": 3.19067859661381e-05, "loss": 0.0, "step": 55952 }, { "epoch": 1.09, "learning_rate": 3.190613900588087e-05, "loss": 0.0, "step": 55954 }, { "epoch": 1.09, "learning_rate": 3.1905492045623633e-05, "loss": 0.0, "step": 55956 }, { "epoch": 1.09, "learning_rate": 3.190484508536641e-05, "loss": 0.0353, "step": 55958 }, { "epoch": 1.09, "learning_rate": 3.190419812510917e-05, "loss": 0.0, "step": 55960 }, { "epoch": 1.09, "learning_rate": 3.190355116485195e-05, "loss": 0.0009, "step": 55962 }, { "epoch": 1.09, "learning_rate": 3.1902904204594717e-05, "loss": 0.0, "step": 55964 }, { "epoch": 1.09, "learning_rate": 3.190225724433748e-05, "loss": 0.0, "step": 55966 }, { "epoch": 1.09, "learning_rate": 3.1901610284080255e-05, "loss": 0.0, "step": 55968 }, { "epoch": 1.09, "learning_rate": 3.190096332382302e-05, "loss": 0.0, "step": 55970 }, { "epoch": 1.09, "learning_rate": 3.1900316363565786e-05, "loss": 0.0, "step": 55972 }, { "epoch": 1.09, "learning_rate": 3.1899669403308555e-05, "loss": 0.0, "step": 55974 }, { "epoch": 1.09, "learning_rate": 3.1899022443051324e-05, "loss": 0.0, "step": 55976 }, { "epoch": 1.09, "learning_rate": 3.1898375482794093e-05, "loss": 0.0, "step": 55978 }, { "epoch": 1.09, "learning_rate": 3.189772852253686e-05, "loss": 0.0002, "step": 55980 }, { "epoch": 1.09, "learning_rate": 3.189708156227963e-05, "loss": 0.0001, "step": 55982 }, { "epoch": 1.09, "learning_rate": 3.18964346020224e-05, "loss": 0.0, "step": 55984 }, { "epoch": 1.09, "learning_rate": 3.189578764176517e-05, "loss": 0.0182, "step": 55986 }, { "epoch": 1.09, "learning_rate": 3.189514068150793e-05, "loss": 0.0, "step": 55988 }, { "epoch": 1.09, "learning_rate": 3.189449372125071e-05, "loss": 0.0, "step": 55990 }, { "epoch": 1.09, "learning_rate": 3.189384676099347e-05, "loss": 0.0, "step": 55992 }, { "epoch": 1.09, "learning_rate": 3.189319980073624e-05, "loss": 0.0, "step": 55994 }, { "epoch": 1.09, "learning_rate": 3.1892552840479015e-05, "loss": 0.0, "step": 55996 }, { "epoch": 1.09, "learning_rate": 3.189190588022178e-05, "loss": 0.0, "step": 55998 }, { "epoch": 1.09, "learning_rate": 3.189125891996455e-05, "loss": 0.0, "step": 56000 }, { "epoch": 1.09, "learning_rate": 3.1890611959707316e-05, "loss": 0.0, "step": 56002 }, { "epoch": 1.09, "learning_rate": 3.1889964999450085e-05, "loss": 0.0, "step": 56004 }, { "epoch": 1.09, "learning_rate": 3.1889318039192854e-05, "loss": 0.0, "step": 56006 }, { "epoch": 1.09, "learning_rate": 3.188867107893562e-05, "loss": 0.0002, "step": 56008 }, { "epoch": 1.09, "learning_rate": 3.188802411867839e-05, "loss": 0.0, "step": 56010 }, { "epoch": 1.09, "learning_rate": 3.188737715842116e-05, "loss": 0.0016, "step": 56012 }, { "epoch": 1.09, "learning_rate": 3.188673019816393e-05, "loss": 0.0, "step": 56014 }, { "epoch": 1.09, "learning_rate": 3.188608323790669e-05, "loss": 0.0013, "step": 56016 }, { "epoch": 1.09, "learning_rate": 3.188543627764947e-05, "loss": 0.0, "step": 56018 }, { "epoch": 1.09, "learning_rate": 3.188478931739223e-05, "loss": 0.0092, "step": 56020 }, { "epoch": 1.09, "learning_rate": 3.1884142357135006e-05, "loss": 0.0, "step": 56022 }, { "epoch": 1.09, "learning_rate": 3.188349539687777e-05, "loss": 0.0, "step": 56024 }, { "epoch": 1.09, "learning_rate": 3.188284843662054e-05, "loss": 0.0002, "step": 56026 }, { "epoch": 1.09, "learning_rate": 3.188220147636331e-05, "loss": 0.0, "step": 56028 }, { "epoch": 1.09, "learning_rate": 3.1881554516106076e-05, "loss": 0.0, "step": 56030 }, { "epoch": 1.09, "learning_rate": 3.1880907555848845e-05, "loss": 0.0, "step": 56032 }, { "epoch": 1.09, "learning_rate": 3.1880260595591614e-05, "loss": 0.0, "step": 56034 }, { "epoch": 1.09, "learning_rate": 3.187961363533438e-05, "loss": 0.0, "step": 56036 }, { "epoch": 1.09, "learning_rate": 3.187896667507715e-05, "loss": 0.0001, "step": 56038 }, { "epoch": 1.09, "learning_rate": 3.187831971481992e-05, "loss": 0.0, "step": 56040 }, { "epoch": 1.09, "learning_rate": 3.187767275456269e-05, "loss": 0.0, "step": 56042 }, { "epoch": 1.09, "learning_rate": 3.187702579430546e-05, "loss": 0.0135, "step": 56044 }, { "epoch": 1.09, "learning_rate": 3.187637883404823e-05, "loss": 0.0087, "step": 56046 }, { "epoch": 1.09, "learning_rate": 3.187573187379099e-05, "loss": 0.0, "step": 56048 }, { "epoch": 1.09, "learning_rate": 3.187508491353377e-05, "loss": 0.0005, "step": 56050 }, { "epoch": 1.09, "learning_rate": 3.187443795327653e-05, "loss": 0.0001, "step": 56052 }, { "epoch": 1.09, "learning_rate": 3.18737909930193e-05, "loss": 0.0003, "step": 56054 }, { "epoch": 1.09, "learning_rate": 3.187314403276207e-05, "loss": 0.0, "step": 56056 }, { "epoch": 1.09, "learning_rate": 3.1872497072504836e-05, "loss": 0.0065, "step": 56058 }, { "epoch": 1.09, "learning_rate": 3.1871850112247606e-05, "loss": 0.0, "step": 56060 }, { "epoch": 1.09, "learning_rate": 3.1871203151990375e-05, "loss": 0.0001, "step": 56062 }, { "epoch": 1.09, "learning_rate": 3.1870556191733144e-05, "loss": 0.0, "step": 56064 }, { "epoch": 1.09, "learning_rate": 3.186990923147591e-05, "loss": 0.0, "step": 56066 }, { "epoch": 1.09, "learning_rate": 3.186926227121868e-05, "loss": 0.0, "step": 56068 }, { "epoch": 1.09, "learning_rate": 3.1868615310961444e-05, "loss": 0.0, "step": 56070 }, { "epoch": 1.09, "learning_rate": 3.186796835070422e-05, "loss": 0.0, "step": 56072 }, { "epoch": 1.09, "learning_rate": 3.186732139044699e-05, "loss": 0.0076, "step": 56074 }, { "epoch": 1.09, "learning_rate": 3.186667443018975e-05, "loss": 0.0, "step": 56076 }, { "epoch": 1.09, "learning_rate": 3.186602746993253e-05, "loss": 0.0, "step": 56078 }, { "epoch": 1.09, "learning_rate": 3.186538050967529e-05, "loss": 0.0, "step": 56080 }, { "epoch": 1.09, "learning_rate": 3.1864733549418065e-05, "loss": 0.0279, "step": 56082 }, { "epoch": 1.09, "learning_rate": 3.186408658916083e-05, "loss": 0.0, "step": 56084 }, { "epoch": 1.09, "learning_rate": 3.18634396289036e-05, "loss": 0.0001, "step": 56086 }, { "epoch": 1.09, "learning_rate": 3.1862792668646366e-05, "loss": 0.0, "step": 56088 }, { "epoch": 1.09, "learning_rate": 3.1862145708389135e-05, "loss": 0.0, "step": 56090 }, { "epoch": 1.09, "learning_rate": 3.1861498748131904e-05, "loss": 0.0001, "step": 56092 }, { "epoch": 1.09, "learning_rate": 3.186085178787467e-05, "loss": 0.0, "step": 56094 }, { "epoch": 1.09, "learning_rate": 3.186020482761744e-05, "loss": 0.0, "step": 56096 }, { "epoch": 1.09, "learning_rate": 3.1859557867360205e-05, "loss": 0.0002, "step": 56098 }, { "epoch": 1.09, "learning_rate": 3.185891090710298e-05, "loss": 0.0, "step": 56100 }, { "epoch": 1.09, "learning_rate": 3.185826394684574e-05, "loss": 0.0002, "step": 56102 }, { "epoch": 1.09, "learning_rate": 3.185761698658852e-05, "loss": 0.0, "step": 56104 }, { "epoch": 1.09, "learning_rate": 3.185697002633128e-05, "loss": 0.0003, "step": 56106 }, { "epoch": 1.09, "learning_rate": 3.185632306607405e-05, "loss": 0.0029, "step": 56108 }, { "epoch": 1.09, "learning_rate": 3.1855676105816826e-05, "loss": 0.0, "step": 56110 }, { "epoch": 1.09, "learning_rate": 3.185502914555959e-05, "loss": 0.0, "step": 56112 }, { "epoch": 1.09, "learning_rate": 3.185438218530236e-05, "loss": 0.0, "step": 56114 }, { "epoch": 1.09, "learning_rate": 3.1853735225045126e-05, "loss": 0.0, "step": 56116 }, { "epoch": 1.09, "learning_rate": 3.1853088264787895e-05, "loss": 0.0, "step": 56118 }, { "epoch": 1.09, "learning_rate": 3.1852441304530665e-05, "loss": 0.0001, "step": 56120 }, { "epoch": 1.09, "learning_rate": 3.1851794344273434e-05, "loss": 0.0, "step": 56122 }, { "epoch": 1.09, "learning_rate": 3.18511473840162e-05, "loss": 0.0, "step": 56124 }, { "epoch": 1.09, "learning_rate": 3.185050042375897e-05, "loss": 0.0, "step": 56126 }, { "epoch": 1.09, "learning_rate": 3.184985346350174e-05, "loss": 0.0037, "step": 56128 }, { "epoch": 1.09, "learning_rate": 3.18492065032445e-05, "loss": 0.0, "step": 56130 }, { "epoch": 1.09, "learning_rate": 3.184855954298728e-05, "loss": 0.0, "step": 56132 }, { "epoch": 1.09, "learning_rate": 3.184791258273004e-05, "loss": 0.0, "step": 56134 }, { "epoch": 1.09, "learning_rate": 3.184726562247281e-05, "loss": 0.0001, "step": 56136 }, { "epoch": 1.09, "learning_rate": 3.184661866221558e-05, "loss": 0.0, "step": 56138 }, { "epoch": 1.09, "learning_rate": 3.184597170195835e-05, "loss": 0.0, "step": 56140 }, { "epoch": 1.09, "learning_rate": 3.1845324741701124e-05, "loss": 0.0, "step": 56142 }, { "epoch": 1.09, "learning_rate": 3.184467778144389e-05, "loss": 0.0, "step": 56144 }, { "epoch": 1.09, "learning_rate": 3.1844030821186656e-05, "loss": 0.0, "step": 56146 }, { "epoch": 1.09, "learning_rate": 3.1843383860929425e-05, "loss": 0.0, "step": 56148 }, { "epoch": 1.09, "learning_rate": 3.1842736900672194e-05, "loss": 0.0, "step": 56150 }, { "epoch": 1.09, "learning_rate": 3.184208994041496e-05, "loss": 0.0015, "step": 56152 }, { "epoch": 1.09, "learning_rate": 3.184144298015773e-05, "loss": 0.0, "step": 56154 }, { "epoch": 1.09, "learning_rate": 3.18407960199005e-05, "loss": 0.0082, "step": 56156 }, { "epoch": 1.09, "learning_rate": 3.1840149059643264e-05, "loss": 0.0, "step": 56158 }, { "epoch": 1.09, "learning_rate": 3.183950209938604e-05, "loss": 0.0, "step": 56160 }, { "epoch": 1.09, "learning_rate": 3.18388551391288e-05, "loss": 0.0, "step": 56162 }, { "epoch": 1.09, "learning_rate": 3.183820817887158e-05, "loss": 0.0, "step": 56164 }, { "epoch": 1.09, "learning_rate": 3.183756121861434e-05, "loss": 0.0, "step": 56166 }, { "epoch": 1.09, "learning_rate": 3.183691425835711e-05, "loss": 0.0322, "step": 56168 }, { "epoch": 1.09, "learning_rate": 3.183626729809988e-05, "loss": 0.0, "step": 56170 }, { "epoch": 1.09, "learning_rate": 3.183562033784265e-05, "loss": 0.0, "step": 56172 }, { "epoch": 1.09, "learning_rate": 3.1834973377585416e-05, "loss": 0.0014, "step": 56174 }, { "epoch": 1.09, "learning_rate": 3.1834326417328185e-05, "loss": 0.0001, "step": 56176 }, { "epoch": 1.09, "learning_rate": 3.1833679457070954e-05, "loss": 0.0, "step": 56178 }, { "epoch": 1.09, "learning_rate": 3.1833032496813723e-05, "loss": 0.0059, "step": 56180 }, { "epoch": 1.09, "learning_rate": 3.183238553655649e-05, "loss": 0.0, "step": 56182 }, { "epoch": 1.09, "learning_rate": 3.1831738576299255e-05, "loss": 0.0, "step": 56184 }, { "epoch": 1.09, "learning_rate": 3.183109161604203e-05, "loss": 0.0, "step": 56186 }, { "epoch": 1.09, "learning_rate": 3.18304446557848e-05, "loss": 0.0, "step": 56188 }, { "epoch": 1.09, "learning_rate": 3.182979769552756e-05, "loss": 0.0035, "step": 56190 }, { "epoch": 1.09, "learning_rate": 3.182915073527034e-05, "loss": 0.0002, "step": 56192 }, { "epoch": 1.09, "learning_rate": 3.18285037750131e-05, "loss": 0.0, "step": 56194 }, { "epoch": 1.09, "learning_rate": 3.182785681475587e-05, "loss": 0.0, "step": 56196 }, { "epoch": 1.09, "learning_rate": 3.182720985449864e-05, "loss": 0.0, "step": 56198 }, { "epoch": 1.09, "learning_rate": 3.182656289424141e-05, "loss": 0.0, "step": 56200 }, { "epoch": 1.09, "learning_rate": 3.182591593398418e-05, "loss": 0.0, "step": 56202 }, { "epoch": 1.09, "learning_rate": 3.1825268973726946e-05, "loss": 0.0, "step": 56204 }, { "epoch": 1.09, "learning_rate": 3.1824622013469715e-05, "loss": 0.004, "step": 56206 }, { "epoch": 1.09, "learning_rate": 3.1823975053212484e-05, "loss": 0.0, "step": 56208 }, { "epoch": 1.09, "learning_rate": 3.182332809295525e-05, "loss": 0.0, "step": 56210 }, { "epoch": 1.09, "learning_rate": 3.1822681132698015e-05, "loss": 0.0001, "step": 56212 }, { "epoch": 1.09, "learning_rate": 3.182203417244079e-05, "loss": 0.0, "step": 56214 }, { "epoch": 1.09, "learning_rate": 3.1821387212183553e-05, "loss": 0.0, "step": 56216 }, { "epoch": 1.09, "learning_rate": 3.182074025192632e-05, "loss": 0.0, "step": 56218 }, { "epoch": 1.09, "learning_rate": 3.18200932916691e-05, "loss": 0.0, "step": 56220 }, { "epoch": 1.09, "learning_rate": 3.181944633141186e-05, "loss": 0.0002, "step": 56222 }, { "epoch": 1.09, "learning_rate": 3.1818799371154637e-05, "loss": 0.0001, "step": 56224 }, { "epoch": 1.09, "learning_rate": 3.18181524108974e-05, "loss": 0.0, "step": 56226 }, { "epoch": 1.09, "learning_rate": 3.181750545064017e-05, "loss": 0.0, "step": 56228 }, { "epoch": 1.09, "learning_rate": 3.181685849038294e-05, "loss": 0.0, "step": 56230 }, { "epoch": 1.09, "learning_rate": 3.1816211530125706e-05, "loss": 0.0, "step": 56232 }, { "epoch": 1.09, "learning_rate": 3.1815564569868475e-05, "loss": 0.0011, "step": 56234 }, { "epoch": 1.09, "learning_rate": 3.1814917609611244e-05, "loss": 0.0, "step": 56236 }, { "epoch": 1.09, "learning_rate": 3.1814270649354013e-05, "loss": 0.0007, "step": 56238 }, { "epoch": 1.09, "learning_rate": 3.1813623689096776e-05, "loss": 0.0, "step": 56240 }, { "epoch": 1.09, "learning_rate": 3.181297672883955e-05, "loss": 0.0047, "step": 56242 }, { "epoch": 1.09, "learning_rate": 3.1812329768582314e-05, "loss": 0.0078, "step": 56244 }, { "epoch": 1.09, "learning_rate": 3.181168280832509e-05, "loss": 0.0, "step": 56246 }, { "epoch": 1.09, "learning_rate": 3.181103584806785e-05, "loss": 0.0001, "step": 56248 }, { "epoch": 1.09, "learning_rate": 3.181038888781062e-05, "loss": 0.0135, "step": 56250 }, { "epoch": 1.09, "learning_rate": 3.18097419275534e-05, "loss": 0.0005, "step": 56252 }, { "epoch": 1.09, "learning_rate": 3.180909496729616e-05, "loss": 0.0001, "step": 56254 }, { "epoch": 1.09, "learning_rate": 3.180844800703893e-05, "loss": 0.0, "step": 56256 }, { "epoch": 1.09, "learning_rate": 3.18078010467817e-05, "loss": 0.0, "step": 56258 }, { "epoch": 1.09, "learning_rate": 3.1807154086524467e-05, "loss": 0.0, "step": 56260 }, { "epoch": 1.09, "learning_rate": 3.1806507126267236e-05, "loss": 0.0, "step": 56262 }, { "epoch": 1.09, "learning_rate": 3.1805860166010005e-05, "loss": 0.0, "step": 56264 }, { "epoch": 1.09, "learning_rate": 3.1805213205752774e-05, "loss": 0.0, "step": 56266 }, { "epoch": 1.09, "learning_rate": 3.180456624549554e-05, "loss": 0.0, "step": 56268 }, { "epoch": 1.09, "learning_rate": 3.180391928523831e-05, "loss": 0.0039, "step": 56270 }, { "epoch": 1.09, "learning_rate": 3.1803272324981074e-05, "loss": 0.0003, "step": 56272 }, { "epoch": 1.09, "learning_rate": 3.180262536472385e-05, "loss": 0.0003, "step": 56274 }, { "epoch": 1.09, "learning_rate": 3.180197840446661e-05, "loss": 0.0, "step": 56276 }, { "epoch": 1.09, "learning_rate": 3.180133144420938e-05, "loss": 0.0, "step": 56278 }, { "epoch": 1.09, "learning_rate": 3.180068448395215e-05, "loss": 0.0, "step": 56280 }, { "epoch": 1.09, "learning_rate": 3.180003752369492e-05, "loss": 0.0, "step": 56282 }, { "epoch": 1.09, "learning_rate": 3.179939056343769e-05, "loss": 0.0, "step": 56284 }, { "epoch": 1.09, "learning_rate": 3.179874360318046e-05, "loss": 0.0, "step": 56286 }, { "epoch": 1.09, "learning_rate": 3.179809664292323e-05, "loss": 0.0, "step": 56288 }, { "epoch": 1.09, "learning_rate": 3.1797449682665996e-05, "loss": 0.0, "step": 56290 }, { "epoch": 1.09, "learning_rate": 3.1796802722408765e-05, "loss": 0.0, "step": 56292 }, { "epoch": 1.09, "learning_rate": 3.179615576215153e-05, "loss": 0.0, "step": 56294 }, { "epoch": 1.09, "learning_rate": 3.17955088018943e-05, "loss": 0.0, "step": 56296 }, { "epoch": 1.09, "learning_rate": 3.179486184163707e-05, "loss": 0.0002, "step": 56298 }, { "epoch": 1.09, "learning_rate": 3.1794214881379835e-05, "loss": 0.0, "step": 56300 }, { "epoch": 1.09, "learning_rate": 3.179356792112261e-05, "loss": 0.0019, "step": 56302 }, { "epoch": 1.09, "learning_rate": 3.179292096086537e-05, "loss": 0.0, "step": 56304 }, { "epoch": 1.09, "learning_rate": 3.179227400060815e-05, "loss": 0.0059, "step": 56306 }, { "epoch": 1.09, "learning_rate": 3.179162704035091e-05, "loss": 0.0, "step": 56308 }, { "epoch": 1.09, "learning_rate": 3.179098008009368e-05, "loss": 0.0001, "step": 56310 }, { "epoch": 1.09, "learning_rate": 3.179033311983645e-05, "loss": 0.0, "step": 56312 }, { "epoch": 1.09, "learning_rate": 3.178968615957922e-05, "loss": 0.0011, "step": 56314 }, { "epoch": 1.09, "learning_rate": 3.178903919932199e-05, "loss": 0.0, "step": 56316 }, { "epoch": 1.09, "learning_rate": 3.1788392239064756e-05, "loss": 0.0007, "step": 56318 }, { "epoch": 1.09, "learning_rate": 3.1787745278807526e-05, "loss": 0.0, "step": 56320 }, { "epoch": 1.09, "learning_rate": 3.1787098318550295e-05, "loss": 0.0, "step": 56322 }, { "epoch": 1.09, "learning_rate": 3.1786451358293064e-05, "loss": 0.0059, "step": 56324 }, { "epoch": 1.09, "learning_rate": 3.1785804398035826e-05, "loss": 0.0, "step": 56326 }, { "epoch": 1.09, "learning_rate": 3.17851574377786e-05, "loss": 0.0001, "step": 56328 }, { "epoch": 1.09, "learning_rate": 3.178451047752137e-05, "loss": 0.0, "step": 56330 }, { "epoch": 1.09, "learning_rate": 3.178386351726413e-05, "loss": 0.0125, "step": 56332 }, { "epoch": 1.09, "learning_rate": 3.178321655700691e-05, "loss": 0.0056, "step": 56334 }, { "epoch": 1.09, "learning_rate": 3.178256959674967e-05, "loss": 0.0, "step": 56336 }, { "epoch": 1.09, "learning_rate": 3.178192263649244e-05, "loss": 0.0, "step": 56338 }, { "epoch": 1.09, "learning_rate": 3.178127567623521e-05, "loss": 0.0, "step": 56340 }, { "epoch": 1.09, "learning_rate": 3.178062871597798e-05, "loss": 0.0001, "step": 56342 }, { "epoch": 1.09, "learning_rate": 3.177998175572075e-05, "loss": 0.0001, "step": 56344 }, { "epoch": 1.09, "learning_rate": 3.177933479546352e-05, "loss": 0.0001, "step": 56346 }, { "epoch": 1.09, "learning_rate": 3.1778687835206286e-05, "loss": 0.0, "step": 56348 }, { "epoch": 1.09, "learning_rate": 3.1778040874949055e-05, "loss": 0.0, "step": 56350 }, { "epoch": 1.09, "learning_rate": 3.1777393914691824e-05, "loss": 0.0001, "step": 56352 }, { "epoch": 1.09, "learning_rate": 3.1776746954434586e-05, "loss": 0.0156, "step": 56354 }, { "epoch": 1.09, "learning_rate": 3.177609999417736e-05, "loss": 0.0, "step": 56356 }, { "epoch": 1.09, "learning_rate": 3.1775453033920125e-05, "loss": 0.0, "step": 56358 }, { "epoch": 1.09, "learning_rate": 3.1774806073662894e-05, "loss": 0.0, "step": 56360 }, { "epoch": 1.09, "learning_rate": 3.177415911340566e-05, "loss": 0.0, "step": 56362 }, { "epoch": 1.09, "learning_rate": 3.177351215314843e-05, "loss": 0.0004, "step": 56364 }, { "epoch": 1.09, "learning_rate": 3.177286519289121e-05, "loss": 0.0001, "step": 56366 }, { "epoch": 1.09, "learning_rate": 3.177221823263397e-05, "loss": 0.0006, "step": 56368 }, { "epoch": 1.09, "learning_rate": 3.177157127237674e-05, "loss": 0.0001, "step": 56370 }, { "epoch": 1.09, "learning_rate": 3.177092431211951e-05, "loss": 0.0, "step": 56372 }, { "epoch": 1.09, "learning_rate": 3.177027735186228e-05, "loss": 0.0001, "step": 56374 }, { "epoch": 1.09, "learning_rate": 3.1769630391605046e-05, "loss": 0.0, "step": 56376 }, { "epoch": 1.09, "learning_rate": 3.1768983431347815e-05, "loss": 0.0, "step": 56378 }, { "epoch": 1.09, "learning_rate": 3.1768336471090585e-05, "loss": 0.0, "step": 56380 }, { "epoch": 1.09, "learning_rate": 3.176768951083335e-05, "loss": 0.0, "step": 56382 }, { "epoch": 1.09, "learning_rate": 3.176704255057612e-05, "loss": 0.0, "step": 56384 }, { "epoch": 1.09, "learning_rate": 3.1766395590318885e-05, "loss": 0.0, "step": 56386 }, { "epoch": 1.09, "learning_rate": 3.176574863006166e-05, "loss": 0.0, "step": 56388 }, { "epoch": 1.09, "learning_rate": 3.176510166980442e-05, "loss": 0.0005, "step": 56390 }, { "epoch": 1.09, "learning_rate": 3.176445470954719e-05, "loss": 0.0, "step": 56392 }, { "epoch": 1.09, "learning_rate": 3.176380774928996e-05, "loss": 0.0001, "step": 56394 }, { "epoch": 1.09, "learning_rate": 3.176316078903273e-05, "loss": 0.0, "step": 56396 }, { "epoch": 1.09, "learning_rate": 3.17625138287755e-05, "loss": 0.0, "step": 56398 }, { "epoch": 1.09, "learning_rate": 3.176186686851827e-05, "loss": 0.0, "step": 56400 }, { "epoch": 1.09, "learning_rate": 3.176121990826104e-05, "loss": 0.0, "step": 56402 }, { "epoch": 1.09, "learning_rate": 3.176057294800381e-05, "loss": 0.0, "step": 56404 }, { "epoch": 1.09, "learning_rate": 3.1759925987746576e-05, "loss": 0.0284, "step": 56406 }, { "epoch": 1.09, "learning_rate": 3.175927902748934e-05, "loss": 0.0001, "step": 56408 }, { "epoch": 1.09, "learning_rate": 3.1758632067232114e-05, "loss": 0.0, "step": 56410 }, { "epoch": 1.09, "learning_rate": 3.175798510697488e-05, "loss": 0.0001, "step": 56412 }, { "epoch": 1.09, "learning_rate": 3.1757338146717645e-05, "loss": 0.0001, "step": 56414 }, { "epoch": 1.09, "learning_rate": 3.175669118646042e-05, "loss": 0.0, "step": 56416 }, { "epoch": 1.1, "learning_rate": 3.1756044226203184e-05, "loss": 0.0, "step": 56418 }, { "epoch": 1.1, "learning_rate": 3.175539726594595e-05, "loss": 0.0, "step": 56420 }, { "epoch": 1.1, "learning_rate": 3.175475030568872e-05, "loss": 0.0, "step": 56422 }, { "epoch": 1.1, "learning_rate": 3.175410334543149e-05, "loss": 0.0, "step": 56424 }, { "epoch": 1.1, "learning_rate": 3.175345638517426e-05, "loss": 0.0051, "step": 56426 }, { "epoch": 1.1, "learning_rate": 3.175280942491703e-05, "loss": 0.0, "step": 56428 }, { "epoch": 1.1, "learning_rate": 3.17521624646598e-05, "loss": 0.0, "step": 56430 }, { "epoch": 1.1, "learning_rate": 3.175151550440257e-05, "loss": 0.0, "step": 56432 }, { "epoch": 1.1, "learning_rate": 3.1750868544145336e-05, "loss": 0.0, "step": 56434 }, { "epoch": 1.1, "learning_rate": 3.17502215838881e-05, "loss": 0.0, "step": 56436 }, { "epoch": 1.1, "learning_rate": 3.1749574623630874e-05, "loss": 0.0, "step": 56438 }, { "epoch": 1.1, "learning_rate": 3.174892766337364e-05, "loss": 0.0002, "step": 56440 }, { "epoch": 1.1, "learning_rate": 3.1748280703116406e-05, "loss": 0.0, "step": 56442 }, { "epoch": 1.1, "learning_rate": 3.174763374285918e-05, "loss": 0.0, "step": 56444 }, { "epoch": 1.1, "learning_rate": 3.1746986782601944e-05, "loss": 0.0, "step": 56446 }, { "epoch": 1.1, "learning_rate": 3.174633982234472e-05, "loss": 0.0003, "step": 56448 }, { "epoch": 1.1, "learning_rate": 3.174569286208748e-05, "loss": 0.0, "step": 56450 }, { "epoch": 1.1, "learning_rate": 3.174504590183025e-05, "loss": 0.0, "step": 56452 }, { "epoch": 1.1, "learning_rate": 3.174439894157302e-05, "loss": 0.0, "step": 56454 }, { "epoch": 1.1, "learning_rate": 3.174375198131579e-05, "loss": 0.0, "step": 56456 }, { "epoch": 1.1, "learning_rate": 3.174310502105856e-05, "loss": 0.0, "step": 56458 }, { "epoch": 1.1, "learning_rate": 3.174245806080133e-05, "loss": 0.0, "step": 56460 }, { "epoch": 1.1, "learning_rate": 3.17418111005441e-05, "loss": 0.0, "step": 56462 }, { "epoch": 1.1, "learning_rate": 3.1741164140286866e-05, "loss": 0.0, "step": 56464 }, { "epoch": 1.1, "learning_rate": 3.1740517180029635e-05, "loss": 0.0, "step": 56466 }, { "epoch": 1.1, "learning_rate": 3.17398702197724e-05, "loss": 0.0, "step": 56468 }, { "epoch": 1.1, "learning_rate": 3.173922325951517e-05, "loss": 0.0001, "step": 56470 }, { "epoch": 1.1, "learning_rate": 3.1738576299257935e-05, "loss": 0.0017, "step": 56472 }, { "epoch": 1.1, "learning_rate": 3.1737929339000704e-05, "loss": 0.0, "step": 56474 }, { "epoch": 1.1, "learning_rate": 3.173728237874348e-05, "loss": 0.0, "step": 56476 }, { "epoch": 1.1, "learning_rate": 3.173663541848624e-05, "loss": 0.0005, "step": 56478 }, { "epoch": 1.1, "learning_rate": 3.173598845822901e-05, "loss": 0.0, "step": 56480 }, { "epoch": 1.1, "learning_rate": 3.173534149797178e-05, "loss": 0.0001, "step": 56482 }, { "epoch": 1.1, "learning_rate": 3.173469453771455e-05, "loss": 0.0008, "step": 56484 }, { "epoch": 1.1, "learning_rate": 3.173404757745732e-05, "loss": 0.0, "step": 56486 }, { "epoch": 1.1, "learning_rate": 3.173340061720009e-05, "loss": 0.0001, "step": 56488 }, { "epoch": 1.1, "learning_rate": 3.173275365694286e-05, "loss": 0.0, "step": 56490 }, { "epoch": 1.1, "learning_rate": 3.1732106696685626e-05, "loss": 0.0001, "step": 56492 }, { "epoch": 1.1, "learning_rate": 3.1731459736428395e-05, "loss": 0.0, "step": 56494 }, { "epoch": 1.1, "learning_rate": 3.173081277617116e-05, "loss": 0.0, "step": 56496 }, { "epoch": 1.1, "learning_rate": 3.173016581591393e-05, "loss": 0.0146, "step": 56498 }, { "epoch": 1.1, "learning_rate": 3.1729518855656696e-05, "loss": 0.0018, "step": 56500 }, { "epoch": 1.1, "learning_rate": 3.1728871895399465e-05, "loss": 0.0045, "step": 56502 }, { "epoch": 1.1, "learning_rate": 3.1728224935142234e-05, "loss": 0.0093, "step": 56504 }, { "epoch": 1.1, "learning_rate": 3.172790145501362e-05, "loss": 0.0761, "step": 56506 }, { "epoch": 1.1, "learning_rate": 3.172725449475639e-05, "loss": 0.0, "step": 56508 }, { "epoch": 1.1, "learning_rate": 3.1726607534499157e-05, "loss": 0.0, "step": 56510 }, { "epoch": 1.1, "learning_rate": 3.1725960574241926e-05, "loss": 0.0046, "step": 56512 }, { "epoch": 1.1, "learning_rate": 3.1725313613984695e-05, "loss": 0.0001, "step": 56514 }, { "epoch": 1.1, "learning_rate": 3.1724666653727464e-05, "loss": 0.0, "step": 56516 }, { "epoch": 1.1, "learning_rate": 3.172401969347023e-05, "loss": 0.0001, "step": 56518 }, { "epoch": 1.1, "learning_rate": 3.1723372733213e-05, "loss": 0.0, "step": 56520 }, { "epoch": 1.1, "learning_rate": 3.172272577295577e-05, "loss": 0.0, "step": 56522 }, { "epoch": 1.1, "learning_rate": 3.172207881269853e-05, "loss": 0.0, "step": 56524 }, { "epoch": 1.1, "learning_rate": 3.172143185244131e-05, "loss": 0.0001, "step": 56526 }, { "epoch": 1.1, "learning_rate": 3.172078489218407e-05, "loss": 0.0, "step": 56528 }, { "epoch": 1.1, "learning_rate": 3.172013793192684e-05, "loss": 0.0, "step": 56530 }, { "epoch": 1.1, "learning_rate": 3.1719490971669616e-05, "loss": 0.0013, "step": 56532 }, { "epoch": 1.1, "learning_rate": 3.171884401141238e-05, "loss": 0.0, "step": 56534 }, { "epoch": 1.1, "learning_rate": 3.1718197051155155e-05, "loss": 0.0, "step": 56536 }, { "epoch": 1.1, "learning_rate": 3.171755009089792e-05, "loss": 0.0004, "step": 56538 }, { "epoch": 1.1, "learning_rate": 3.1716903130640686e-05, "loss": 0.0, "step": 56540 }, { "epoch": 1.1, "learning_rate": 3.1716256170383455e-05, "loss": 0.0004, "step": 56542 }, { "epoch": 1.1, "learning_rate": 3.1715609210126224e-05, "loss": 0.0, "step": 56544 }, { "epoch": 1.1, "learning_rate": 3.171496224986899e-05, "loss": 0.0006, "step": 56546 }, { "epoch": 1.1, "learning_rate": 3.171431528961176e-05, "loss": 0.0001, "step": 56548 }, { "epoch": 1.1, "learning_rate": 3.171366832935453e-05, "loss": 0.0, "step": 56550 }, { "epoch": 1.1, "learning_rate": 3.1713021369097294e-05, "loss": 0.0003, "step": 56552 }, { "epoch": 1.1, "learning_rate": 3.171237440884007e-05, "loss": 0.0002, "step": 56554 }, { "epoch": 1.1, "learning_rate": 3.171172744858283e-05, "loss": 0.0, "step": 56556 }, { "epoch": 1.1, "learning_rate": 3.171108048832561e-05, "loss": 0.0, "step": 56558 }, { "epoch": 1.1, "learning_rate": 3.171043352806837e-05, "loss": 0.0001, "step": 56560 }, { "epoch": 1.1, "learning_rate": 3.170978656781114e-05, "loss": 0.0045, "step": 56562 }, { "epoch": 1.1, "learning_rate": 3.170913960755391e-05, "loss": 0.0074, "step": 56564 }, { "epoch": 1.1, "learning_rate": 3.170849264729668e-05, "loss": 0.0076, "step": 56566 }, { "epoch": 1.1, "learning_rate": 3.1707845687039446e-05, "loss": 0.0, "step": 56568 }, { "epoch": 1.1, "learning_rate": 3.1707198726782216e-05, "loss": 0.0, "step": 56570 }, { "epoch": 1.1, "learning_rate": 3.1706551766524985e-05, "loss": 0.0, "step": 56572 }, { "epoch": 1.1, "learning_rate": 3.170590480626775e-05, "loss": 0.0, "step": 56574 }, { "epoch": 1.1, "learning_rate": 3.170525784601052e-05, "loss": 0.0, "step": 56576 }, { "epoch": 1.1, "learning_rate": 3.170461088575329e-05, "loss": 0.0, "step": 56578 }, { "epoch": 1.1, "learning_rate": 3.170396392549606e-05, "loss": 0.0, "step": 56580 }, { "epoch": 1.1, "learning_rate": 3.170331696523883e-05, "loss": 0.0, "step": 56582 }, { "epoch": 1.1, "learning_rate": 3.170267000498159e-05, "loss": 0.0002, "step": 56584 }, { "epoch": 1.1, "learning_rate": 3.170202304472437e-05, "loss": 0.0, "step": 56586 }, { "epoch": 1.1, "learning_rate": 3.170137608446713e-05, "loss": 0.0, "step": 56588 }, { "epoch": 1.1, "learning_rate": 3.17007291242099e-05, "loss": 0.0, "step": 56590 }, { "epoch": 1.1, "learning_rate": 3.170008216395267e-05, "loss": 0.0, "step": 56592 }, { "epoch": 1.1, "learning_rate": 3.169943520369544e-05, "loss": 0.0006, "step": 56594 }, { "epoch": 1.1, "learning_rate": 3.169878824343821e-05, "loss": 0.0, "step": 56596 }, { "epoch": 1.1, "learning_rate": 3.1698141283180976e-05, "loss": 0.0001, "step": 56598 }, { "epoch": 1.1, "learning_rate": 3.1697494322923745e-05, "loss": 0.0001, "step": 56600 }, { "epoch": 1.1, "learning_rate": 3.1696847362666514e-05, "loss": 0.0, "step": 56602 }, { "epoch": 1.1, "learning_rate": 3.169620040240928e-05, "loss": 0.0019, "step": 56604 }, { "epoch": 1.1, "learning_rate": 3.1695553442152046e-05, "loss": 0.0, "step": 56606 }, { "epoch": 1.1, "learning_rate": 3.169490648189482e-05, "loss": 0.0051, "step": 56608 }, { "epoch": 1.1, "learning_rate": 3.1694259521637584e-05, "loss": 0.0, "step": 56610 }, { "epoch": 1.1, "learning_rate": 3.169361256138035e-05, "loss": 0.0, "step": 56612 }, { "epoch": 1.1, "learning_rate": 3.169296560112313e-05, "loss": 0.0, "step": 56614 }, { "epoch": 1.1, "learning_rate": 3.169231864086589e-05, "loss": 0.0, "step": 56616 }, { "epoch": 1.1, "learning_rate": 3.169167168060867e-05, "loss": 0.0065, "step": 56618 }, { "epoch": 1.1, "learning_rate": 3.169102472035143e-05, "loss": 0.0, "step": 56620 }, { "epoch": 1.1, "learning_rate": 3.16903777600942e-05, "loss": 0.0, "step": 56622 }, { "epoch": 1.1, "learning_rate": 3.168973079983697e-05, "loss": 0.0002, "step": 56624 }, { "epoch": 1.1, "learning_rate": 3.1689083839579736e-05, "loss": 0.0031, "step": 56626 }, { "epoch": 1.1, "learning_rate": 3.1688436879322505e-05, "loss": 0.0004, "step": 56628 }, { "epoch": 1.1, "learning_rate": 3.1687789919065275e-05, "loss": 0.0, "step": 56630 }, { "epoch": 1.1, "learning_rate": 3.1687142958808044e-05, "loss": 0.0, "step": 56632 }, { "epoch": 1.1, "learning_rate": 3.1686495998550806e-05, "loss": 0.0, "step": 56634 }, { "epoch": 1.1, "learning_rate": 3.168584903829358e-05, "loss": 0.0, "step": 56636 }, { "epoch": 1.1, "learning_rate": 3.1685202078036344e-05, "loss": 0.0, "step": 56638 }, { "epoch": 1.1, "learning_rate": 3.168455511777912e-05, "loss": 0.0006, "step": 56640 }, { "epoch": 1.1, "learning_rate": 3.168390815752188e-05, "loss": 0.0025, "step": 56642 }, { "epoch": 1.1, "learning_rate": 3.168326119726465e-05, "loss": 0.0001, "step": 56644 }, { "epoch": 1.1, "learning_rate": 3.168261423700743e-05, "loss": 0.0, "step": 56646 }, { "epoch": 1.1, "learning_rate": 3.168196727675019e-05, "loss": 0.0001, "step": 56648 }, { "epoch": 1.1, "learning_rate": 3.168132031649296e-05, "loss": 0.0, "step": 56650 }, { "epoch": 1.1, "learning_rate": 3.168067335623573e-05, "loss": 0.0004, "step": 56652 }, { "epoch": 1.1, "learning_rate": 3.16800263959785e-05, "loss": 0.0002, "step": 56654 }, { "epoch": 1.1, "learning_rate": 3.1679379435721266e-05, "loss": 0.0, "step": 56656 }, { "epoch": 1.1, "learning_rate": 3.1678732475464035e-05, "loss": 0.0019, "step": 56658 }, { "epoch": 1.1, "learning_rate": 3.1678085515206804e-05, "loss": 0.0, "step": 56660 }, { "epoch": 1.1, "learning_rate": 3.167743855494957e-05, "loss": 0.0, "step": 56662 }, { "epoch": 1.1, "learning_rate": 3.167679159469234e-05, "loss": 0.0048, "step": 56664 }, { "epoch": 1.1, "learning_rate": 3.1676144634435104e-05, "loss": 0.0002, "step": 56666 }, { "epoch": 1.1, "learning_rate": 3.167549767417788e-05, "loss": 0.0, "step": 56668 }, { "epoch": 1.1, "learning_rate": 3.167485071392064e-05, "loss": 0.0, "step": 56670 }, { "epoch": 1.1, "learning_rate": 3.167420375366341e-05, "loss": 0.0007, "step": 56672 }, { "epoch": 1.1, "learning_rate": 3.167355679340618e-05, "loss": 0.0, "step": 56674 }, { "epoch": 1.1, "learning_rate": 3.167290983314895e-05, "loss": 0.0, "step": 56676 }, { "epoch": 1.1, "learning_rate": 3.167226287289172e-05, "loss": 0.0, "step": 56678 }, { "epoch": 1.1, "learning_rate": 3.167161591263449e-05, "loss": 0.0059, "step": 56680 }, { "epoch": 1.1, "learning_rate": 3.167096895237726e-05, "loss": 0.0, "step": 56682 }, { "epoch": 1.1, "learning_rate": 3.1670321992120026e-05, "loss": 0.0, "step": 56684 }, { "epoch": 1.1, "learning_rate": 3.1669675031862795e-05, "loss": 0.0, "step": 56686 }, { "epoch": 1.1, "learning_rate": 3.166902807160556e-05, "loss": 0.0, "step": 56688 }, { "epoch": 1.1, "learning_rate": 3.1668381111348333e-05, "loss": 0.0069, "step": 56690 }, { "epoch": 1.1, "learning_rate": 3.16677341510911e-05, "loss": 0.0, "step": 56692 }, { "epoch": 1.1, "learning_rate": 3.1667087190833865e-05, "loss": 0.0, "step": 56694 }, { "epoch": 1.1, "learning_rate": 3.166644023057664e-05, "loss": 0.0052, "step": 56696 }, { "epoch": 1.1, "learning_rate": 3.16657932703194e-05, "loss": 0.0002, "step": 56698 }, { "epoch": 1.1, "learning_rate": 3.166514631006218e-05, "loss": 0.0001, "step": 56700 }, { "epoch": 1.1, "learning_rate": 3.166449934980494e-05, "loss": 0.0018, "step": 56702 }, { "epoch": 1.1, "learning_rate": 3.166385238954771e-05, "loss": 0.0, "step": 56704 }, { "epoch": 1.1, "learning_rate": 3.166320542929048e-05, "loss": 0.0, "step": 56706 }, { "epoch": 1.1, "learning_rate": 3.166255846903325e-05, "loss": 0.0, "step": 56708 }, { "epoch": 1.1, "learning_rate": 3.166191150877602e-05, "loss": 0.0002, "step": 56710 }, { "epoch": 1.1, "learning_rate": 3.166126454851879e-05, "loss": 0.0, "step": 56712 }, { "epoch": 1.1, "learning_rate": 3.1660617588261556e-05, "loss": 0.0, "step": 56714 }, { "epoch": 1.1, "learning_rate": 3.165997062800432e-05, "loss": 0.0035, "step": 56716 }, { "epoch": 1.1, "learning_rate": 3.1659323667747094e-05, "loss": 0.0, "step": 56718 }, { "epoch": 1.1, "learning_rate": 3.1658676707489856e-05, "loss": 0.0, "step": 56720 }, { "epoch": 1.1, "learning_rate": 3.165802974723263e-05, "loss": 0.0, "step": 56722 }, { "epoch": 1.1, "learning_rate": 3.16573827869754e-05, "loss": 0.0, "step": 56724 }, { "epoch": 1.1, "learning_rate": 3.1656735826718163e-05, "loss": 0.0001, "step": 56726 }, { "epoch": 1.1, "learning_rate": 3.165608886646094e-05, "loss": 0.0036, "step": 56728 }, { "epoch": 1.1, "learning_rate": 3.16554419062037e-05, "loss": 0.0001, "step": 56730 }, { "epoch": 1.1, "learning_rate": 3.165479494594647e-05, "loss": 0.0007, "step": 56732 }, { "epoch": 1.1, "learning_rate": 3.165414798568924e-05, "loss": 0.0, "step": 56734 }, { "epoch": 1.1, "learning_rate": 3.165350102543201e-05, "loss": 0.0, "step": 56736 }, { "epoch": 1.1, "learning_rate": 3.165285406517478e-05, "loss": 0.0003, "step": 56738 }, { "epoch": 1.1, "learning_rate": 3.165220710491755e-05, "loss": 0.0, "step": 56740 }, { "epoch": 1.1, "learning_rate": 3.1651560144660316e-05, "loss": 0.0, "step": 56742 }, { "epoch": 1.1, "learning_rate": 3.1650913184403085e-05, "loss": 0.0, "step": 56744 }, { "epoch": 1.1, "learning_rate": 3.1650266224145854e-05, "loss": 0.0, "step": 56746 }, { "epoch": 1.1, "learning_rate": 3.164961926388862e-05, "loss": 0.0, "step": 56748 }, { "epoch": 1.1, "learning_rate": 3.164897230363139e-05, "loss": 0.0002, "step": 56750 }, { "epoch": 1.1, "learning_rate": 3.1648325343374155e-05, "loss": 0.0049, "step": 56752 }, { "epoch": 1.1, "learning_rate": 3.1647678383116924e-05, "loss": 0.0, "step": 56754 }, { "epoch": 1.1, "learning_rate": 3.16470314228597e-05, "loss": 0.0, "step": 56756 }, { "epoch": 1.1, "learning_rate": 3.164638446260246e-05, "loss": 0.0, "step": 56758 }, { "epoch": 1.1, "learning_rate": 3.164573750234524e-05, "loss": 0.0, "step": 56760 }, { "epoch": 1.1, "learning_rate": 3.1645090542088e-05, "loss": 0.0, "step": 56762 }, { "epoch": 1.1, "learning_rate": 3.164444358183077e-05, "loss": 0.0, "step": 56764 }, { "epoch": 1.1, "learning_rate": 3.164379662157354e-05, "loss": 0.0123, "step": 56766 }, { "epoch": 1.1, "learning_rate": 3.164314966131631e-05, "loss": 0.0078, "step": 56768 }, { "epoch": 1.1, "learning_rate": 3.1642502701059077e-05, "loss": 0.0004, "step": 56770 }, { "epoch": 1.1, "learning_rate": 3.1641855740801846e-05, "loss": 0.0, "step": 56772 }, { "epoch": 1.1, "learning_rate": 3.1641208780544615e-05, "loss": 0.0001, "step": 56774 }, { "epoch": 1.1, "learning_rate": 3.164056182028738e-05, "loss": 0.0001, "step": 56776 }, { "epoch": 1.1, "learning_rate": 3.163991486003015e-05, "loss": 0.0, "step": 56778 }, { "epoch": 1.1, "learning_rate": 3.1639267899772915e-05, "loss": 0.0, "step": 56780 }, { "epoch": 1.1, "learning_rate": 3.163862093951569e-05, "loss": 0.0, "step": 56782 }, { "epoch": 1.1, "learning_rate": 3.163797397925845e-05, "loss": 0.0, "step": 56784 }, { "epoch": 1.1, "learning_rate": 3.163732701900122e-05, "loss": 0.0, "step": 56786 }, { "epoch": 1.1, "learning_rate": 3.163668005874399e-05, "loss": 0.0, "step": 56788 }, { "epoch": 1.1, "learning_rate": 3.163603309848676e-05, "loss": 0.0002, "step": 56790 }, { "epoch": 1.1, "learning_rate": 3.163538613822953e-05, "loss": 0.0, "step": 56792 }, { "epoch": 1.1, "learning_rate": 3.16347391779723e-05, "loss": 0.0, "step": 56794 }, { "epoch": 1.1, "learning_rate": 3.163409221771507e-05, "loss": 0.0003, "step": 56796 }, { "epoch": 1.1, "learning_rate": 3.163344525745783e-05, "loss": 0.0006, "step": 56798 }, { "epoch": 1.1, "learning_rate": 3.1632798297200606e-05, "loss": 0.0, "step": 56800 }, { "epoch": 1.1, "learning_rate": 3.1632151336943375e-05, "loss": 0.0001, "step": 56802 }, { "epoch": 1.1, "learning_rate": 3.1631504376686144e-05, "loss": 0.0, "step": 56804 }, { "epoch": 1.1, "learning_rate": 3.163085741642891e-05, "loss": 0.0798, "step": 56806 }, { "epoch": 1.1, "learning_rate": 3.1630210456171676e-05, "loss": 0.0, "step": 56808 }, { "epoch": 1.1, "learning_rate": 3.162956349591445e-05, "loss": 0.0027, "step": 56810 }, { "epoch": 1.1, "learning_rate": 3.1628916535657214e-05, "loss": 0.0179, "step": 56812 }, { "epoch": 1.1, "learning_rate": 3.162826957539998e-05, "loss": 0.0, "step": 56814 }, { "epoch": 1.1, "learning_rate": 3.162762261514275e-05, "loss": 0.0, "step": 56816 }, { "epoch": 1.1, "learning_rate": 3.162697565488552e-05, "loss": 0.0, "step": 56818 }, { "epoch": 1.1, "learning_rate": 3.162632869462829e-05, "loss": 0.0018, "step": 56820 }, { "epoch": 1.1, "learning_rate": 3.162568173437106e-05, "loss": 0.0, "step": 56822 }, { "epoch": 1.1, "learning_rate": 3.162503477411383e-05, "loss": 0.0045, "step": 56824 }, { "epoch": 1.1, "learning_rate": 3.16243878138566e-05, "loss": 0.0088, "step": 56826 }, { "epoch": 1.1, "learning_rate": 3.1623740853599366e-05, "loss": 0.0002, "step": 56828 }, { "epoch": 1.1, "learning_rate": 3.162309389334213e-05, "loss": 0.0316, "step": 56830 }, { "epoch": 1.1, "learning_rate": 3.1622446933084905e-05, "loss": 0.0, "step": 56832 }, { "epoch": 1.1, "learning_rate": 3.1621799972827674e-05, "loss": 0.0, "step": 56834 }, { "epoch": 1.1, "learning_rate": 3.1621153012570436e-05, "loss": 0.0001, "step": 56836 }, { "epoch": 1.1, "learning_rate": 3.162050605231321e-05, "loss": 0.0001, "step": 56838 }, { "epoch": 1.1, "learning_rate": 3.1619859092055974e-05, "loss": 0.0007, "step": 56840 }, { "epoch": 1.1, "learning_rate": 3.161921213179875e-05, "loss": 0.0, "step": 56842 }, { "epoch": 1.1, "learning_rate": 3.161856517154151e-05, "loss": 0.0, "step": 56844 }, { "epoch": 1.1, "learning_rate": 3.161791821128428e-05, "loss": 0.0, "step": 56846 }, { "epoch": 1.1, "learning_rate": 3.161727125102705e-05, "loss": 0.0, "step": 56848 }, { "epoch": 1.1, "learning_rate": 3.161662429076982e-05, "loss": 0.0, "step": 56850 }, { "epoch": 1.1, "learning_rate": 3.161597733051259e-05, "loss": 0.0, "step": 56852 }, { "epoch": 1.1, "learning_rate": 3.161533037025536e-05, "loss": 0.0, "step": 56854 }, { "epoch": 1.1, "learning_rate": 3.161468340999813e-05, "loss": 0.0, "step": 56856 }, { "epoch": 1.1, "learning_rate": 3.161403644974089e-05, "loss": 0.0, "step": 56858 }, { "epoch": 1.1, "learning_rate": 3.1613389489483665e-05, "loss": 0.0095, "step": 56860 }, { "epoch": 1.1, "learning_rate": 3.161274252922643e-05, "loss": 0.0004, "step": 56862 }, { "epoch": 1.1, "learning_rate": 3.16120955689692e-05, "loss": 0.0, "step": 56864 }, { "epoch": 1.1, "learning_rate": 3.1611448608711965e-05, "loss": 0.0, "step": 56866 }, { "epoch": 1.1, "learning_rate": 3.1610801648454735e-05, "loss": 0.0, "step": 56868 }, { "epoch": 1.1, "learning_rate": 3.161015468819751e-05, "loss": 0.0, "step": 56870 }, { "epoch": 1.1, "learning_rate": 3.160950772794027e-05, "loss": 0.0, "step": 56872 }, { "epoch": 1.1, "learning_rate": 3.160886076768304e-05, "loss": 0.0, "step": 56874 }, { "epoch": 1.1, "learning_rate": 3.160821380742581e-05, "loss": 0.0, "step": 56876 }, { "epoch": 1.1, "learning_rate": 3.160756684716858e-05, "loss": 0.0, "step": 56878 }, { "epoch": 1.1, "learning_rate": 3.160691988691135e-05, "loss": 0.0, "step": 56880 }, { "epoch": 1.1, "learning_rate": 3.160627292665412e-05, "loss": 0.0, "step": 56882 }, { "epoch": 1.1, "learning_rate": 3.160562596639689e-05, "loss": 0.0, "step": 56884 }, { "epoch": 1.1, "learning_rate": 3.1604979006139656e-05, "loss": 0.0, "step": 56886 }, { "epoch": 1.1, "learning_rate": 3.1604332045882425e-05, "loss": 0.0001, "step": 56888 }, { "epoch": 1.1, "learning_rate": 3.160368508562519e-05, "loss": 0.0, "step": 56890 }, { "epoch": 1.1, "learning_rate": 3.1603038125367964e-05, "loss": 0.0, "step": 56892 }, { "epoch": 1.1, "learning_rate": 3.1602391165110726e-05, "loss": 0.0, "step": 56894 }, { "epoch": 1.1, "learning_rate": 3.1601744204853495e-05, "loss": 0.0, "step": 56896 }, { "epoch": 1.1, "learning_rate": 3.1601097244596264e-05, "loss": 0.0012, "step": 56898 }, { "epoch": 1.1, "learning_rate": 3.160045028433903e-05, "loss": 0.0, "step": 56900 }, { "epoch": 1.1, "learning_rate": 3.159980332408181e-05, "loss": 0.0, "step": 56902 }, { "epoch": 1.1, "learning_rate": 3.159915636382457e-05, "loss": 0.0042, "step": 56904 }, { "epoch": 1.1, "learning_rate": 3.159850940356734e-05, "loss": 0.0, "step": 56906 }, { "epoch": 1.1, "learning_rate": 3.159786244331011e-05, "loss": 0.0, "step": 56908 }, { "epoch": 1.1, "learning_rate": 3.159721548305288e-05, "loss": 0.0, "step": 56910 }, { "epoch": 1.1, "learning_rate": 3.159656852279565e-05, "loss": 0.0001, "step": 56912 }, { "epoch": 1.1, "learning_rate": 3.159592156253842e-05, "loss": 0.003, "step": 56914 }, { "epoch": 1.1, "learning_rate": 3.1595274602281186e-05, "loss": 0.0, "step": 56916 }, { "epoch": 1.1, "learning_rate": 3.159462764202395e-05, "loss": 0.0, "step": 56918 }, { "epoch": 1.1, "learning_rate": 3.1593980681766724e-05, "loss": 0.0, "step": 56920 }, { "epoch": 1.1, "learning_rate": 3.1593333721509486e-05, "loss": 0.0, "step": 56922 }, { "epoch": 1.1, "learning_rate": 3.159268676125226e-05, "loss": 0.0, "step": 56924 }, { "epoch": 1.1, "learning_rate": 3.1592039800995024e-05, "loss": 0.0, "step": 56926 }, { "epoch": 1.1, "learning_rate": 3.1591392840737794e-05, "loss": 0.0004, "step": 56928 }, { "epoch": 1.1, "learning_rate": 3.159074588048056e-05, "loss": 0.0, "step": 56930 }, { "epoch": 1.1, "learning_rate": 3.159009892022333e-05, "loss": 0.0, "step": 56932 }, { "epoch": 1.11, "learning_rate": 3.15894519599661e-05, "loss": 0.0001, "step": 56934 }, { "epoch": 1.11, "learning_rate": 3.158880499970887e-05, "loss": 0.0, "step": 56936 }, { "epoch": 1.11, "learning_rate": 3.158815803945164e-05, "loss": 0.0058, "step": 56938 }, { "epoch": 1.11, "learning_rate": 3.15875110791944e-05, "loss": 0.0032, "step": 56940 }, { "epoch": 1.11, "learning_rate": 3.158686411893718e-05, "loss": 0.0004, "step": 56942 }, { "epoch": 1.11, "learning_rate": 3.158621715867994e-05, "loss": 0.0, "step": 56944 }, { "epoch": 1.11, "learning_rate": 3.1585570198422715e-05, "loss": 0.0003, "step": 56946 }, { "epoch": 1.11, "learning_rate": 3.1584923238165484e-05, "loss": 0.0005, "step": 56948 }, { "epoch": 1.11, "learning_rate": 3.158427627790825e-05, "loss": 0.0001, "step": 56950 }, { "epoch": 1.11, "learning_rate": 3.158362931765102e-05, "loss": 0.0, "step": 56952 }, { "epoch": 1.11, "learning_rate": 3.1582982357393785e-05, "loss": 0.0, "step": 56954 }, { "epoch": 1.11, "learning_rate": 3.1582335397136554e-05, "loss": 0.0, "step": 56956 }, { "epoch": 1.11, "learning_rate": 3.158168843687932e-05, "loss": 0.0001, "step": 56958 }, { "epoch": 1.11, "learning_rate": 3.158104147662209e-05, "loss": 0.0002, "step": 56960 }, { "epoch": 1.11, "learning_rate": 3.158039451636486e-05, "loss": 0.0, "step": 56962 }, { "epoch": 1.11, "learning_rate": 3.157974755610763e-05, "loss": 0.0014, "step": 56964 }, { "epoch": 1.11, "learning_rate": 3.15791005958504e-05, "loss": 0.0, "step": 56966 }, { "epoch": 1.11, "learning_rate": 3.157845363559317e-05, "loss": 0.0, "step": 56968 }, { "epoch": 1.11, "learning_rate": 3.157780667533594e-05, "loss": 0.0, "step": 56970 }, { "epoch": 1.11, "learning_rate": 3.15771597150787e-05, "loss": 0.0, "step": 56972 }, { "epoch": 1.11, "learning_rate": 3.1576512754821476e-05, "loss": 0.0, "step": 56974 }, { "epoch": 1.11, "learning_rate": 3.157586579456424e-05, "loss": 0.0001, "step": 56976 }, { "epoch": 1.11, "learning_rate": 3.157521883430701e-05, "loss": 0.0035, "step": 56978 }, { "epoch": 1.11, "learning_rate": 3.157457187404978e-05, "loss": 0.0, "step": 56980 }, { "epoch": 1.11, "learning_rate": 3.1573924913792545e-05, "loss": 0.0, "step": 56982 }, { "epoch": 1.11, "learning_rate": 3.157327795353532e-05, "loss": 0.0049, "step": 56984 }, { "epoch": 1.11, "learning_rate": 3.1572630993278083e-05, "loss": 0.004, "step": 56986 }, { "epoch": 1.11, "learning_rate": 3.157198403302085e-05, "loss": 0.0, "step": 56988 }, { "epoch": 1.11, "learning_rate": 3.157133707276362e-05, "loss": 0.0, "step": 56990 }, { "epoch": 1.11, "learning_rate": 3.157069011250639e-05, "loss": 0.0, "step": 56992 }, { "epoch": 1.11, "learning_rate": 3.157004315224916e-05, "loss": 0.0, "step": 56994 }, { "epoch": 1.11, "learning_rate": 3.156939619199193e-05, "loss": 0.0, "step": 56996 }, { "epoch": 1.11, "learning_rate": 3.15687492317347e-05, "loss": 0.0015, "step": 56998 }, { "epoch": 1.11, "learning_rate": 3.156810227147746e-05, "loss": 0.0, "step": 57000 }, { "epoch": 1.11, "learning_rate": 3.1567455311220236e-05, "loss": 0.0, "step": 57002 }, { "epoch": 1.11, "learning_rate": 3.1566808350963e-05, "loss": 0.0, "step": 57004 }, { "epoch": 1.11, "learning_rate": 3.1566161390705774e-05, "loss": 0.0, "step": 57006 }, { "epoch": 1.11, "learning_rate": 3.1565514430448537e-05, "loss": 0.0, "step": 57008 }, { "epoch": 1.11, "learning_rate": 3.1564867470191306e-05, "loss": 0.0, "step": 57010 }, { "epoch": 1.11, "learning_rate": 3.156422050993408e-05, "loss": 0.0, "step": 57012 }, { "epoch": 1.11, "learning_rate": 3.1563573549676844e-05, "loss": 0.0, "step": 57014 }, { "epoch": 1.11, "learning_rate": 3.156292658941961e-05, "loss": 0.0, "step": 57016 }, { "epoch": 1.11, "learning_rate": 3.156227962916238e-05, "loss": 0.0001, "step": 57018 }, { "epoch": 1.11, "learning_rate": 3.156163266890515e-05, "loss": 0.0, "step": 57020 }, { "epoch": 1.11, "learning_rate": 3.1560985708647913e-05, "loss": 0.0, "step": 57022 }, { "epoch": 1.11, "learning_rate": 3.156033874839069e-05, "loss": 0.0, "step": 57024 }, { "epoch": 1.11, "learning_rate": 3.155969178813346e-05, "loss": 0.0001, "step": 57026 }, { "epoch": 1.11, "learning_rate": 3.155904482787623e-05, "loss": 0.0, "step": 57028 }, { "epoch": 1.11, "learning_rate": 3.1558397867618997e-05, "loss": 0.0, "step": 57030 }, { "epoch": 1.11, "learning_rate": 3.155775090736176e-05, "loss": 0.0, "step": 57032 }, { "epoch": 1.11, "learning_rate": 3.1557103947104535e-05, "loss": 0.0, "step": 57034 }, { "epoch": 1.11, "learning_rate": 3.15564569868473e-05, "loss": 0.0004, "step": 57036 }, { "epoch": 1.11, "learning_rate": 3.1555810026590066e-05, "loss": 0.0, "step": 57038 }, { "epoch": 1.11, "learning_rate": 3.1555163066332835e-05, "loss": 0.0, "step": 57040 }, { "epoch": 1.11, "learning_rate": 3.1554516106075604e-05, "loss": 0.0924, "step": 57042 }, { "epoch": 1.11, "learning_rate": 3.155386914581837e-05, "loss": 0.0074, "step": 57044 }, { "epoch": 1.11, "learning_rate": 3.155322218556114e-05, "loss": 0.0, "step": 57046 }, { "epoch": 1.11, "learning_rate": 3.155257522530391e-05, "loss": 0.0001, "step": 57048 }, { "epoch": 1.11, "learning_rate": 3.155192826504668e-05, "loss": 0.0, "step": 57050 }, { "epoch": 1.11, "learning_rate": 3.155128130478945e-05, "loss": 0.0, "step": 57052 }, { "epoch": 1.11, "learning_rate": 3.155063434453221e-05, "loss": 0.001, "step": 57054 }, { "epoch": 1.11, "learning_rate": 3.154998738427499e-05, "loss": 0.0001, "step": 57056 }, { "epoch": 1.11, "learning_rate": 3.154934042401776e-05, "loss": 0.0, "step": 57058 }, { "epoch": 1.11, "learning_rate": 3.154869346376052e-05, "loss": 0.0, "step": 57060 }, { "epoch": 1.11, "learning_rate": 3.1548046503503295e-05, "loss": 0.0, "step": 57062 }, { "epoch": 1.11, "learning_rate": 3.154739954324606e-05, "loss": 0.0, "step": 57064 }, { "epoch": 1.11, "learning_rate": 3.154675258298883e-05, "loss": 0.002, "step": 57066 }, { "epoch": 1.11, "learning_rate": 3.1546105622731596e-05, "loss": 0.0014, "step": 57068 }, { "epoch": 1.11, "learning_rate": 3.1545458662474365e-05, "loss": 0.0055, "step": 57070 }, { "epoch": 1.11, "learning_rate": 3.1544811702217134e-05, "loss": 0.0002, "step": 57072 }, { "epoch": 1.11, "learning_rate": 3.15441647419599e-05, "loss": 0.0032, "step": 57074 }, { "epoch": 1.11, "learning_rate": 3.154351778170267e-05, "loss": 0.0, "step": 57076 }, { "epoch": 1.11, "learning_rate": 3.154287082144544e-05, "loss": 0.0004, "step": 57078 }, { "epoch": 1.11, "learning_rate": 3.154222386118821e-05, "loss": 0.0, "step": 57080 }, { "epoch": 1.11, "learning_rate": 3.154157690093097e-05, "loss": 0.0081, "step": 57082 }, { "epoch": 1.11, "learning_rate": 3.154092994067375e-05, "loss": 0.0, "step": 57084 }, { "epoch": 1.11, "learning_rate": 3.154028298041651e-05, "loss": 0.0026, "step": 57086 }, { "epoch": 1.11, "learning_rate": 3.1539636020159286e-05, "loss": 0.0112, "step": 57088 }, { "epoch": 1.11, "learning_rate": 3.153898905990205e-05, "loss": 0.0191, "step": 57090 }, { "epoch": 1.11, "learning_rate": 3.153834209964482e-05, "loss": 0.0, "step": 57092 }, { "epoch": 1.11, "learning_rate": 3.1537695139387594e-05, "loss": 0.0, "step": 57094 }, { "epoch": 1.11, "learning_rate": 3.1537048179130356e-05, "loss": 0.0, "step": 57096 }, { "epoch": 1.11, "learning_rate": 3.1536401218873125e-05, "loss": 0.0087, "step": 57098 }, { "epoch": 1.11, "learning_rate": 3.1535754258615894e-05, "loss": 0.0, "step": 57100 }, { "epoch": 1.11, "learning_rate": 3.153510729835866e-05, "loss": 0.0, "step": 57102 }, { "epoch": 1.11, "learning_rate": 3.153446033810143e-05, "loss": 0.0, "step": 57104 }, { "epoch": 1.11, "learning_rate": 3.15338133778442e-05, "loss": 0.0003, "step": 57106 }, { "epoch": 1.11, "learning_rate": 3.153316641758697e-05, "loss": 0.0, "step": 57108 }, { "epoch": 1.11, "learning_rate": 3.153251945732974e-05, "loss": 0.0005, "step": 57110 }, { "epoch": 1.11, "learning_rate": 3.153187249707251e-05, "loss": 0.0, "step": 57112 }, { "epoch": 1.11, "learning_rate": 3.153122553681527e-05, "loss": 0.0001, "step": 57114 }, { "epoch": 1.11, "learning_rate": 3.153057857655805e-05, "loss": 0.0, "step": 57116 }, { "epoch": 1.11, "learning_rate": 3.152993161630081e-05, "loss": 0.0002, "step": 57118 }, { "epoch": 1.11, "learning_rate": 3.152928465604358e-05, "loss": 0.0002, "step": 57120 }, { "epoch": 1.11, "learning_rate": 3.152863769578635e-05, "loss": 0.0, "step": 57122 }, { "epoch": 1.11, "learning_rate": 3.1527990735529116e-05, "loss": 0.0001, "step": 57124 }, { "epoch": 1.11, "learning_rate": 3.152734377527189e-05, "loss": 0.0147, "step": 57126 }, { "epoch": 1.11, "learning_rate": 3.1526696815014655e-05, "loss": 0.0012, "step": 57128 }, { "epoch": 1.11, "learning_rate": 3.1526049854757424e-05, "loss": 0.0, "step": 57130 }, { "epoch": 1.11, "learning_rate": 3.152540289450019e-05, "loss": 0.0003, "step": 57132 }, { "epoch": 1.11, "learning_rate": 3.152475593424296e-05, "loss": 0.0052, "step": 57134 }, { "epoch": 1.11, "learning_rate": 3.152410897398573e-05, "loss": 0.0, "step": 57136 }, { "epoch": 1.11, "learning_rate": 3.15234620137285e-05, "loss": 0.0, "step": 57138 }, { "epoch": 1.11, "learning_rate": 3.152281505347127e-05, "loss": 0.0, "step": 57140 }, { "epoch": 1.11, "learning_rate": 3.152216809321403e-05, "loss": 0.0001, "step": 57142 }, { "epoch": 1.11, "learning_rate": 3.152152113295681e-05, "loss": 0.0, "step": 57144 }, { "epoch": 1.11, "learning_rate": 3.152087417269957e-05, "loss": 0.003, "step": 57146 }, { "epoch": 1.11, "learning_rate": 3.1520227212442345e-05, "loss": 0.0009, "step": 57148 }, { "epoch": 1.11, "learning_rate": 3.151958025218511e-05, "loss": 0.0001, "step": 57150 }, { "epoch": 1.11, "learning_rate": 3.151893329192788e-05, "loss": 0.0, "step": 57152 }, { "epoch": 1.11, "learning_rate": 3.1518286331670646e-05, "loss": 0.0, "step": 57154 }, { "epoch": 1.11, "learning_rate": 3.1517639371413415e-05, "loss": 0.0, "step": 57156 }, { "epoch": 1.11, "learning_rate": 3.1516992411156184e-05, "loss": 0.0003, "step": 57158 }, { "epoch": 1.11, "learning_rate": 3.151634545089895e-05, "loss": 0.0, "step": 57160 }, { "epoch": 1.11, "learning_rate": 3.151569849064172e-05, "loss": 0.0, "step": 57162 }, { "epoch": 1.11, "learning_rate": 3.1515051530384485e-05, "loss": 0.0003, "step": 57164 }, { "epoch": 1.11, "learning_rate": 3.151440457012726e-05, "loss": 0.0, "step": 57166 }, { "epoch": 1.11, "learning_rate": 3.151375760987002e-05, "loss": 0.0, "step": 57168 }, { "epoch": 1.11, "learning_rate": 3.15131106496128e-05, "loss": 0.0, "step": 57170 }, { "epoch": 1.11, "learning_rate": 3.151246368935557e-05, "loss": 0.0, "step": 57172 }, { "epoch": 1.11, "learning_rate": 3.151181672909833e-05, "loss": 0.0, "step": 57174 }, { "epoch": 1.11, "learning_rate": 3.1511169768841106e-05, "loss": 0.0, "step": 57176 }, { "epoch": 1.11, "learning_rate": 3.151052280858387e-05, "loss": 0.0001, "step": 57178 }, { "epoch": 1.11, "learning_rate": 3.150987584832664e-05, "loss": 0.0, "step": 57180 }, { "epoch": 1.11, "learning_rate": 3.1509228888069406e-05, "loss": 0.0, "step": 57182 }, { "epoch": 1.11, "learning_rate": 3.1508581927812175e-05, "loss": 0.0, "step": 57184 }, { "epoch": 1.11, "learning_rate": 3.1507934967554944e-05, "loss": 0.0, "step": 57186 }, { "epoch": 1.11, "learning_rate": 3.1507288007297714e-05, "loss": 0.0, "step": 57188 }, { "epoch": 1.11, "learning_rate": 3.150664104704048e-05, "loss": 0.0, "step": 57190 }, { "epoch": 1.11, "learning_rate": 3.150599408678325e-05, "loss": 0.0, "step": 57192 }, { "epoch": 1.11, "learning_rate": 3.150534712652602e-05, "loss": 0.0, "step": 57194 }, { "epoch": 1.11, "learning_rate": 3.150470016626878e-05, "loss": 0.0, "step": 57196 }, { "epoch": 1.11, "learning_rate": 3.150405320601156e-05, "loss": 0.0002, "step": 57198 }, { "epoch": 1.11, "learning_rate": 3.150340624575432e-05, "loss": 0.0, "step": 57200 }, { "epoch": 1.11, "learning_rate": 3.150275928549709e-05, "loss": 0.0, "step": 57202 }, { "epoch": 1.11, "learning_rate": 3.1502112325239866e-05, "loss": 0.0, "step": 57204 }, { "epoch": 1.11, "learning_rate": 3.150146536498263e-05, "loss": 0.0001, "step": 57206 }, { "epoch": 1.11, "learning_rate": 3.1500818404725404e-05, "loss": 0.0, "step": 57208 }, { "epoch": 1.11, "learning_rate": 3.150017144446817e-05, "loss": 0.0, "step": 57210 }, { "epoch": 1.11, "learning_rate": 3.1499524484210936e-05, "loss": 0.0001, "step": 57212 }, { "epoch": 1.11, "learning_rate": 3.1498877523953705e-05, "loss": 0.0, "step": 57214 }, { "epoch": 1.11, "learning_rate": 3.1498230563696474e-05, "loss": 0.0, "step": 57216 }, { "epoch": 1.11, "learning_rate": 3.149758360343924e-05, "loss": 0.0, "step": 57218 }, { "epoch": 1.11, "learning_rate": 3.149693664318201e-05, "loss": 0.0001, "step": 57220 }, { "epoch": 1.11, "learning_rate": 3.149628968292478e-05, "loss": 0.0, "step": 57222 }, { "epoch": 1.11, "learning_rate": 3.1495642722667544e-05, "loss": 0.0, "step": 57224 }, { "epoch": 1.11, "learning_rate": 3.149499576241032e-05, "loss": 0.0064, "step": 57226 }, { "epoch": 1.11, "learning_rate": 3.149434880215308e-05, "loss": 0.0, "step": 57228 }, { "epoch": 1.11, "learning_rate": 3.149370184189586e-05, "loss": 0.0001, "step": 57230 }, { "epoch": 1.11, "learning_rate": 3.149305488163862e-05, "loss": 0.0, "step": 57232 }, { "epoch": 1.11, "learning_rate": 3.149240792138139e-05, "loss": 0.0, "step": 57234 }, { "epoch": 1.11, "learning_rate": 3.1491760961124165e-05, "loss": 0.0, "step": 57236 }, { "epoch": 1.11, "learning_rate": 3.149111400086693e-05, "loss": 0.0, "step": 57238 }, { "epoch": 1.11, "learning_rate": 3.1490467040609696e-05, "loss": 0.0, "step": 57240 }, { "epoch": 1.11, "learning_rate": 3.1489820080352465e-05, "loss": 0.0, "step": 57242 }, { "epoch": 1.11, "learning_rate": 3.1489173120095234e-05, "loss": 0.0003, "step": 57244 }, { "epoch": 1.11, "learning_rate": 3.1488526159838e-05, "loss": 0.0, "step": 57246 }, { "epoch": 1.11, "learning_rate": 3.148787919958077e-05, "loss": 0.0107, "step": 57248 }, { "epoch": 1.11, "learning_rate": 3.148723223932354e-05, "loss": 0.0001, "step": 57250 }, { "epoch": 1.11, "learning_rate": 3.148658527906631e-05, "loss": 0.0, "step": 57252 }, { "epoch": 1.11, "learning_rate": 3.148593831880908e-05, "loss": 0.0, "step": 57254 }, { "epoch": 1.11, "learning_rate": 3.148529135855184e-05, "loss": 0.0032, "step": 57256 }, { "epoch": 1.11, "learning_rate": 3.148464439829462e-05, "loss": 0.0, "step": 57258 }, { "epoch": 1.11, "learning_rate": 3.148399743803738e-05, "loss": 0.0, "step": 57260 }, { "epoch": 1.11, "learning_rate": 3.148335047778015e-05, "loss": 0.0, "step": 57262 }, { "epoch": 1.11, "learning_rate": 3.148270351752292e-05, "loss": 0.0001, "step": 57264 }, { "epoch": 1.11, "learning_rate": 3.148205655726569e-05, "loss": 0.0, "step": 57266 }, { "epoch": 1.11, "learning_rate": 3.1481409597008457e-05, "loss": 0.0001, "step": 57268 }, { "epoch": 1.11, "learning_rate": 3.1480762636751226e-05, "loss": 0.0035, "step": 57270 }, { "epoch": 1.11, "learning_rate": 3.1480115676493995e-05, "loss": 0.0, "step": 57272 }, { "epoch": 1.11, "learning_rate": 3.1479468716236764e-05, "loss": 0.0, "step": 57274 }, { "epoch": 1.11, "learning_rate": 3.147882175597953e-05, "loss": 0.0, "step": 57276 }, { "epoch": 1.11, "learning_rate": 3.1478174795722295e-05, "loss": 0.0, "step": 57278 }, { "epoch": 1.11, "learning_rate": 3.147752783546507e-05, "loss": 0.0, "step": 57280 }, { "epoch": 1.11, "learning_rate": 3.147688087520784e-05, "loss": 0.0001, "step": 57282 }, { "epoch": 1.11, "learning_rate": 3.14762339149506e-05, "loss": 0.0043, "step": 57284 }, { "epoch": 1.11, "learning_rate": 3.147558695469338e-05, "loss": 0.0, "step": 57286 }, { "epoch": 1.11, "learning_rate": 3.147493999443614e-05, "loss": 0.0, "step": 57288 }, { "epoch": 1.11, "learning_rate": 3.1474293034178917e-05, "loss": 0.0, "step": 57290 }, { "epoch": 1.11, "learning_rate": 3.147364607392168e-05, "loss": 0.0, "step": 57292 }, { "epoch": 1.11, "learning_rate": 3.147299911366445e-05, "loss": 0.0, "step": 57294 }, { "epoch": 1.11, "learning_rate": 3.147235215340722e-05, "loss": 0.0, "step": 57296 }, { "epoch": 1.11, "learning_rate": 3.1471705193149986e-05, "loss": 0.0001, "step": 57298 }, { "epoch": 1.11, "learning_rate": 3.1471058232892755e-05, "loss": 0.0, "step": 57300 }, { "epoch": 1.11, "learning_rate": 3.1470411272635524e-05, "loss": 0.0312, "step": 57302 }, { "epoch": 1.11, "learning_rate": 3.146976431237829e-05, "loss": 0.0, "step": 57304 }, { "epoch": 1.11, "learning_rate": 3.1469117352121056e-05, "loss": 0.0001, "step": 57306 }, { "epoch": 1.11, "learning_rate": 3.146847039186383e-05, "loss": 0.0001, "step": 57308 }, { "epoch": 1.11, "learning_rate": 3.1467823431606594e-05, "loss": 0.0, "step": 57310 }, { "epoch": 1.11, "learning_rate": 3.146717647134937e-05, "loss": 0.0, "step": 57312 }, { "epoch": 1.11, "learning_rate": 3.146652951109214e-05, "loss": 0.0, "step": 57314 }, { "epoch": 1.11, "learning_rate": 3.14658825508349e-05, "loss": 0.0, "step": 57316 }, { "epoch": 1.11, "learning_rate": 3.146523559057768e-05, "loss": 0.0, "step": 57318 }, { "epoch": 1.11, "learning_rate": 3.146458863032044e-05, "loss": 0.0001, "step": 57320 }, { "epoch": 1.11, "learning_rate": 3.146394167006321e-05, "loss": 0.002, "step": 57322 }, { "epoch": 1.11, "learning_rate": 3.146329470980598e-05, "loss": 0.0, "step": 57324 }, { "epoch": 1.11, "learning_rate": 3.1462647749548746e-05, "loss": 0.0, "step": 57326 }, { "epoch": 1.11, "learning_rate": 3.1462000789291516e-05, "loss": 0.0, "step": 57328 }, { "epoch": 1.11, "learning_rate": 3.1461353829034285e-05, "loss": 0.0, "step": 57330 }, { "epoch": 1.11, "learning_rate": 3.1460706868777054e-05, "loss": 0.0009, "step": 57332 }, { "epoch": 1.11, "learning_rate": 3.146005990851982e-05, "loss": 0.0, "step": 57334 }, { "epoch": 1.11, "learning_rate": 3.145941294826259e-05, "loss": 0.0006, "step": 57336 }, { "epoch": 1.11, "learning_rate": 3.1458765988005354e-05, "loss": 0.0, "step": 57338 }, { "epoch": 1.11, "learning_rate": 3.145811902774813e-05, "loss": 0.0001, "step": 57340 }, { "epoch": 1.11, "learning_rate": 3.145747206749089e-05, "loss": 0.0, "step": 57342 }, { "epoch": 1.11, "learning_rate": 3.145682510723366e-05, "loss": 0.0, "step": 57344 }, { "epoch": 1.11, "learning_rate": 3.145617814697643e-05, "loss": 0.0008, "step": 57346 }, { "epoch": 1.11, "learning_rate": 3.14555311867192e-05, "loss": 0.0, "step": 57348 }, { "epoch": 1.11, "learning_rate": 3.1454884226461975e-05, "loss": 0.0, "step": 57350 }, { "epoch": 1.11, "learning_rate": 3.145423726620474e-05, "loss": 0.0, "step": 57352 }, { "epoch": 1.11, "learning_rate": 3.145359030594751e-05, "loss": 0.0, "step": 57354 }, { "epoch": 1.11, "learning_rate": 3.1452943345690276e-05, "loss": 0.0001, "step": 57356 }, { "epoch": 1.11, "learning_rate": 3.1452296385433045e-05, "loss": 0.0, "step": 57358 }, { "epoch": 1.11, "learning_rate": 3.1451649425175814e-05, "loss": 0.0033, "step": 57360 }, { "epoch": 1.11, "learning_rate": 3.145100246491858e-05, "loss": 0.0, "step": 57362 }, { "epoch": 1.11, "learning_rate": 3.145035550466135e-05, "loss": 0.0, "step": 57364 }, { "epoch": 1.11, "learning_rate": 3.1449708544404115e-05, "loss": 0.0006, "step": 57366 }, { "epoch": 1.11, "learning_rate": 3.144906158414689e-05, "loss": 0.0, "step": 57368 }, { "epoch": 1.11, "learning_rate": 3.144841462388965e-05, "loss": 0.0003, "step": 57370 }, { "epoch": 1.11, "learning_rate": 3.144776766363243e-05, "loss": 0.0003, "step": 57372 }, { "epoch": 1.11, "learning_rate": 3.144712070337519e-05, "loss": 0.0, "step": 57374 }, { "epoch": 1.11, "learning_rate": 3.144647374311796e-05, "loss": 0.0, "step": 57376 }, { "epoch": 1.11, "learning_rate": 3.144582678286073e-05, "loss": 0.0, "step": 57378 }, { "epoch": 1.11, "learning_rate": 3.14451798226035e-05, "loss": 0.0, "step": 57380 }, { "epoch": 1.11, "learning_rate": 3.144453286234627e-05, "loss": 0.0219, "step": 57382 }, { "epoch": 1.11, "learning_rate": 3.1443885902089036e-05, "loss": 0.0, "step": 57384 }, { "epoch": 1.11, "learning_rate": 3.1443238941831805e-05, "loss": 0.0, "step": 57386 }, { "epoch": 1.11, "learning_rate": 3.144259198157457e-05, "loss": 0.0, "step": 57388 }, { "epoch": 1.11, "learning_rate": 3.1441945021317344e-05, "loss": 0.0, "step": 57390 }, { "epoch": 1.11, "learning_rate": 3.144129806106011e-05, "loss": 0.0002, "step": 57392 }, { "epoch": 1.11, "learning_rate": 3.144065110080288e-05, "loss": 0.0, "step": 57394 }, { "epoch": 1.11, "learning_rate": 3.144000414054565e-05, "loss": 0.0, "step": 57396 }, { "epoch": 1.11, "learning_rate": 3.143935718028841e-05, "loss": 0.0, "step": 57398 }, { "epoch": 1.11, "learning_rate": 3.143871022003119e-05, "loss": 0.0, "step": 57400 }, { "epoch": 1.11, "learning_rate": 3.143806325977395e-05, "loss": 0.0001, "step": 57402 }, { "epoch": 1.11, "learning_rate": 3.143741629951672e-05, "loss": 0.0, "step": 57404 }, { "epoch": 1.11, "learning_rate": 3.143676933925949e-05, "loss": 0.0, "step": 57406 }, { "epoch": 1.11, "learning_rate": 3.143612237900226e-05, "loss": 0.0, "step": 57408 }, { "epoch": 1.11, "learning_rate": 3.143547541874503e-05, "loss": 0.003, "step": 57410 }, { "epoch": 1.11, "learning_rate": 3.14348284584878e-05, "loss": 0.0001, "step": 57412 }, { "epoch": 1.11, "learning_rate": 3.1434181498230566e-05, "loss": 0.0, "step": 57414 }, { "epoch": 1.11, "learning_rate": 3.1433534537973335e-05, "loss": 0.0, "step": 57416 }, { "epoch": 1.11, "learning_rate": 3.1432887577716104e-05, "loss": 0.0, "step": 57418 }, { "epoch": 1.11, "learning_rate": 3.1432240617458866e-05, "loss": 0.0062, "step": 57420 }, { "epoch": 1.11, "learning_rate": 3.143159365720164e-05, "loss": 0.0, "step": 57422 }, { "epoch": 1.11, "learning_rate": 3.1430946696944405e-05, "loss": 0.0, "step": 57424 }, { "epoch": 1.11, "learning_rate": 3.1430299736687174e-05, "loss": 0.0, "step": 57426 }, { "epoch": 1.11, "learning_rate": 3.142965277642995e-05, "loss": 0.0001, "step": 57428 }, { "epoch": 1.11, "learning_rate": 3.142900581617271e-05, "loss": 0.0, "step": 57430 }, { "epoch": 1.11, "learning_rate": 3.142835885591549e-05, "loss": 0.0002, "step": 57432 }, { "epoch": 1.11, "learning_rate": 3.142771189565825e-05, "loss": 0.0041, "step": 57434 }, { "epoch": 1.11, "learning_rate": 3.142706493540102e-05, "loss": 0.0, "step": 57436 }, { "epoch": 1.11, "learning_rate": 3.142641797514379e-05, "loss": 0.0, "step": 57438 }, { "epoch": 1.11, "learning_rate": 3.142577101488656e-05, "loss": 0.0009, "step": 57440 }, { "epoch": 1.11, "learning_rate": 3.1425124054629326e-05, "loss": 0.0, "step": 57442 }, { "epoch": 1.11, "learning_rate": 3.1424477094372095e-05, "loss": 0.0003, "step": 57444 }, { "epoch": 1.11, "learning_rate": 3.1423830134114864e-05, "loss": 0.0, "step": 57446 }, { "epoch": 1.11, "learning_rate": 3.142318317385763e-05, "loss": 0.0056, "step": 57448 }, { "epoch": 1.12, "learning_rate": 3.14225362136004e-05, "loss": 0.0, "step": 57450 }, { "epoch": 1.12, "learning_rate": 3.1421889253343165e-05, "loss": 0.0079, "step": 57452 }, { "epoch": 1.12, "learning_rate": 3.142124229308594e-05, "loss": 0.0, "step": 57454 }, { "epoch": 1.12, "learning_rate": 3.14205953328287e-05, "loss": 0.006, "step": 57456 }, { "epoch": 1.12, "learning_rate": 3.141994837257147e-05, "loss": 0.0, "step": 57458 }, { "epoch": 1.12, "learning_rate": 3.141930141231425e-05, "loss": 0.0001, "step": 57460 }, { "epoch": 1.12, "learning_rate": 3.141865445205701e-05, "loss": 0.0, "step": 57462 }, { "epoch": 1.12, "learning_rate": 3.141800749179978e-05, "loss": 0.0, "step": 57464 }, { "epoch": 1.12, "learning_rate": 3.141736053154255e-05, "loss": 0.0025, "step": 57466 }, { "epoch": 1.12, "learning_rate": 3.141671357128532e-05, "loss": 0.0002, "step": 57468 }, { "epoch": 1.12, "learning_rate": 3.141606661102809e-05, "loss": 0.0, "step": 57470 }, { "epoch": 1.12, "learning_rate": 3.1415419650770856e-05, "loss": 0.0055, "step": 57472 }, { "epoch": 1.12, "learning_rate": 3.1414772690513625e-05, "loss": 0.0, "step": 57474 }, { "epoch": 1.12, "learning_rate": 3.1414125730256394e-05, "loss": 0.0, "step": 57476 }, { "epoch": 1.12, "learning_rate": 3.141347876999916e-05, "loss": 0.0, "step": 57478 }, { "epoch": 1.12, "learning_rate": 3.1412831809741925e-05, "loss": 0.0, "step": 57480 }, { "epoch": 1.12, "learning_rate": 3.14121848494847e-05, "loss": 0.0, "step": 57482 }, { "epoch": 1.12, "learning_rate": 3.1411537889227463e-05, "loss": 0.0, "step": 57484 }, { "epoch": 1.12, "learning_rate": 3.141089092897023e-05, "loss": 0.0001, "step": 57486 }, { "epoch": 1.12, "learning_rate": 3.1410243968713e-05, "loss": 0.0, "step": 57488 }, { "epoch": 1.12, "learning_rate": 3.140959700845577e-05, "loss": 0.0016, "step": 57490 }, { "epoch": 1.12, "learning_rate": 3.1408950048198547e-05, "loss": 0.0, "step": 57492 }, { "epoch": 1.12, "learning_rate": 3.140830308794131e-05, "loss": 0.0001, "step": 57494 }, { "epoch": 1.12, "learning_rate": 3.140765612768408e-05, "loss": 0.0004, "step": 57496 }, { "epoch": 1.12, "learning_rate": 3.140700916742685e-05, "loss": 0.0, "step": 57498 }, { "epoch": 1.12, "learning_rate": 3.1406362207169616e-05, "loss": 0.0, "step": 57500 }, { "epoch": 1.12, "learning_rate": 3.140571524691238e-05, "loss": 0.0, "step": 57502 }, { "epoch": 1.12, "learning_rate": 3.1405068286655154e-05, "loss": 0.0, "step": 57504 }, { "epoch": 1.12, "learning_rate": 3.1404421326397923e-05, "loss": 0.0, "step": 57506 }, { "epoch": 1.12, "learning_rate": 3.1403774366140686e-05, "loss": 0.0, "step": 57508 }, { "epoch": 1.12, "learning_rate": 3.140312740588346e-05, "loss": 0.0, "step": 57510 }, { "epoch": 1.12, "learning_rate": 3.1402480445626224e-05, "loss": 0.0, "step": 57512 }, { "epoch": 1.12, "learning_rate": 3.1401833485369e-05, "loss": 0.0, "step": 57514 }, { "epoch": 1.12, "learning_rate": 3.140118652511176e-05, "loss": 0.0, "step": 57516 }, { "epoch": 1.12, "learning_rate": 3.140053956485453e-05, "loss": 0.0, "step": 57518 }, { "epoch": 1.12, "learning_rate": 3.13998926045973e-05, "loss": 0.0002, "step": 57520 }, { "epoch": 1.12, "learning_rate": 3.139924564434007e-05, "loss": 0.0, "step": 57522 }, { "epoch": 1.12, "learning_rate": 3.139859868408284e-05, "loss": 0.0, "step": 57524 }, { "epoch": 1.12, "learning_rate": 3.139795172382561e-05, "loss": 0.0, "step": 57526 }, { "epoch": 1.12, "learning_rate": 3.1397304763568377e-05, "loss": 0.0003, "step": 57528 }, { "epoch": 1.12, "learning_rate": 3.139665780331114e-05, "loss": 0.0, "step": 57530 }, { "epoch": 1.12, "learning_rate": 3.1396010843053915e-05, "loss": 0.0, "step": 57532 }, { "epoch": 1.12, "learning_rate": 3.139536388279668e-05, "loss": 0.0001, "step": 57534 }, { "epoch": 1.12, "learning_rate": 3.139471692253945e-05, "loss": 0.0, "step": 57536 }, { "epoch": 1.12, "learning_rate": 3.139406996228222e-05, "loss": 0.0, "step": 57538 }, { "epoch": 1.12, "learning_rate": 3.1393423002024984e-05, "loss": 0.0009, "step": 57540 }, { "epoch": 1.12, "learning_rate": 3.139277604176776e-05, "loss": 0.0, "step": 57542 }, { "epoch": 1.12, "learning_rate": 3.139212908151052e-05, "loss": 0.0, "step": 57544 }, { "epoch": 1.12, "learning_rate": 3.139148212125329e-05, "loss": 0.0, "step": 57546 }, { "epoch": 1.12, "learning_rate": 3.139083516099606e-05, "loss": 0.0006, "step": 57548 }, { "epoch": 1.12, "learning_rate": 3.139018820073883e-05, "loss": 0.0, "step": 57550 }, { "epoch": 1.12, "learning_rate": 3.13895412404816e-05, "loss": 0.0, "step": 57552 }, { "epoch": 1.12, "learning_rate": 3.138889428022437e-05, "loss": 0.0, "step": 57554 }, { "epoch": 1.12, "learning_rate": 3.138824731996714e-05, "loss": 0.0009, "step": 57556 }, { "epoch": 1.12, "learning_rate": 3.1387600359709906e-05, "loss": 0.0, "step": 57558 }, { "epoch": 1.12, "learning_rate": 3.1386953399452675e-05, "loss": 0.0003, "step": 57560 }, { "epoch": 1.12, "learning_rate": 3.138630643919544e-05, "loss": 0.0002, "step": 57562 }, { "epoch": 1.12, "learning_rate": 3.138565947893821e-05, "loss": 0.0, "step": 57564 }, { "epoch": 1.12, "learning_rate": 3.1385012518680976e-05, "loss": 0.0, "step": 57566 }, { "epoch": 1.12, "learning_rate": 3.1384365558423745e-05, "loss": 0.0, "step": 57568 }, { "epoch": 1.12, "learning_rate": 3.1383718598166514e-05, "loss": 0.0, "step": 57570 }, { "epoch": 1.12, "learning_rate": 3.138307163790928e-05, "loss": 0.0025, "step": 57572 }, { "epoch": 1.12, "learning_rate": 3.138242467765206e-05, "loss": 0.0001, "step": 57574 }, { "epoch": 1.12, "learning_rate": 3.138177771739482e-05, "loss": 0.0001, "step": 57576 }, { "epoch": 1.12, "learning_rate": 3.138113075713759e-05, "loss": 0.0003, "step": 57578 }, { "epoch": 1.12, "learning_rate": 3.138048379688036e-05, "loss": 0.0001, "step": 57580 }, { "epoch": 1.12, "learning_rate": 3.137983683662313e-05, "loss": 0.0, "step": 57582 }, { "epoch": 1.12, "learning_rate": 3.13791898763659e-05, "loss": 0.0001, "step": 57584 }, { "epoch": 1.12, "learning_rate": 3.1378542916108666e-05, "loss": 0.0, "step": 57586 }, { "epoch": 1.12, "learning_rate": 3.1377895955851436e-05, "loss": 0.0, "step": 57588 }, { "epoch": 1.12, "learning_rate": 3.13772489955942e-05, "loss": 0.0003, "step": 57590 }, { "epoch": 1.12, "learning_rate": 3.1376602035336974e-05, "loss": 0.0, "step": 57592 }, { "epoch": 1.12, "learning_rate": 3.1375955075079736e-05, "loss": 0.0, "step": 57594 }, { "epoch": 1.12, "learning_rate": 3.137530811482251e-05, "loss": 0.0, "step": 57596 }, { "epoch": 1.12, "learning_rate": 3.1374661154565274e-05, "loss": 0.0, "step": 57598 }, { "epoch": 1.12, "learning_rate": 3.137401419430804e-05, "loss": 0.0001, "step": 57600 }, { "epoch": 1.12, "learning_rate": 3.137336723405081e-05, "loss": 0.0048, "step": 57602 }, { "epoch": 1.12, "learning_rate": 3.137272027379358e-05, "loss": 0.0, "step": 57604 }, { "epoch": 1.12, "learning_rate": 3.137207331353635e-05, "loss": 0.0001, "step": 57606 }, { "epoch": 1.12, "learning_rate": 3.137142635327912e-05, "loss": 0.0005, "step": 57608 }, { "epoch": 1.12, "learning_rate": 3.137077939302189e-05, "loss": 0.0, "step": 57610 }, { "epoch": 1.12, "learning_rate": 3.137013243276466e-05, "loss": 0.0001, "step": 57612 }, { "epoch": 1.12, "learning_rate": 3.136948547250743e-05, "loss": 0.0, "step": 57614 }, { "epoch": 1.12, "learning_rate": 3.1368838512250196e-05, "loss": 0.0, "step": 57616 }, { "epoch": 1.12, "learning_rate": 3.1368191551992965e-05, "loss": 0.0, "step": 57618 }, { "epoch": 1.12, "learning_rate": 3.1367544591735734e-05, "loss": 0.0, "step": 57620 }, { "epoch": 1.12, "learning_rate": 3.1366897631478496e-05, "loss": 0.0, "step": 57622 }, { "epoch": 1.12, "learning_rate": 3.136625067122127e-05, "loss": 0.0, "step": 57624 }, { "epoch": 1.12, "learning_rate": 3.1365603710964035e-05, "loss": 0.0, "step": 57626 }, { "epoch": 1.12, "learning_rate": 3.1364956750706804e-05, "loss": 0.0, "step": 57628 }, { "epoch": 1.12, "learning_rate": 3.136430979044957e-05, "loss": 0.0, "step": 57630 }, { "epoch": 1.12, "learning_rate": 3.136366283019234e-05, "loss": 0.0, "step": 57632 }, { "epoch": 1.12, "learning_rate": 3.136301586993511e-05, "loss": 0.0, "step": 57634 }, { "epoch": 1.12, "learning_rate": 3.136236890967788e-05, "loss": 0.0, "step": 57636 }, { "epoch": 1.12, "learning_rate": 3.136172194942065e-05, "loss": 0.0, "step": 57638 }, { "epoch": 1.12, "learning_rate": 3.136107498916342e-05, "loss": 0.0, "step": 57640 }, { "epoch": 1.12, "learning_rate": 3.136042802890619e-05, "loss": 0.0, "step": 57642 }, { "epoch": 1.12, "learning_rate": 3.135978106864895e-05, "loss": 0.0081, "step": 57644 }, { "epoch": 1.12, "learning_rate": 3.1359134108391725e-05, "loss": 0.0, "step": 57646 }, { "epoch": 1.12, "learning_rate": 3.135848714813449e-05, "loss": 0.0, "step": 57648 }, { "epoch": 1.12, "learning_rate": 3.135784018787726e-05, "loss": 0.0, "step": 57650 }, { "epoch": 1.12, "learning_rate": 3.135719322762003e-05, "loss": 0.0017, "step": 57652 }, { "epoch": 1.12, "learning_rate": 3.1356546267362795e-05, "loss": 0.0, "step": 57654 }, { "epoch": 1.12, "learning_rate": 3.135589930710557e-05, "loss": 0.0, "step": 57656 }, { "epoch": 1.12, "learning_rate": 3.135525234684833e-05, "loss": 0.0, "step": 57658 }, { "epoch": 1.12, "learning_rate": 3.13546053865911e-05, "loss": 0.0121, "step": 57660 }, { "epoch": 1.12, "learning_rate": 3.135395842633387e-05, "loss": 0.0, "step": 57662 }, { "epoch": 1.12, "learning_rate": 3.135331146607664e-05, "loss": 0.0168, "step": 57664 }, { "epoch": 1.12, "learning_rate": 3.135266450581941e-05, "loss": 0.0003, "step": 57666 }, { "epoch": 1.12, "learning_rate": 3.135201754556218e-05, "loss": 0.0, "step": 57668 }, { "epoch": 1.12, "learning_rate": 3.135137058530495e-05, "loss": 0.0, "step": 57670 }, { "epoch": 1.12, "learning_rate": 3.135072362504771e-05, "loss": 0.0, "step": 57672 }, { "epoch": 1.12, "learning_rate": 3.1350076664790486e-05, "loss": 0.0, "step": 57674 }, { "epoch": 1.12, "learning_rate": 3.134942970453325e-05, "loss": 0.0001, "step": 57676 }, { "epoch": 1.12, "learning_rate": 3.1348782744276024e-05, "loss": 0.0, "step": 57678 }, { "epoch": 1.12, "learning_rate": 3.1348135784018786e-05, "loss": 0.0, "step": 57680 }, { "epoch": 1.12, "learning_rate": 3.1347488823761555e-05, "loss": 0.0, "step": 57682 }, { "epoch": 1.12, "learning_rate": 3.134684186350433e-05, "loss": 0.0019, "step": 57684 }, { "epoch": 1.12, "learning_rate": 3.1346194903247094e-05, "loss": 0.0, "step": 57686 }, { "epoch": 1.12, "learning_rate": 3.134554794298986e-05, "loss": 0.0001, "step": 57688 }, { "epoch": 1.12, "learning_rate": 3.134490098273263e-05, "loss": 0.0113, "step": 57690 }, { "epoch": 1.12, "learning_rate": 3.13442540224754e-05, "loss": 0.0001, "step": 57692 }, { "epoch": 1.12, "learning_rate": 3.134360706221817e-05, "loss": 0.0, "step": 57694 }, { "epoch": 1.12, "learning_rate": 3.134296010196094e-05, "loss": 0.0, "step": 57696 }, { "epoch": 1.12, "learning_rate": 3.134231314170371e-05, "loss": 0.0, "step": 57698 }, { "epoch": 1.12, "learning_rate": 3.134166618144648e-05, "loss": 0.0035, "step": 57700 }, { "epoch": 1.12, "learning_rate": 3.1341019221189246e-05, "loss": 0.0, "step": 57702 }, { "epoch": 1.12, "learning_rate": 3.134037226093201e-05, "loss": 0.0, "step": 57704 }, { "epoch": 1.12, "learning_rate": 3.1339725300674784e-05, "loss": 0.0, "step": 57706 }, { "epoch": 1.12, "learning_rate": 3.133907834041755e-05, "loss": 0.0, "step": 57708 }, { "epoch": 1.12, "learning_rate": 3.1338431380160316e-05, "loss": 0.0, "step": 57710 }, { "epoch": 1.12, "learning_rate": 3.1337784419903085e-05, "loss": 0.0001, "step": 57712 }, { "epoch": 1.12, "learning_rate": 3.1337137459645854e-05, "loss": 0.0006, "step": 57714 }, { "epoch": 1.12, "learning_rate": 3.133649049938863e-05, "loss": 0.0003, "step": 57716 }, { "epoch": 1.12, "learning_rate": 3.133584353913139e-05, "loss": 0.0, "step": 57718 }, { "epoch": 1.12, "learning_rate": 3.133519657887416e-05, "loss": 0.0, "step": 57720 }, { "epoch": 1.12, "learning_rate": 3.133454961861693e-05, "loss": 0.0053, "step": 57722 }, { "epoch": 1.12, "learning_rate": 3.13339026583597e-05, "loss": 0.0, "step": 57724 }, { "epoch": 1.12, "learning_rate": 3.133325569810246e-05, "loss": 0.0111, "step": 57726 }, { "epoch": 1.12, "learning_rate": 3.133260873784524e-05, "loss": 0.0024, "step": 57728 }, { "epoch": 1.12, "learning_rate": 3.133196177758801e-05, "loss": 0.0, "step": 57730 }, { "epoch": 1.12, "learning_rate": 3.133131481733077e-05, "loss": 0.0001, "step": 57732 }, { "epoch": 1.12, "learning_rate": 3.1330667857073545e-05, "loss": 0.0, "step": 57734 }, { "epoch": 1.12, "learning_rate": 3.133002089681631e-05, "loss": 0.0, "step": 57736 }, { "epoch": 1.12, "learning_rate": 3.132937393655908e-05, "loss": 0.0021, "step": 57738 }, { "epoch": 1.12, "learning_rate": 3.1328726976301845e-05, "loss": 0.0002, "step": 57740 }, { "epoch": 1.12, "learning_rate": 3.1328080016044614e-05, "loss": 0.0009, "step": 57742 }, { "epoch": 1.12, "learning_rate": 3.1327433055787383e-05, "loss": 0.0, "step": 57744 }, { "epoch": 1.12, "learning_rate": 3.132678609553015e-05, "loss": 0.0, "step": 57746 }, { "epoch": 1.12, "learning_rate": 3.132613913527292e-05, "loss": 0.0001, "step": 57748 }, { "epoch": 1.12, "learning_rate": 3.132549217501569e-05, "loss": 0.0032, "step": 57750 }, { "epoch": 1.12, "learning_rate": 3.132484521475846e-05, "loss": 0.0, "step": 57752 }, { "epoch": 1.12, "learning_rate": 3.132419825450123e-05, "loss": 0.0, "step": 57754 }, { "epoch": 1.12, "learning_rate": 3.1323551294244e-05, "loss": 0.0, "step": 57756 }, { "epoch": 1.12, "learning_rate": 3.132290433398676e-05, "loss": 0.0001, "step": 57758 }, { "epoch": 1.12, "learning_rate": 3.1322257373729536e-05, "loss": 0.0001, "step": 57760 }, { "epoch": 1.12, "learning_rate": 3.1321610413472305e-05, "loss": 0.0055, "step": 57762 }, { "epoch": 1.12, "learning_rate": 3.132096345321507e-05, "loss": 0.0, "step": 57764 }, { "epoch": 1.12, "learning_rate": 3.1320316492957843e-05, "loss": 0.0002, "step": 57766 }, { "epoch": 1.12, "learning_rate": 3.1319669532700606e-05, "loss": 0.0001, "step": 57768 }, { "epoch": 1.12, "learning_rate": 3.1319022572443375e-05, "loss": 0.0017, "step": 57770 }, { "epoch": 1.12, "learning_rate": 3.1318375612186144e-05, "loss": 0.0, "step": 57772 }, { "epoch": 1.12, "learning_rate": 3.131772865192891e-05, "loss": 0.0001, "step": 57774 }, { "epoch": 1.12, "learning_rate": 3.131708169167168e-05, "loss": 0.0, "step": 57776 }, { "epoch": 1.12, "learning_rate": 3.131643473141445e-05, "loss": 0.0002, "step": 57778 }, { "epoch": 1.12, "learning_rate": 3.131578777115722e-05, "loss": 0.0, "step": 57780 }, { "epoch": 1.12, "learning_rate": 3.131514081089999e-05, "loss": 0.0001, "step": 57782 }, { "epoch": 1.12, "learning_rate": 3.131449385064276e-05, "loss": 0.0, "step": 57784 }, { "epoch": 1.12, "learning_rate": 3.131384689038552e-05, "loss": 0.0, "step": 57786 }, { "epoch": 1.12, "learning_rate": 3.1313199930128297e-05, "loss": 0.0, "step": 57788 }, { "epoch": 1.12, "learning_rate": 3.131255296987106e-05, "loss": 0.0, "step": 57790 }, { "epoch": 1.12, "learning_rate": 3.131190600961383e-05, "loss": 0.0, "step": 57792 }, { "epoch": 1.12, "learning_rate": 3.1311259049356604e-05, "loss": 0.0, "step": 57794 }, { "epoch": 1.12, "learning_rate": 3.1310612089099366e-05, "loss": 0.0, "step": 57796 }, { "epoch": 1.12, "learning_rate": 3.130996512884214e-05, "loss": 0.0, "step": 57798 }, { "epoch": 1.12, "learning_rate": 3.1309318168584904e-05, "loss": 0.0065, "step": 57800 }, { "epoch": 1.12, "learning_rate": 3.130867120832767e-05, "loss": 0.0001, "step": 57802 }, { "epoch": 1.12, "learning_rate": 3.130802424807044e-05, "loss": 0.0, "step": 57804 }, { "epoch": 1.12, "learning_rate": 3.130737728781321e-05, "loss": 0.0, "step": 57806 }, { "epoch": 1.12, "learning_rate": 3.130673032755598e-05, "loss": 0.0107, "step": 57808 }, { "epoch": 1.12, "learning_rate": 3.130608336729875e-05, "loss": 0.0, "step": 57810 }, { "epoch": 1.12, "learning_rate": 3.130543640704152e-05, "loss": 0.007, "step": 57812 }, { "epoch": 1.12, "learning_rate": 3.130478944678428e-05, "loss": 0.0, "step": 57814 }, { "epoch": 1.12, "learning_rate": 3.130414248652706e-05, "loss": 0.0, "step": 57816 }, { "epoch": 1.12, "learning_rate": 3.130349552626982e-05, "loss": 0.0001, "step": 57818 }, { "epoch": 1.12, "learning_rate": 3.1302848566012595e-05, "loss": 0.0, "step": 57820 }, { "epoch": 1.12, "learning_rate": 3.130220160575536e-05, "loss": 0.0041, "step": 57822 }, { "epoch": 1.12, "learning_rate": 3.1301554645498127e-05, "loss": 0.0, "step": 57824 }, { "epoch": 1.12, "learning_rate": 3.1300907685240896e-05, "loss": 0.0, "step": 57826 }, { "epoch": 1.12, "learning_rate": 3.1300260724983665e-05, "loss": 0.0, "step": 57828 }, { "epoch": 1.12, "learning_rate": 3.1299613764726434e-05, "loss": 0.0002, "step": 57830 }, { "epoch": 1.12, "learning_rate": 3.12989668044692e-05, "loss": 0.0001, "step": 57832 }, { "epoch": 1.12, "learning_rate": 3.129831984421197e-05, "loss": 0.0, "step": 57834 }, { "epoch": 1.12, "learning_rate": 3.129767288395474e-05, "loss": 0.0, "step": 57836 }, { "epoch": 1.12, "learning_rate": 3.129702592369751e-05, "loss": 0.0029, "step": 57838 }, { "epoch": 1.12, "learning_rate": 3.129637896344028e-05, "loss": 0.0, "step": 57840 }, { "epoch": 1.12, "learning_rate": 3.129573200318305e-05, "loss": 0.0, "step": 57842 }, { "epoch": 1.12, "learning_rate": 3.129508504292582e-05, "loss": 0.0, "step": 57844 }, { "epoch": 1.12, "learning_rate": 3.129443808266858e-05, "loss": 0.0001, "step": 57846 }, { "epoch": 1.12, "learning_rate": 3.1293791122411356e-05, "loss": 0.0, "step": 57848 }, { "epoch": 1.12, "learning_rate": 3.129314416215412e-05, "loss": 0.0, "step": 57850 }, { "epoch": 1.12, "learning_rate": 3.129249720189689e-05, "loss": 0.0041, "step": 57852 }, { "epoch": 1.12, "learning_rate": 3.1291850241639656e-05, "loss": 0.0, "step": 57854 }, { "epoch": 1.12, "learning_rate": 3.1291203281382425e-05, "loss": 0.0, "step": 57856 }, { "epoch": 1.12, "learning_rate": 3.1290556321125194e-05, "loss": 0.0, "step": 57858 }, { "epoch": 1.12, "learning_rate": 3.128990936086796e-05, "loss": 0.0, "step": 57860 }, { "epoch": 1.12, "learning_rate": 3.128926240061073e-05, "loss": 0.0093, "step": 57862 }, { "epoch": 1.12, "learning_rate": 3.12886154403535e-05, "loss": 0.0085, "step": 57864 }, { "epoch": 1.12, "learning_rate": 3.128796848009627e-05, "loss": 0.0, "step": 57866 }, { "epoch": 1.12, "learning_rate": 3.128732151983903e-05, "loss": 0.0049, "step": 57868 }, { "epoch": 1.12, "learning_rate": 3.128667455958181e-05, "loss": 0.0, "step": 57870 }, { "epoch": 1.12, "learning_rate": 3.128602759932458e-05, "loss": 0.0, "step": 57872 }, { "epoch": 1.12, "learning_rate": 3.128538063906734e-05, "loss": 0.0, "step": 57874 }, { "epoch": 1.12, "learning_rate": 3.1284733678810116e-05, "loss": 0.0013, "step": 57876 }, { "epoch": 1.12, "learning_rate": 3.128408671855288e-05, "loss": 0.0, "step": 57878 }, { "epoch": 1.12, "learning_rate": 3.1283439758295654e-05, "loss": 0.0, "step": 57880 }, { "epoch": 1.12, "learning_rate": 3.1282792798038416e-05, "loss": 0.0, "step": 57882 }, { "epoch": 1.12, "learning_rate": 3.1282145837781185e-05, "loss": 0.0, "step": 57884 }, { "epoch": 1.12, "learning_rate": 3.1281498877523955e-05, "loss": 0.0031, "step": 57886 }, { "epoch": 1.12, "learning_rate": 3.1280851917266724e-05, "loss": 0.0, "step": 57888 }, { "epoch": 1.12, "learning_rate": 3.128020495700949e-05, "loss": 0.0, "step": 57890 }, { "epoch": 1.12, "learning_rate": 3.127955799675226e-05, "loss": 0.0, "step": 57892 }, { "epoch": 1.12, "learning_rate": 3.127891103649503e-05, "loss": 0.0, "step": 57894 }, { "epoch": 1.12, "learning_rate": 3.12782640762378e-05, "loss": 0.0, "step": 57896 }, { "epoch": 1.12, "learning_rate": 3.127761711598057e-05, "loss": 0.0003, "step": 57898 }, { "epoch": 1.12, "learning_rate": 3.127697015572333e-05, "loss": 0.0002, "step": 57900 }, { "epoch": 1.12, "learning_rate": 3.127632319546611e-05, "loss": 0.0, "step": 57902 }, { "epoch": 1.12, "learning_rate": 3.127567623520887e-05, "loss": 0.0, "step": 57904 }, { "epoch": 1.12, "learning_rate": 3.127502927495164e-05, "loss": 0.0, "step": 57906 }, { "epoch": 1.12, "learning_rate": 3.1274382314694415e-05, "loss": 0.0024, "step": 57908 }, { "epoch": 1.12, "learning_rate": 3.127373535443718e-05, "loss": 0.0016, "step": 57910 }, { "epoch": 1.12, "learning_rate": 3.1273088394179946e-05, "loss": 0.0, "step": 57912 }, { "epoch": 1.12, "learning_rate": 3.1272441433922715e-05, "loss": 0.0, "step": 57914 }, { "epoch": 1.12, "learning_rate": 3.1271794473665484e-05, "loss": 0.0001, "step": 57916 }, { "epoch": 1.12, "learning_rate": 3.127114751340825e-05, "loss": 0.0, "step": 57918 }, { "epoch": 1.12, "learning_rate": 3.127050055315102e-05, "loss": 0.0, "step": 57920 }, { "epoch": 1.12, "learning_rate": 3.126985359289379e-05, "loss": 0.0, "step": 57922 }, { "epoch": 1.12, "learning_rate": 3.126920663263656e-05, "loss": 0.0046, "step": 57924 }, { "epoch": 1.12, "learning_rate": 3.126855967237933e-05, "loss": 0.0, "step": 57926 }, { "epoch": 1.12, "learning_rate": 3.126791271212209e-05, "loss": 0.0, "step": 57928 }, { "epoch": 1.12, "learning_rate": 3.126726575186487e-05, "loss": 0.0, "step": 57930 }, { "epoch": 1.12, "learning_rate": 3.126661879160763e-05, "loss": 0.0251, "step": 57932 }, { "epoch": 1.12, "learning_rate": 3.12659718313504e-05, "loss": 0.0052, "step": 57934 }, { "epoch": 1.12, "learning_rate": 3.126532487109317e-05, "loss": 0.0033, "step": 57936 }, { "epoch": 1.12, "learning_rate": 3.126467791083594e-05, "loss": 0.0, "step": 57938 }, { "epoch": 1.12, "learning_rate": 3.126403095057871e-05, "loss": 0.0008, "step": 57940 }, { "epoch": 1.12, "learning_rate": 3.1263383990321475e-05, "loss": 0.0, "step": 57942 }, { "epoch": 1.12, "learning_rate": 3.1262737030064244e-05, "loss": 0.0, "step": 57944 }, { "epoch": 1.12, "learning_rate": 3.1262090069807014e-05, "loss": 0.0, "step": 57946 }, { "epoch": 1.12, "learning_rate": 3.126144310954978e-05, "loss": 0.0, "step": 57948 }, { "epoch": 1.12, "learning_rate": 3.1260796149292545e-05, "loss": 0.0001, "step": 57950 }, { "epoch": 1.12, "learning_rate": 3.126014918903532e-05, "loss": 0.0, "step": 57952 }, { "epoch": 1.12, "learning_rate": 3.125950222877809e-05, "loss": 0.0, "step": 57954 }, { "epoch": 1.12, "learning_rate": 3.125885526852085e-05, "loss": 0.0015, "step": 57956 }, { "epoch": 1.12, "learning_rate": 3.125820830826363e-05, "loss": 0.0, "step": 57958 }, { "epoch": 1.12, "learning_rate": 3.125756134800639e-05, "loss": 0.0, "step": 57960 }, { "epoch": 1.12, "learning_rate": 3.1256914387749166e-05, "loss": 0.0, "step": 57962 }, { "epoch": 1.13, "learning_rate": 3.125626742749193e-05, "loss": 0.0324, "step": 57964 }, { "epoch": 1.13, "learning_rate": 3.12556204672347e-05, "loss": 0.0, "step": 57966 }, { "epoch": 1.13, "learning_rate": 3.125497350697747e-05, "loss": 0.0004, "step": 57968 }, { "epoch": 1.13, "learning_rate": 3.1254326546720236e-05, "loss": 0.0017, "step": 57970 }, { "epoch": 1.13, "learning_rate": 3.1253679586463005e-05, "loss": 0.0073, "step": 57972 }, { "epoch": 1.13, "learning_rate": 3.1253032626205774e-05, "loss": 0.0039, "step": 57974 }, { "epoch": 1.13, "learning_rate": 3.125238566594854e-05, "loss": 0.001, "step": 57976 }, { "epoch": 1.13, "learning_rate": 3.125173870569131e-05, "loss": 0.0, "step": 57978 }, { "epoch": 1.13, "learning_rate": 3.125109174543408e-05, "loss": 0.0, "step": 57980 }, { "epoch": 1.13, "learning_rate": 3.1250444785176844e-05, "loss": 0.0, "step": 57982 }, { "epoch": 1.13, "learning_rate": 3.124979782491962e-05, "loss": 0.0, "step": 57984 }, { "epoch": 1.13, "learning_rate": 3.124915086466239e-05, "loss": 0.0001, "step": 57986 }, { "epoch": 1.13, "learning_rate": 3.124850390440515e-05, "loss": 0.0, "step": 57988 }, { "epoch": 1.13, "learning_rate": 3.124785694414793e-05, "loss": 0.0, "step": 57990 }, { "epoch": 1.13, "learning_rate": 3.124720998389069e-05, "loss": 0.0003, "step": 57992 }, { "epoch": 1.13, "learning_rate": 3.124656302363346e-05, "loss": 0.001, "step": 57994 }, { "epoch": 1.13, "learning_rate": 3.124591606337623e-05, "loss": 0.0, "step": 57996 }, { "epoch": 1.13, "learning_rate": 3.1245269103118996e-05, "loss": 0.0, "step": 57998 }, { "epoch": 1.13, "learning_rate": 3.1244622142861765e-05, "loss": 0.0, "step": 58000 }, { "epoch": 1.13, "learning_rate": 3.1243975182604534e-05, "loss": 0.0, "step": 58002 }, { "epoch": 1.13, "learning_rate": 3.1243328222347303e-05, "loss": 0.0, "step": 58004 }, { "epoch": 1.13, "learning_rate": 3.124268126209007e-05, "loss": 0.0, "step": 58006 }, { "epoch": 1.13, "learning_rate": 3.124203430183284e-05, "loss": 0.001, "step": 58008 }, { "epoch": 1.13, "learning_rate": 3.1241387341575604e-05, "loss": 0.0002, "step": 58010 }, { "epoch": 1.13, "learning_rate": 3.124074038131838e-05, "loss": 0.0269, "step": 58012 }, { "epoch": 1.13, "learning_rate": 3.124009342106114e-05, "loss": 0.0063, "step": 58014 }, { "epoch": 1.13, "learning_rate": 3.123944646080391e-05, "loss": 0.0, "step": 58016 }, { "epoch": 1.13, "learning_rate": 3.123879950054669e-05, "loss": 0.0, "step": 58018 }, { "epoch": 1.13, "learning_rate": 3.123815254028945e-05, "loss": 0.0, "step": 58020 }, { "epoch": 1.13, "learning_rate": 3.1237505580032225e-05, "loss": 0.0, "step": 58022 }, { "epoch": 1.13, "learning_rate": 3.123685861977499e-05, "loss": 0.0, "step": 58024 }, { "epoch": 1.13, "learning_rate": 3.1236211659517757e-05, "loss": 0.0, "step": 58026 }, { "epoch": 1.13, "learning_rate": 3.1235564699260526e-05, "loss": 0.0, "step": 58028 }, { "epoch": 1.13, "learning_rate": 3.1234917739003295e-05, "loss": 0.0, "step": 58030 }, { "epoch": 1.13, "learning_rate": 3.1234270778746064e-05, "loss": 0.0, "step": 58032 }, { "epoch": 1.13, "learning_rate": 3.123362381848883e-05, "loss": 0.0, "step": 58034 }, { "epoch": 1.13, "learning_rate": 3.12329768582316e-05, "loss": 0.0, "step": 58036 }, { "epoch": 1.13, "learning_rate": 3.123232989797437e-05, "loss": 0.0002, "step": 58038 }, { "epoch": 1.13, "learning_rate": 3.123168293771714e-05, "loss": 0.0, "step": 58040 }, { "epoch": 1.13, "learning_rate": 3.12310359774599e-05, "loss": 0.0, "step": 58042 }, { "epoch": 1.13, "learning_rate": 3.123038901720268e-05, "loss": 0.0001, "step": 58044 }, { "epoch": 1.13, "learning_rate": 3.122974205694544e-05, "loss": 0.0, "step": 58046 }, { "epoch": 1.13, "learning_rate": 3.122909509668821e-05, "loss": 0.0, "step": 58048 }, { "epoch": 1.13, "learning_rate": 3.122844813643098e-05, "loss": 0.0, "step": 58050 }, { "epoch": 1.13, "learning_rate": 3.122780117617375e-05, "loss": 0.0, "step": 58052 }, { "epoch": 1.13, "learning_rate": 3.122715421591652e-05, "loss": 0.0, "step": 58054 }, { "epoch": 1.13, "learning_rate": 3.1226507255659286e-05, "loss": 0.0001, "step": 58056 }, { "epoch": 1.13, "learning_rate": 3.1225860295402055e-05, "loss": 0.0, "step": 58058 }, { "epoch": 1.13, "learning_rate": 3.1225213335144824e-05, "loss": 0.0, "step": 58060 }, { "epoch": 1.13, "learning_rate": 3.122456637488759e-05, "loss": 0.0, "step": 58062 }, { "epoch": 1.13, "learning_rate": 3.122391941463036e-05, "loss": 0.0004, "step": 58064 }, { "epoch": 1.13, "learning_rate": 3.122327245437313e-05, "loss": 0.0, "step": 58066 }, { "epoch": 1.13, "learning_rate": 3.12226254941159e-05, "loss": 0.0001, "step": 58068 }, { "epoch": 1.13, "learning_rate": 3.122197853385866e-05, "loss": 0.0, "step": 58070 }, { "epoch": 1.13, "learning_rate": 3.122133157360144e-05, "loss": 0.0, "step": 58072 }, { "epoch": 1.13, "learning_rate": 3.12206846133442e-05, "loss": 0.0, "step": 58074 }, { "epoch": 1.13, "learning_rate": 3.122003765308697e-05, "loss": 0.0, "step": 58076 }, { "epoch": 1.13, "learning_rate": 3.121939069282974e-05, "loss": 0.0001, "step": 58078 }, { "epoch": 1.13, "learning_rate": 3.121874373257251e-05, "loss": 0.0058, "step": 58080 }, { "epoch": 1.13, "learning_rate": 3.121809677231528e-05, "loss": 0.0, "step": 58082 }, { "epoch": 1.13, "learning_rate": 3.1217449812058046e-05, "loss": 0.0059, "step": 58084 }, { "epoch": 1.13, "learning_rate": 3.1216802851800816e-05, "loss": 0.0001, "step": 58086 }, { "epoch": 1.13, "learning_rate": 3.1216155891543585e-05, "loss": 0.0, "step": 58088 }, { "epoch": 1.13, "learning_rate": 3.1215508931286354e-05, "loss": 0.0008, "step": 58090 }, { "epoch": 1.13, "learning_rate": 3.1214861971029116e-05, "loss": 0.0042, "step": 58092 }, { "epoch": 1.13, "learning_rate": 3.121421501077189e-05, "loss": 0.0, "step": 58094 }, { "epoch": 1.13, "learning_rate": 3.121356805051466e-05, "loss": 0.0, "step": 58096 }, { "epoch": 1.13, "learning_rate": 3.121292109025742e-05, "loss": 0.0, "step": 58098 }, { "epoch": 1.13, "learning_rate": 3.12122741300002e-05, "loss": 0.0, "step": 58100 }, { "epoch": 1.13, "learning_rate": 3.121162716974296e-05, "loss": 0.0001, "step": 58102 }, { "epoch": 1.13, "learning_rate": 3.121098020948574e-05, "loss": 0.0, "step": 58104 }, { "epoch": 1.13, "learning_rate": 3.12103332492285e-05, "loss": 0.0, "step": 58106 }, { "epoch": 1.13, "learning_rate": 3.120968628897127e-05, "loss": 0.0002, "step": 58108 }, { "epoch": 1.13, "learning_rate": 3.120903932871404e-05, "loss": 0.0, "step": 58110 }, { "epoch": 1.13, "learning_rate": 3.120839236845681e-05, "loss": 0.0008, "step": 58112 }, { "epoch": 1.13, "learning_rate": 3.1207745408199576e-05, "loss": 0.0, "step": 58114 }, { "epoch": 1.13, "learning_rate": 3.1207098447942345e-05, "loss": 0.0019, "step": 58116 }, { "epoch": 1.13, "learning_rate": 3.1206451487685114e-05, "loss": 0.0, "step": 58118 }, { "epoch": 1.13, "learning_rate": 3.120580452742788e-05, "loss": 0.0, "step": 58120 }, { "epoch": 1.13, "learning_rate": 3.120515756717065e-05, "loss": 0.0047, "step": 58122 }, { "epoch": 1.13, "learning_rate": 3.1204510606913415e-05, "loss": 0.0002, "step": 58124 }, { "epoch": 1.13, "learning_rate": 3.120386364665619e-05, "loss": 0.0001, "step": 58126 }, { "epoch": 1.13, "learning_rate": 3.120321668639895e-05, "loss": 0.0012, "step": 58128 }, { "epoch": 1.13, "learning_rate": 3.120256972614172e-05, "loss": 0.0001, "step": 58130 }, { "epoch": 1.13, "learning_rate": 3.12019227658845e-05, "loss": 0.0003, "step": 58132 }, { "epoch": 1.13, "learning_rate": 3.120127580562726e-05, "loss": 0.0, "step": 58134 }, { "epoch": 1.13, "learning_rate": 3.120062884537003e-05, "loss": 0.0, "step": 58136 }, { "epoch": 1.13, "learning_rate": 3.11999818851128e-05, "loss": 0.0, "step": 58138 }, { "epoch": 1.13, "learning_rate": 3.119933492485557e-05, "loss": 0.0, "step": 58140 }, { "epoch": 1.13, "learning_rate": 3.1198687964598336e-05, "loss": 0.0, "step": 58142 }, { "epoch": 1.13, "learning_rate": 3.1198041004341105e-05, "loss": 0.0, "step": 58144 }, { "epoch": 1.13, "learning_rate": 3.1197394044083875e-05, "loss": 0.0, "step": 58146 }, { "epoch": 1.13, "learning_rate": 3.1196747083826644e-05, "loss": 0.0, "step": 58148 }, { "epoch": 1.13, "learning_rate": 3.119610012356941e-05, "loss": 0.0002, "step": 58150 }, { "epoch": 1.13, "learning_rate": 3.1195453163312175e-05, "loss": 0.0, "step": 58152 }, { "epoch": 1.13, "learning_rate": 3.119480620305495e-05, "loss": 0.0001, "step": 58154 }, { "epoch": 1.13, "learning_rate": 3.119415924279771e-05, "loss": 0.0, "step": 58156 }, { "epoch": 1.13, "learning_rate": 3.119351228254048e-05, "loss": 0.0006, "step": 58158 }, { "epoch": 1.13, "learning_rate": 3.119286532228325e-05, "loss": 0.0078, "step": 58160 }, { "epoch": 1.13, "learning_rate": 3.119221836202602e-05, "loss": 0.0003, "step": 58162 }, { "epoch": 1.13, "learning_rate": 3.1191571401768796e-05, "loss": 0.0, "step": 58164 }, { "epoch": 1.13, "learning_rate": 3.119092444151156e-05, "loss": 0.0002, "step": 58166 }, { "epoch": 1.13, "learning_rate": 3.119027748125433e-05, "loss": 0.0002, "step": 58168 }, { "epoch": 1.13, "learning_rate": 3.11896305209971e-05, "loss": 0.0, "step": 58170 }, { "epoch": 1.13, "learning_rate": 3.1188983560739866e-05, "loss": 0.0, "step": 58172 }, { "epoch": 1.13, "learning_rate": 3.1188336600482635e-05, "loss": 0.0, "step": 58174 }, { "epoch": 1.13, "learning_rate": 3.1187689640225404e-05, "loss": 0.0, "step": 58176 }, { "epoch": 1.13, "learning_rate": 3.118704267996817e-05, "loss": 0.0001, "step": 58178 }, { "epoch": 1.13, "learning_rate": 3.118639571971094e-05, "loss": 0.0, "step": 58180 }, { "epoch": 1.13, "learning_rate": 3.118574875945371e-05, "loss": 0.0, "step": 58182 }, { "epoch": 1.13, "learning_rate": 3.1185101799196474e-05, "loss": 0.0, "step": 58184 }, { "epoch": 1.13, "learning_rate": 3.118445483893925e-05, "loss": 0.0003, "step": 58186 }, { "epoch": 1.13, "learning_rate": 3.118380787868201e-05, "loss": 0.0002, "step": 58188 }, { "epoch": 1.13, "learning_rate": 3.118316091842478e-05, "loss": 0.003, "step": 58190 }, { "epoch": 1.13, "learning_rate": 3.118251395816755e-05, "loss": 0.0, "step": 58192 }, { "epoch": 1.13, "learning_rate": 3.118186699791032e-05, "loss": 0.0, "step": 58194 }, { "epoch": 1.13, "learning_rate": 3.118122003765309e-05, "loss": 0.0, "step": 58196 }, { "epoch": 1.13, "learning_rate": 3.118057307739586e-05, "loss": 0.0, "step": 58198 }, { "epoch": 1.13, "learning_rate": 3.1179926117138626e-05, "loss": 0.0, "step": 58200 }, { "epoch": 1.13, "learning_rate": 3.1179279156881395e-05, "loss": 0.0, "step": 58202 }, { "epoch": 1.13, "learning_rate": 3.1178632196624164e-05, "loss": 0.0001, "step": 58204 }, { "epoch": 1.13, "learning_rate": 3.117798523636693e-05, "loss": 0.0, "step": 58206 }, { "epoch": 1.13, "learning_rate": 3.11773382761097e-05, "loss": 0.0006, "step": 58208 }, { "epoch": 1.13, "learning_rate": 3.117669131585247e-05, "loss": 0.0001, "step": 58210 }, { "epoch": 1.13, "learning_rate": 3.1176044355595234e-05, "loss": 0.0008, "step": 58212 }, { "epoch": 1.13, "learning_rate": 3.117539739533801e-05, "loss": 0.0007, "step": 58214 }, { "epoch": 1.13, "learning_rate": 3.117475043508077e-05, "loss": 0.0001, "step": 58216 }, { "epoch": 1.13, "learning_rate": 3.117410347482354e-05, "loss": 0.0, "step": 58218 }, { "epoch": 1.13, "learning_rate": 3.117345651456631e-05, "loss": 0.0, "step": 58220 }, { "epoch": 1.13, "learning_rate": 3.117280955430908e-05, "loss": 0.0011, "step": 58222 }, { "epoch": 1.13, "learning_rate": 3.117216259405185e-05, "loss": 0.0271, "step": 58224 }, { "epoch": 1.13, "learning_rate": 3.117151563379462e-05, "loss": 0.0, "step": 58226 }, { "epoch": 1.13, "learning_rate": 3.117086867353739e-05, "loss": 0.0, "step": 58228 }, { "epoch": 1.13, "learning_rate": 3.1170221713280156e-05, "loss": 0.0001, "step": 58230 }, { "epoch": 1.13, "learning_rate": 3.1169574753022925e-05, "loss": 0.0, "step": 58232 }, { "epoch": 1.13, "learning_rate": 3.116892779276569e-05, "loss": 0.0, "step": 58234 }, { "epoch": 1.13, "learning_rate": 3.116828083250846e-05, "loss": 0.0, "step": 58236 }, { "epoch": 1.13, "learning_rate": 3.1167633872251225e-05, "loss": 0.0, "step": 58238 }, { "epoch": 1.13, "learning_rate": 3.1166986911993994e-05, "loss": 0.0012, "step": 58240 }, { "epoch": 1.13, "learning_rate": 3.116633995173677e-05, "loss": 0.0, "step": 58242 }, { "epoch": 1.13, "learning_rate": 3.116569299147953e-05, "loss": 0.0, "step": 58244 }, { "epoch": 1.13, "learning_rate": 3.116504603122231e-05, "loss": 0.0, "step": 58246 }, { "epoch": 1.13, "learning_rate": 3.116439907096507e-05, "loss": 0.0, "step": 58248 }, { "epoch": 1.13, "learning_rate": 3.116375211070784e-05, "loss": 0.0014, "step": 58250 }, { "epoch": 1.13, "learning_rate": 3.116310515045061e-05, "loss": 0.0, "step": 58252 }, { "epoch": 1.13, "learning_rate": 3.116245819019338e-05, "loss": 0.0, "step": 58254 }, { "epoch": 1.13, "learning_rate": 3.116181122993615e-05, "loss": 0.0, "step": 58256 }, { "epoch": 1.13, "learning_rate": 3.1161164269678916e-05, "loss": 0.0, "step": 58258 }, { "epoch": 1.13, "learning_rate": 3.1160517309421685e-05, "loss": 0.0066, "step": 58260 }, { "epoch": 1.13, "learning_rate": 3.1159870349164454e-05, "loss": 0.0, "step": 58262 }, { "epoch": 1.13, "learning_rate": 3.1159223388907223e-05, "loss": 0.0, "step": 58264 }, { "epoch": 1.13, "learning_rate": 3.1158576428649986e-05, "loss": 0.0076, "step": 58266 }, { "epoch": 1.13, "learning_rate": 3.115792946839276e-05, "loss": 0.0, "step": 58268 }, { "epoch": 1.13, "learning_rate": 3.1157282508135524e-05, "loss": 0.0023, "step": 58270 }, { "epoch": 1.13, "learning_rate": 3.115663554787829e-05, "loss": 0.0, "step": 58272 }, { "epoch": 1.13, "learning_rate": 3.115598858762107e-05, "loss": 0.0, "step": 58274 }, { "epoch": 1.13, "learning_rate": 3.115534162736383e-05, "loss": 0.0, "step": 58276 }, { "epoch": 1.13, "learning_rate": 3.11546946671066e-05, "loss": 0.0, "step": 58278 }, { "epoch": 1.13, "learning_rate": 3.115404770684937e-05, "loss": 0.0, "step": 58280 }, { "epoch": 1.13, "learning_rate": 3.115340074659214e-05, "loss": 0.0, "step": 58282 }, { "epoch": 1.13, "learning_rate": 3.115275378633491e-05, "loss": 0.0, "step": 58284 }, { "epoch": 1.13, "learning_rate": 3.1152106826077677e-05, "loss": 0.0, "step": 58286 }, { "epoch": 1.13, "learning_rate": 3.1151459865820446e-05, "loss": 0.0, "step": 58288 }, { "epoch": 1.13, "learning_rate": 3.1150812905563215e-05, "loss": 0.0001, "step": 58290 }, { "epoch": 1.13, "learning_rate": 3.1150165945305984e-05, "loss": 0.0, "step": 58292 }, { "epoch": 1.13, "learning_rate": 3.1149518985048746e-05, "loss": 0.0013, "step": 58294 }, { "epoch": 1.13, "learning_rate": 3.114887202479152e-05, "loss": 0.0, "step": 58296 }, { "epoch": 1.13, "learning_rate": 3.1148225064534284e-05, "loss": 0.0, "step": 58298 }, { "epoch": 1.13, "learning_rate": 3.1147578104277053e-05, "loss": 0.0, "step": 58300 }, { "epoch": 1.13, "learning_rate": 3.114693114401982e-05, "loss": 0.0, "step": 58302 }, { "epoch": 1.13, "learning_rate": 3.114628418376259e-05, "loss": 0.0003, "step": 58304 }, { "epoch": 1.13, "learning_rate": 3.114563722350536e-05, "loss": 0.001, "step": 58306 }, { "epoch": 1.13, "learning_rate": 3.114499026324813e-05, "loss": 0.0, "step": 58308 }, { "epoch": 1.13, "learning_rate": 3.11443433029909e-05, "loss": 0.0, "step": 58310 }, { "epoch": 1.13, "learning_rate": 3.114369634273367e-05, "loss": 0.0, "step": 58312 }, { "epoch": 1.13, "learning_rate": 3.114304938247644e-05, "loss": 0.0, "step": 58314 }, { "epoch": 1.13, "learning_rate": 3.11424024222192e-05, "loss": 0.0, "step": 58316 }, { "epoch": 1.13, "learning_rate": 3.1141755461961975e-05, "loss": 0.0, "step": 58318 }, { "epoch": 1.13, "learning_rate": 3.1141108501704744e-05, "loss": 0.0001, "step": 58320 }, { "epoch": 1.13, "learning_rate": 3.1140461541447507e-05, "loss": 0.0002, "step": 58322 }, { "epoch": 1.13, "learning_rate": 3.113981458119028e-05, "loss": 0.0, "step": 58324 }, { "epoch": 1.13, "learning_rate": 3.1139167620933045e-05, "loss": 0.0, "step": 58326 }, { "epoch": 1.13, "learning_rate": 3.113852066067582e-05, "loss": 0.0117, "step": 58328 }, { "epoch": 1.13, "learning_rate": 3.113787370041858e-05, "loss": 0.0, "step": 58330 }, { "epoch": 1.13, "learning_rate": 3.113722674016135e-05, "loss": 0.0, "step": 58332 }, { "epoch": 1.13, "learning_rate": 3.113657977990412e-05, "loss": 0.0, "step": 58334 }, { "epoch": 1.13, "learning_rate": 3.113593281964689e-05, "loss": 0.0, "step": 58336 }, { "epoch": 1.13, "learning_rate": 3.113528585938966e-05, "loss": 0.0006, "step": 58338 }, { "epoch": 1.13, "learning_rate": 3.113463889913243e-05, "loss": 0.0, "step": 58340 }, { "epoch": 1.13, "learning_rate": 3.11339919388752e-05, "loss": 0.0, "step": 58342 }, { "epoch": 1.13, "learning_rate": 3.1133344978617966e-05, "loss": 0.0, "step": 58344 }, { "epoch": 1.13, "learning_rate": 3.1132698018360736e-05, "loss": 0.0021, "step": 58346 }, { "epoch": 1.13, "learning_rate": 3.11320510581035e-05, "loss": 0.0, "step": 58348 }, { "epoch": 1.13, "learning_rate": 3.1131404097846274e-05, "loss": 0.0, "step": 58350 }, { "epoch": 1.13, "learning_rate": 3.113075713758904e-05, "loss": 0.0, "step": 58352 }, { "epoch": 1.13, "learning_rate": 3.1130110177331805e-05, "loss": 0.0, "step": 58354 }, { "epoch": 1.13, "learning_rate": 3.112946321707458e-05, "loss": 0.0, "step": 58356 }, { "epoch": 1.13, "learning_rate": 3.112881625681734e-05, "loss": 0.0, "step": 58358 }, { "epoch": 1.13, "learning_rate": 3.112816929656011e-05, "loss": 0.0, "step": 58360 }, { "epoch": 1.13, "learning_rate": 3.112752233630288e-05, "loss": 0.0, "step": 58362 }, { "epoch": 1.13, "learning_rate": 3.112687537604565e-05, "loss": 0.0172, "step": 58364 }, { "epoch": 1.13, "learning_rate": 3.112622841578842e-05, "loss": 0.0031, "step": 58366 }, { "epoch": 1.13, "learning_rate": 3.112558145553119e-05, "loss": 0.0037, "step": 58368 }, { "epoch": 1.13, "learning_rate": 3.112493449527396e-05, "loss": 0.0, "step": 58370 }, { "epoch": 1.13, "learning_rate": 3.112428753501673e-05, "loss": 0.0, "step": 58372 }, { "epoch": 1.13, "learning_rate": 3.1123640574759496e-05, "loss": 0.0005, "step": 58374 }, { "epoch": 1.13, "learning_rate": 3.112299361450226e-05, "loss": 0.0, "step": 58376 }, { "epoch": 1.13, "learning_rate": 3.1122346654245034e-05, "loss": 0.0, "step": 58378 }, { "epoch": 1.13, "learning_rate": 3.1121699693987796e-05, "loss": 0.0017, "step": 58380 }, { "epoch": 1.13, "learning_rate": 3.1121052733730566e-05, "loss": 0.0, "step": 58382 }, { "epoch": 1.13, "learning_rate": 3.1120405773473335e-05, "loss": 0.0, "step": 58384 }, { "epoch": 1.13, "learning_rate": 3.1119758813216104e-05, "loss": 0.0032, "step": 58386 }, { "epoch": 1.13, "learning_rate": 3.111911185295888e-05, "loss": 0.0, "step": 58388 }, { "epoch": 1.13, "learning_rate": 3.111846489270164e-05, "loss": 0.0, "step": 58390 }, { "epoch": 1.13, "learning_rate": 3.111781793244441e-05, "loss": 0.0, "step": 58392 }, { "epoch": 1.13, "learning_rate": 3.111717097218718e-05, "loss": 0.0, "step": 58394 }, { "epoch": 1.13, "learning_rate": 3.111652401192995e-05, "loss": 0.0, "step": 58396 }, { "epoch": 1.13, "learning_rate": 3.111587705167272e-05, "loss": 0.0, "step": 58398 }, { "epoch": 1.13, "learning_rate": 3.111523009141549e-05, "loss": 0.0, "step": 58400 }, { "epoch": 1.13, "learning_rate": 3.1114583131158256e-05, "loss": 0.0, "step": 58402 }, { "epoch": 1.13, "learning_rate": 3.1113936170901025e-05, "loss": 0.0, "step": 58404 }, { "epoch": 1.13, "learning_rate": 3.1113289210643795e-05, "loss": 0.0, "step": 58406 }, { "epoch": 1.13, "learning_rate": 3.111264225038656e-05, "loss": 0.0, "step": 58408 }, { "epoch": 1.13, "learning_rate": 3.111199529012933e-05, "loss": 0.0002, "step": 58410 }, { "epoch": 1.13, "learning_rate": 3.1111348329872095e-05, "loss": 0.0, "step": 58412 }, { "epoch": 1.13, "learning_rate": 3.1110701369614864e-05, "loss": 0.0, "step": 58414 }, { "epoch": 1.13, "learning_rate": 3.111005440935763e-05, "loss": 0.0, "step": 58416 }, { "epoch": 1.13, "learning_rate": 3.11094074491004e-05, "loss": 0.0, "step": 58418 }, { "epoch": 1.13, "learning_rate": 3.110876048884317e-05, "loss": 0.0, "step": 58420 }, { "epoch": 1.13, "learning_rate": 3.110811352858594e-05, "loss": 0.0, "step": 58422 }, { "epoch": 1.13, "learning_rate": 3.110746656832871e-05, "loss": 0.0, "step": 58424 }, { "epoch": 1.13, "learning_rate": 3.110681960807148e-05, "loss": 0.0001, "step": 58426 }, { "epoch": 1.13, "learning_rate": 3.110617264781425e-05, "loss": 0.0001, "step": 58428 }, { "epoch": 1.13, "learning_rate": 3.110552568755701e-05, "loss": 0.0, "step": 58430 }, { "epoch": 1.13, "learning_rate": 3.1104878727299786e-05, "loss": 0.0, "step": 58432 }, { "epoch": 1.13, "learning_rate": 3.1104231767042555e-05, "loss": 0.0, "step": 58434 }, { "epoch": 1.13, "learning_rate": 3.110358480678532e-05, "loss": 0.0, "step": 58436 }, { "epoch": 1.13, "learning_rate": 3.110293784652809e-05, "loss": 0.0, "step": 58438 }, { "epoch": 1.13, "learning_rate": 3.1102290886270855e-05, "loss": 0.0, "step": 58440 }, { "epoch": 1.13, "learning_rate": 3.1101643926013625e-05, "loss": 0.0, "step": 58442 }, { "epoch": 1.13, "learning_rate": 3.1100996965756394e-05, "loss": 0.0, "step": 58444 }, { "epoch": 1.13, "learning_rate": 3.110035000549916e-05, "loss": 0.0, "step": 58446 }, { "epoch": 1.13, "learning_rate": 3.109970304524193e-05, "loss": 0.0032, "step": 58448 }, { "epoch": 1.13, "learning_rate": 3.10990560849847e-05, "loss": 0.0, "step": 58450 }, { "epoch": 1.13, "learning_rate": 3.109840912472747e-05, "loss": 0.0001, "step": 58452 }, { "epoch": 1.13, "learning_rate": 3.109776216447024e-05, "loss": 0.0002, "step": 58454 }, { "epoch": 1.13, "learning_rate": 3.109711520421301e-05, "loss": 0.0, "step": 58456 }, { "epoch": 1.13, "learning_rate": 3.109646824395577e-05, "loss": 0.0, "step": 58458 }, { "epoch": 1.13, "learning_rate": 3.1095821283698546e-05, "loss": 0.0, "step": 58460 }, { "epoch": 1.13, "learning_rate": 3.109517432344131e-05, "loss": 0.0, "step": 58462 }, { "epoch": 1.13, "learning_rate": 3.109452736318408e-05, "loss": 0.0, "step": 58464 }, { "epoch": 1.13, "learning_rate": 3.1093880402926854e-05, "loss": 0.0009, "step": 58466 }, { "epoch": 1.13, "learning_rate": 3.1093233442669616e-05, "loss": 0.0, "step": 58468 }, { "epoch": 1.13, "learning_rate": 3.109258648241239e-05, "loss": 0.0, "step": 58470 }, { "epoch": 1.13, "learning_rate": 3.1091939522155154e-05, "loss": 0.0, "step": 58472 }, { "epoch": 1.13, "learning_rate": 3.109129256189792e-05, "loss": 0.0008, "step": 58474 }, { "epoch": 1.13, "learning_rate": 3.109064560164069e-05, "loss": 0.0, "step": 58476 }, { "epoch": 1.13, "learning_rate": 3.108999864138346e-05, "loss": 0.0002, "step": 58478 }, { "epoch": 1.14, "learning_rate": 3.108935168112623e-05, "loss": 0.0, "step": 58480 }, { "epoch": 1.14, "learning_rate": 3.1088704720869e-05, "loss": 0.0003, "step": 58482 }, { "epoch": 1.14, "learning_rate": 3.108805776061177e-05, "loss": 0.0, "step": 58484 }, { "epoch": 1.14, "learning_rate": 3.108741080035454e-05, "loss": 0.0001, "step": 58486 }, { "epoch": 1.14, "learning_rate": 3.108676384009731e-05, "loss": 0.0008, "step": 58488 }, { "epoch": 1.14, "learning_rate": 3.108611687984007e-05, "loss": 0.0002, "step": 58490 }, { "epoch": 1.14, "learning_rate": 3.1085469919582845e-05, "loss": 0.0, "step": 58492 }, { "epoch": 1.14, "learning_rate": 3.108482295932561e-05, "loss": 0.0001, "step": 58494 }, { "epoch": 1.14, "learning_rate": 3.1084175999068376e-05, "loss": 0.0, "step": 58496 }, { "epoch": 1.14, "learning_rate": 3.108352903881115e-05, "loss": 0.001, "step": 58498 }, { "epoch": 1.14, "learning_rate": 3.1082882078553914e-05, "loss": 0.0, "step": 58500 }, { "epoch": 1.14, "learning_rate": 3.1082235118296683e-05, "loss": 0.0001, "step": 58502 }, { "epoch": 1.14, "learning_rate": 3.108158815803945e-05, "loss": 0.0067, "step": 58504 }, { "epoch": 1.14, "learning_rate": 3.108094119778222e-05, "loss": 0.0029, "step": 58506 }, { "epoch": 1.14, "learning_rate": 3.108029423752499e-05, "loss": 0.0, "step": 58508 }, { "epoch": 1.14, "learning_rate": 3.107964727726776e-05, "loss": 0.0, "step": 58510 }, { "epoch": 1.14, "learning_rate": 3.107900031701053e-05, "loss": 0.0, "step": 58512 }, { "epoch": 1.14, "learning_rate": 3.10783533567533e-05, "loss": 0.0, "step": 58514 }, { "epoch": 1.14, "learning_rate": 3.107770639649607e-05, "loss": 0.0, "step": 58516 }, { "epoch": 1.14, "learning_rate": 3.107705943623883e-05, "loss": 0.0003, "step": 58518 }, { "epoch": 1.14, "learning_rate": 3.1076412475981605e-05, "loss": 0.0036, "step": 58520 }, { "epoch": 1.14, "learning_rate": 3.107576551572437e-05, "loss": 0.0, "step": 58522 }, { "epoch": 1.14, "learning_rate": 3.107511855546714e-05, "loss": 0.0, "step": 58524 }, { "epoch": 1.14, "learning_rate": 3.1074471595209906e-05, "loss": 0.0, "step": 58526 }, { "epoch": 1.14, "learning_rate": 3.1073824634952675e-05, "loss": 0.0, "step": 58528 }, { "epoch": 1.14, "learning_rate": 3.1073177674695444e-05, "loss": 0.0, "step": 58530 }, { "epoch": 1.14, "learning_rate": 3.107253071443821e-05, "loss": 0.0001, "step": 58532 }, { "epoch": 1.14, "learning_rate": 3.107188375418098e-05, "loss": 0.0, "step": 58534 }, { "epoch": 1.14, "learning_rate": 3.107123679392375e-05, "loss": 0.0, "step": 58536 }, { "epoch": 1.14, "learning_rate": 3.107058983366652e-05, "loss": 0.0, "step": 58538 }, { "epoch": 1.14, "learning_rate": 3.106994287340928e-05, "loss": 0.0219, "step": 58540 }, { "epoch": 1.14, "learning_rate": 3.106929591315206e-05, "loss": 0.0, "step": 58542 }, { "epoch": 1.14, "learning_rate": 3.106864895289483e-05, "loss": 0.0, "step": 58544 }, { "epoch": 1.14, "learning_rate": 3.1068001992637597e-05, "loss": 0.0, "step": 58546 }, { "epoch": 1.14, "learning_rate": 3.1067355032380366e-05, "loss": 0.0, "step": 58548 }, { "epoch": 1.14, "learning_rate": 3.106670807212313e-05, "loss": 0.0, "step": 58550 }, { "epoch": 1.14, "learning_rate": 3.1066061111865904e-05, "loss": 0.0, "step": 58552 }, { "epoch": 1.14, "learning_rate": 3.1065414151608666e-05, "loss": 0.0, "step": 58554 }, { "epoch": 1.14, "learning_rate": 3.106509067148005e-05, "loss": 0.0366, "step": 58556 }, { "epoch": 1.14, "learning_rate": 3.1064443711222827e-05, "loss": 0.0, "step": 58558 }, { "epoch": 1.14, "learning_rate": 3.106379675096559e-05, "loss": 0.0001, "step": 58560 }, { "epoch": 1.14, "learning_rate": 3.106314979070836e-05, "loss": 0.0025, "step": 58562 }, { "epoch": 1.14, "learning_rate": 3.106250283045113e-05, "loss": 0.0002, "step": 58564 }, { "epoch": 1.14, "learning_rate": 3.1061855870193896e-05, "loss": 0.0, "step": 58566 }, { "epoch": 1.14, "learning_rate": 3.1061208909936665e-05, "loss": 0.0, "step": 58568 }, { "epoch": 1.14, "learning_rate": 3.1060561949679434e-05, "loss": 0.0, "step": 58570 }, { "epoch": 1.14, "learning_rate": 3.10599149894222e-05, "loss": 0.0, "step": 58572 }, { "epoch": 1.14, "learning_rate": 3.1059268029164966e-05, "loss": 0.0, "step": 58574 }, { "epoch": 1.14, "learning_rate": 3.105862106890774e-05, "loss": 0.0, "step": 58576 }, { "epoch": 1.14, "learning_rate": 3.1057974108650504e-05, "loss": 0.0, "step": 58578 }, { "epoch": 1.14, "learning_rate": 3.105732714839328e-05, "loss": 0.0, "step": 58580 }, { "epoch": 1.14, "learning_rate": 3.105668018813604e-05, "loss": 0.0, "step": 58582 }, { "epoch": 1.14, "learning_rate": 3.105603322787881e-05, "loss": 0.0, "step": 58584 }, { "epoch": 1.14, "learning_rate": 3.105538626762158e-05, "loss": 0.0, "step": 58586 }, { "epoch": 1.14, "learning_rate": 3.105473930736435e-05, "loss": 0.0, "step": 58588 }, { "epoch": 1.14, "learning_rate": 3.105409234710712e-05, "loss": 0.0, "step": 58590 }, { "epoch": 1.14, "learning_rate": 3.105344538684989e-05, "loss": 0.0003, "step": 58592 }, { "epoch": 1.14, "learning_rate": 3.1052798426592656e-05, "loss": 0.0443, "step": 58594 }, { "epoch": 1.14, "learning_rate": 3.105215146633542e-05, "loss": 0.0, "step": 58596 }, { "epoch": 1.14, "learning_rate": 3.1051504506078195e-05, "loss": 0.0353, "step": 58598 }, { "epoch": 1.14, "learning_rate": 3.1050857545820964e-05, "loss": 0.0, "step": 58600 }, { "epoch": 1.14, "learning_rate": 3.105021058556373e-05, "loss": 0.0, "step": 58602 }, { "epoch": 1.14, "learning_rate": 3.10495636253065e-05, "loss": 0.0, "step": 58604 }, { "epoch": 1.14, "learning_rate": 3.1048916665049264e-05, "loss": 0.0, "step": 58606 }, { "epoch": 1.14, "learning_rate": 3.104826970479204e-05, "loss": 0.008, "step": 58608 }, { "epoch": 1.14, "learning_rate": 3.10476227445348e-05, "loss": 0.0, "step": 58610 }, { "epoch": 1.14, "learning_rate": 3.104697578427757e-05, "loss": 0.0003, "step": 58612 }, { "epoch": 1.14, "learning_rate": 3.104632882402034e-05, "loss": 0.0, "step": 58614 }, { "epoch": 1.14, "learning_rate": 3.104568186376311e-05, "loss": 0.0, "step": 58616 }, { "epoch": 1.14, "learning_rate": 3.104503490350588e-05, "loss": 0.0, "step": 58618 }, { "epoch": 1.14, "learning_rate": 3.104438794324865e-05, "loss": 0.0, "step": 58620 }, { "epoch": 1.14, "learning_rate": 3.104374098299142e-05, "loss": 0.0, "step": 58622 }, { "epoch": 1.14, "learning_rate": 3.1043094022734186e-05, "loss": 0.0, "step": 58624 }, { "epoch": 1.14, "learning_rate": 3.1042447062476955e-05, "loss": 0.0, "step": 58626 }, { "epoch": 1.14, "learning_rate": 3.104180010221972e-05, "loss": 0.0, "step": 58628 }, { "epoch": 1.14, "learning_rate": 3.104115314196249e-05, "loss": 0.0, "step": 58630 }, { "epoch": 1.14, "learning_rate": 3.1040506181705256e-05, "loss": 0.0015, "step": 58632 }, { "epoch": 1.14, "learning_rate": 3.1039859221448025e-05, "loss": 0.0, "step": 58634 }, { "epoch": 1.14, "learning_rate": 3.10392122611908e-05, "loss": 0.0, "step": 58636 }, { "epoch": 1.14, "learning_rate": 3.103856530093356e-05, "loss": 0.0001, "step": 58638 }, { "epoch": 1.14, "learning_rate": 3.103791834067634e-05, "loss": 0.0, "step": 58640 }, { "epoch": 1.14, "learning_rate": 3.10372713804191e-05, "loss": 0.0, "step": 58642 }, { "epoch": 1.14, "learning_rate": 3.103662442016187e-05, "loss": 0.0031, "step": 58644 }, { "epoch": 1.14, "learning_rate": 3.103597745990464e-05, "loss": 0.0, "step": 58646 }, { "epoch": 1.14, "learning_rate": 3.103533049964741e-05, "loss": 0.0, "step": 58648 }, { "epoch": 1.14, "learning_rate": 3.103468353939018e-05, "loss": 0.0, "step": 58650 }, { "epoch": 1.14, "learning_rate": 3.1034036579132946e-05, "loss": 0.0, "step": 58652 }, { "epoch": 1.14, "learning_rate": 3.1033389618875715e-05, "loss": 0.0, "step": 58654 }, { "epoch": 1.14, "learning_rate": 3.103274265861848e-05, "loss": 0.0, "step": 58656 }, { "epoch": 1.14, "learning_rate": 3.1032095698361254e-05, "loss": 0.0, "step": 58658 }, { "epoch": 1.14, "learning_rate": 3.1031448738104016e-05, "loss": 0.0, "step": 58660 }, { "epoch": 1.14, "learning_rate": 3.103080177784679e-05, "loss": 0.0, "step": 58662 }, { "epoch": 1.14, "learning_rate": 3.1030154817589554e-05, "loss": 0.0002, "step": 58664 }, { "epoch": 1.14, "learning_rate": 3.102950785733232e-05, "loss": 0.0, "step": 58666 }, { "epoch": 1.14, "learning_rate": 3.10288608970751e-05, "loss": 0.0, "step": 58668 }, { "epoch": 1.14, "learning_rate": 3.102821393681786e-05, "loss": 0.0, "step": 58670 }, { "epoch": 1.14, "learning_rate": 3.102756697656063e-05, "loss": 0.0029, "step": 58672 }, { "epoch": 1.14, "learning_rate": 3.10269200163034e-05, "loss": 0.0096, "step": 58674 }, { "epoch": 1.14, "learning_rate": 3.102627305604617e-05, "loss": 0.0005, "step": 58676 }, { "epoch": 1.14, "learning_rate": 3.102562609578894e-05, "loss": 0.0, "step": 58678 }, { "epoch": 1.14, "learning_rate": 3.102497913553171e-05, "loss": 0.0001, "step": 58680 }, { "epoch": 1.14, "learning_rate": 3.1024332175274476e-05, "loss": 0.0, "step": 58682 }, { "epoch": 1.14, "learning_rate": 3.1023685215017245e-05, "loss": 0.0, "step": 58684 }, { "epoch": 1.14, "learning_rate": 3.1023038254760014e-05, "loss": 0.0, "step": 58686 }, { "epoch": 1.14, "learning_rate": 3.1022391294502776e-05, "loss": 0.0001, "step": 58688 }, { "epoch": 1.14, "learning_rate": 3.102174433424555e-05, "loss": 0.0003, "step": 58690 }, { "epoch": 1.14, "learning_rate": 3.1021097373988315e-05, "loss": 0.002, "step": 58692 }, { "epoch": 1.14, "learning_rate": 3.1020450413731084e-05, "loss": 0.0, "step": 58694 }, { "epoch": 1.14, "learning_rate": 3.101980345347385e-05, "loss": 0.0001, "step": 58696 }, { "epoch": 1.14, "learning_rate": 3.101915649321662e-05, "loss": 0.0, "step": 58698 }, { "epoch": 1.14, "learning_rate": 3.10185095329594e-05, "loss": 0.0, "step": 58700 }, { "epoch": 1.14, "learning_rate": 3.101786257270216e-05, "loss": 0.0004, "step": 58702 }, { "epoch": 1.14, "learning_rate": 3.101721561244493e-05, "loss": 0.0, "step": 58704 }, { "epoch": 1.14, "learning_rate": 3.10165686521877e-05, "loss": 0.0, "step": 58706 }, { "epoch": 1.14, "learning_rate": 3.101592169193047e-05, "loss": 0.0, "step": 58708 }, { "epoch": 1.14, "learning_rate": 3.101527473167323e-05, "loss": 0.0, "step": 58710 }, { "epoch": 1.14, "learning_rate": 3.1014627771416005e-05, "loss": 0.0004, "step": 58712 }, { "epoch": 1.14, "learning_rate": 3.1013980811158774e-05, "loss": 0.0, "step": 58714 }, { "epoch": 1.14, "learning_rate": 3.101333385090154e-05, "loss": 0.0, "step": 58716 }, { "epoch": 1.14, "learning_rate": 3.101268689064431e-05, "loss": 0.0, "step": 58718 }, { "epoch": 1.14, "learning_rate": 3.1012039930387075e-05, "loss": 0.0, "step": 58720 }, { "epoch": 1.14, "learning_rate": 3.101139297012985e-05, "loss": 0.0031, "step": 58722 }, { "epoch": 1.14, "learning_rate": 3.101074600987261e-05, "loss": 0.0001, "step": 58724 }, { "epoch": 1.14, "learning_rate": 3.101009904961538e-05, "loss": 0.0001, "step": 58726 }, { "epoch": 1.14, "learning_rate": 3.100945208935815e-05, "loss": 0.0001, "step": 58728 }, { "epoch": 1.14, "learning_rate": 3.100880512910092e-05, "loss": 0.0, "step": 58730 }, { "epoch": 1.14, "learning_rate": 3.100815816884369e-05, "loss": 0.0013, "step": 58732 }, { "epoch": 1.14, "learning_rate": 3.100751120858646e-05, "loss": 0.0022, "step": 58734 }, { "epoch": 1.14, "learning_rate": 3.100686424832923e-05, "loss": 0.0002, "step": 58736 }, { "epoch": 1.14, "learning_rate": 3.100621728807199e-05, "loss": 0.0, "step": 58738 }, { "epoch": 1.14, "learning_rate": 3.1005570327814766e-05, "loss": 0.0, "step": 58740 }, { "epoch": 1.14, "learning_rate": 3.100492336755753e-05, "loss": 0.0, "step": 58742 }, { "epoch": 1.14, "learning_rate": 3.1004276407300304e-05, "loss": 0.0021, "step": 58744 }, { "epoch": 1.14, "learning_rate": 3.100362944704307e-05, "loss": 0.0024, "step": 58746 }, { "epoch": 1.14, "learning_rate": 3.1002982486785835e-05, "loss": 0.0001, "step": 58748 }, { "epoch": 1.14, "learning_rate": 3.100233552652861e-05, "loss": 0.0, "step": 58750 }, { "epoch": 1.14, "learning_rate": 3.1001688566271373e-05, "loss": 0.0108, "step": 58752 }, { "epoch": 1.14, "learning_rate": 3.100104160601414e-05, "loss": 0.0, "step": 58754 }, { "epoch": 1.14, "learning_rate": 3.100039464575691e-05, "loss": 0.0, "step": 58756 }, { "epoch": 1.14, "learning_rate": 3.099974768549968e-05, "loss": 0.0, "step": 58758 }, { "epoch": 1.14, "learning_rate": 3.099910072524245e-05, "loss": 0.0, "step": 58760 }, { "epoch": 1.14, "learning_rate": 3.099845376498522e-05, "loss": 0.0, "step": 58762 }, { "epoch": 1.14, "learning_rate": 3.099780680472799e-05, "loss": 0.0, "step": 58764 }, { "epoch": 1.14, "learning_rate": 3.099715984447076e-05, "loss": 0.0, "step": 58766 }, { "epoch": 1.14, "learning_rate": 3.0996512884213526e-05, "loss": 0.0, "step": 58768 }, { "epoch": 1.14, "learning_rate": 3.099586592395629e-05, "loss": 0.0, "step": 58770 }, { "epoch": 1.14, "learning_rate": 3.0995218963699064e-05, "loss": 0.0355, "step": 58772 }, { "epoch": 1.14, "learning_rate": 3.099457200344183e-05, "loss": 0.0, "step": 58774 }, { "epoch": 1.14, "learning_rate": 3.0993925043184596e-05, "loss": 0.0, "step": 58776 }, { "epoch": 1.14, "learning_rate": 3.099327808292737e-05, "loss": 0.0, "step": 58778 }, { "epoch": 1.14, "learning_rate": 3.0992631122670134e-05, "loss": 0.0, "step": 58780 }, { "epoch": 1.14, "learning_rate": 3.099198416241291e-05, "loss": 0.0, "step": 58782 }, { "epoch": 1.14, "learning_rate": 3.099133720215567e-05, "loss": 0.0, "step": 58784 }, { "epoch": 1.14, "learning_rate": 3.099069024189844e-05, "loss": 0.0, "step": 58786 }, { "epoch": 1.14, "learning_rate": 3.099004328164121e-05, "loss": 0.0, "step": 58788 }, { "epoch": 1.14, "learning_rate": 3.098939632138398e-05, "loss": 0.0001, "step": 58790 }, { "epoch": 1.14, "learning_rate": 3.098874936112675e-05, "loss": 0.0, "step": 58792 }, { "epoch": 1.14, "learning_rate": 3.098810240086952e-05, "loss": 0.0, "step": 58794 }, { "epoch": 1.14, "learning_rate": 3.0987455440612287e-05, "loss": 0.0, "step": 58796 }, { "epoch": 1.14, "learning_rate": 3.098680848035505e-05, "loss": 0.0, "step": 58798 }, { "epoch": 1.14, "learning_rate": 3.0986161520097825e-05, "loss": 0.0022, "step": 58800 }, { "epoch": 1.14, "learning_rate": 3.098551455984059e-05, "loss": 0.0, "step": 58802 }, { "epoch": 1.14, "learning_rate": 3.098486759958336e-05, "loss": 0.0, "step": 58804 }, { "epoch": 1.14, "learning_rate": 3.0984220639326125e-05, "loss": 0.0, "step": 58806 }, { "epoch": 1.14, "learning_rate": 3.0983573679068894e-05, "loss": 0.0, "step": 58808 }, { "epoch": 1.14, "learning_rate": 3.0982926718811663e-05, "loss": 0.0, "step": 58810 }, { "epoch": 1.14, "learning_rate": 3.098227975855443e-05, "loss": 0.0, "step": 58812 }, { "epoch": 1.14, "learning_rate": 3.09816327982972e-05, "loss": 0.0014, "step": 58814 }, { "epoch": 1.14, "learning_rate": 3.098098583803997e-05, "loss": 0.0, "step": 58816 }, { "epoch": 1.14, "learning_rate": 3.098033887778274e-05, "loss": 0.0, "step": 58818 }, { "epoch": 1.14, "learning_rate": 3.09796919175255e-05, "loss": 0.0, "step": 58820 }, { "epoch": 1.14, "learning_rate": 3.097904495726828e-05, "loss": 0.0016, "step": 58822 }, { "epoch": 1.14, "learning_rate": 3.097839799701105e-05, "loss": 0.0023, "step": 58824 }, { "epoch": 1.14, "learning_rate": 3.0977751036753816e-05, "loss": 0.006, "step": 58826 }, { "epoch": 1.14, "learning_rate": 3.0977104076496585e-05, "loss": 0.0, "step": 58828 }, { "epoch": 1.14, "learning_rate": 3.097645711623935e-05, "loss": 0.0065, "step": 58830 }, { "epoch": 1.14, "learning_rate": 3.097581015598212e-05, "loss": 0.0055, "step": 58832 }, { "epoch": 1.14, "learning_rate": 3.0975163195724886e-05, "loss": 0.0001, "step": 58834 }, { "epoch": 1.14, "learning_rate": 3.0974516235467655e-05, "loss": 0.0, "step": 58836 }, { "epoch": 1.14, "learning_rate": 3.0973869275210424e-05, "loss": 0.0, "step": 58838 }, { "epoch": 1.14, "learning_rate": 3.097322231495319e-05, "loss": 0.0052, "step": 58840 }, { "epoch": 1.14, "learning_rate": 3.097257535469596e-05, "loss": 0.0, "step": 58842 }, { "epoch": 1.14, "learning_rate": 3.097192839443873e-05, "loss": 0.0052, "step": 58844 }, { "epoch": 1.14, "learning_rate": 3.09712814341815e-05, "loss": 0.0, "step": 58846 }, { "epoch": 1.14, "learning_rate": 3.097063447392427e-05, "loss": 0.0, "step": 58848 }, { "epoch": 1.14, "learning_rate": 3.096998751366704e-05, "loss": 0.0003, "step": 58850 }, { "epoch": 1.14, "learning_rate": 3.09693405534098e-05, "loss": 0.0, "step": 58852 }, { "epoch": 1.14, "learning_rate": 3.0968693593152576e-05, "loss": 0.0, "step": 58854 }, { "epoch": 1.14, "learning_rate": 3.0968046632895346e-05, "loss": 0.0024, "step": 58856 }, { "epoch": 1.14, "learning_rate": 3.096739967263811e-05, "loss": 0.0, "step": 58858 }, { "epoch": 1.14, "learning_rate": 3.0966752712380884e-05, "loss": 0.0, "step": 58860 }, { "epoch": 1.14, "learning_rate": 3.0966105752123646e-05, "loss": 0.0, "step": 58862 }, { "epoch": 1.14, "learning_rate": 3.096545879186642e-05, "loss": 0.0, "step": 58864 }, { "epoch": 1.14, "learning_rate": 3.0964811831609184e-05, "loss": 0.0002, "step": 58866 }, { "epoch": 1.14, "learning_rate": 3.096416487135195e-05, "loss": 0.0, "step": 58868 }, { "epoch": 1.14, "learning_rate": 3.096351791109472e-05, "loss": 0.0001, "step": 58870 }, { "epoch": 1.14, "learning_rate": 3.096287095083749e-05, "loss": 0.0, "step": 58872 }, { "epoch": 1.14, "learning_rate": 3.096222399058026e-05, "loss": 0.0, "step": 58874 }, { "epoch": 1.14, "learning_rate": 3.096157703032303e-05, "loss": 0.0, "step": 58876 }, { "epoch": 1.14, "learning_rate": 3.09609300700658e-05, "loss": 0.0, "step": 58878 }, { "epoch": 1.14, "learning_rate": 3.096028310980856e-05, "loss": 0.0001, "step": 58880 }, { "epoch": 1.14, "learning_rate": 3.095963614955134e-05, "loss": 0.0, "step": 58882 }, { "epoch": 1.14, "learning_rate": 3.09589891892941e-05, "loss": 0.0017, "step": 58884 }, { "epoch": 1.14, "learning_rate": 3.0958342229036875e-05, "loss": 0.0, "step": 58886 }, { "epoch": 1.14, "learning_rate": 3.095769526877964e-05, "loss": 0.0001, "step": 58888 }, { "epoch": 1.14, "learning_rate": 3.0957048308522406e-05, "loss": 0.0, "step": 58890 }, { "epoch": 1.14, "learning_rate": 3.095640134826518e-05, "loss": 0.0, "step": 58892 }, { "epoch": 1.14, "learning_rate": 3.0955754388007945e-05, "loss": 0.0, "step": 58894 }, { "epoch": 1.14, "learning_rate": 3.0955107427750714e-05, "loss": 0.0, "step": 58896 }, { "epoch": 1.14, "learning_rate": 3.095446046749348e-05, "loss": 0.0, "step": 58898 }, { "epoch": 1.14, "learning_rate": 3.095381350723625e-05, "loss": 0.0, "step": 58900 }, { "epoch": 1.14, "learning_rate": 3.095316654697902e-05, "loss": 0.0007, "step": 58902 }, { "epoch": 1.14, "learning_rate": 3.095251958672179e-05, "loss": 0.0, "step": 58904 }, { "epoch": 1.14, "learning_rate": 3.095187262646456e-05, "loss": 0.0003, "step": 58906 }, { "epoch": 1.14, "learning_rate": 3.095122566620733e-05, "loss": 0.0002, "step": 58908 }, { "epoch": 1.14, "learning_rate": 3.09505787059501e-05, "loss": 0.0, "step": 58910 }, { "epoch": 1.14, "learning_rate": 3.094993174569286e-05, "loss": 0.0, "step": 58912 }, { "epoch": 1.14, "learning_rate": 3.0949284785435635e-05, "loss": 0.0, "step": 58914 }, { "epoch": 1.14, "learning_rate": 3.09486378251784e-05, "loss": 0.0, "step": 58916 }, { "epoch": 1.14, "learning_rate": 3.094799086492117e-05, "loss": 0.0, "step": 58918 }, { "epoch": 1.14, "learning_rate": 3.0947343904663936e-05, "loss": 0.007, "step": 58920 }, { "epoch": 1.14, "learning_rate": 3.0946696944406705e-05, "loss": 0.002, "step": 58922 }, { "epoch": 1.14, "learning_rate": 3.094604998414948e-05, "loss": 0.0081, "step": 58924 }, { "epoch": 1.14, "learning_rate": 3.094540302389224e-05, "loss": 0.0002, "step": 58926 }, { "epoch": 1.14, "learning_rate": 3.094475606363501e-05, "loss": 0.0, "step": 58928 }, { "epoch": 1.14, "learning_rate": 3.094410910337778e-05, "loss": 0.0, "step": 58930 }, { "epoch": 1.14, "learning_rate": 3.094346214312055e-05, "loss": 0.0003, "step": 58932 }, { "epoch": 1.14, "learning_rate": 3.094281518286332e-05, "loss": 0.0, "step": 58934 }, { "epoch": 1.14, "learning_rate": 3.094216822260609e-05, "loss": 0.0, "step": 58936 }, { "epoch": 1.14, "learning_rate": 3.094152126234886e-05, "loss": 0.0, "step": 58938 }, { "epoch": 1.14, "learning_rate": 3.094087430209162e-05, "loss": 0.0, "step": 58940 }, { "epoch": 1.14, "learning_rate": 3.0940227341834396e-05, "loss": 0.0032, "step": 58942 }, { "epoch": 1.14, "learning_rate": 3.093958038157716e-05, "loss": 0.0, "step": 58944 }, { "epoch": 1.14, "learning_rate": 3.0938933421319934e-05, "loss": 0.0, "step": 58946 }, { "epoch": 1.14, "learning_rate": 3.0938286461062696e-05, "loss": 0.0001, "step": 58948 }, { "epoch": 1.14, "learning_rate": 3.0937639500805465e-05, "loss": 0.0, "step": 58950 }, { "epoch": 1.14, "learning_rate": 3.0936992540548235e-05, "loss": 0.0, "step": 58952 }, { "epoch": 1.14, "learning_rate": 3.0936345580291004e-05, "loss": 0.0, "step": 58954 }, { "epoch": 1.14, "learning_rate": 3.093569862003377e-05, "loss": 0.0, "step": 58956 }, { "epoch": 1.14, "learning_rate": 3.093505165977654e-05, "loss": 0.0, "step": 58958 }, { "epoch": 1.14, "learning_rate": 3.093440469951931e-05, "loss": 0.0001, "step": 58960 }, { "epoch": 1.14, "learning_rate": 3.093375773926207e-05, "loss": 0.0, "step": 58962 }, { "epoch": 1.14, "learning_rate": 3.093311077900485e-05, "loss": 0.0, "step": 58964 }, { "epoch": 1.14, "learning_rate": 3.093246381874761e-05, "loss": 0.0, "step": 58966 }, { "epoch": 1.14, "learning_rate": 3.093181685849039e-05, "loss": 0.0, "step": 58968 }, { "epoch": 1.14, "learning_rate": 3.0931169898233156e-05, "loss": 0.0, "step": 58970 }, { "epoch": 1.14, "learning_rate": 3.093052293797592e-05, "loss": 0.0009, "step": 58972 }, { "epoch": 1.14, "learning_rate": 3.0929875977718694e-05, "loss": 0.0, "step": 58974 }, { "epoch": 1.14, "learning_rate": 3.092922901746146e-05, "loss": 0.031, "step": 58976 }, { "epoch": 1.14, "learning_rate": 3.0928582057204226e-05, "loss": 0.0, "step": 58978 }, { "epoch": 1.14, "learning_rate": 3.0927935096946995e-05, "loss": 0.0, "step": 58980 }, { "epoch": 1.14, "learning_rate": 3.0927288136689764e-05, "loss": 0.0, "step": 58982 }, { "epoch": 1.14, "learning_rate": 3.092664117643253e-05, "loss": 0.0001, "step": 58984 }, { "epoch": 1.14, "learning_rate": 3.09259942161753e-05, "loss": 0.0, "step": 58986 }, { "epoch": 1.14, "learning_rate": 3.092534725591807e-05, "loss": 0.0, "step": 58988 }, { "epoch": 1.14, "learning_rate": 3.092470029566084e-05, "loss": 0.0, "step": 58990 }, { "epoch": 1.14, "learning_rate": 3.092405333540361e-05, "loss": 0.0, "step": 58992 }, { "epoch": 1.15, "learning_rate": 3.092340637514637e-05, "loss": 0.0, "step": 58994 }, { "epoch": 1.15, "learning_rate": 3.092275941488915e-05, "loss": 0.0, "step": 58996 }, { "epoch": 1.15, "learning_rate": 3.092211245463191e-05, "loss": 0.0, "step": 58998 }, { "epoch": 1.15, "learning_rate": 3.092146549437468e-05, "loss": 0.0001, "step": 59000 }, { "epoch": 1.15, "learning_rate": 3.0920818534117455e-05, "loss": 0.0001, "step": 59002 }, { "epoch": 1.15, "learning_rate": 3.092017157386022e-05, "loss": 0.0, "step": 59004 }, { "epoch": 1.15, "learning_rate": 3.091952461360299e-05, "loss": 0.0, "step": 59006 }, { "epoch": 1.15, "learning_rate": 3.0918877653345755e-05, "loss": 0.0008, "step": 59008 }, { "epoch": 1.15, "learning_rate": 3.0918230693088524e-05, "loss": 0.0, "step": 59010 }, { "epoch": 1.15, "learning_rate": 3.0917583732831293e-05, "loss": 0.0, "step": 59012 }, { "epoch": 1.15, "learning_rate": 3.091693677257406e-05, "loss": 0.0005, "step": 59014 }, { "epoch": 1.15, "learning_rate": 3.091628981231683e-05, "loss": 0.0053, "step": 59016 }, { "epoch": 1.15, "learning_rate": 3.09156428520596e-05, "loss": 0.0, "step": 59018 }, { "epoch": 1.15, "learning_rate": 3.091499589180237e-05, "loss": 0.0001, "step": 59020 }, { "epoch": 1.15, "learning_rate": 3.091434893154513e-05, "loss": 0.0, "step": 59022 }, { "epoch": 1.15, "learning_rate": 3.091370197128791e-05, "loss": 0.0, "step": 59024 }, { "epoch": 1.15, "learning_rate": 3.091305501103067e-05, "loss": 0.0009, "step": 59026 }, { "epoch": 1.15, "learning_rate": 3.0912408050773446e-05, "loss": 0.0011, "step": 59028 }, { "epoch": 1.15, "learning_rate": 3.091176109051621e-05, "loss": 0.0002, "step": 59030 }, { "epoch": 1.15, "learning_rate": 3.091111413025898e-05, "loss": 0.0, "step": 59032 }, { "epoch": 1.15, "learning_rate": 3.0910467170001753e-05, "loss": 0.0, "step": 59034 }, { "epoch": 1.15, "learning_rate": 3.0909820209744516e-05, "loss": 0.0, "step": 59036 }, { "epoch": 1.15, "learning_rate": 3.0909173249487285e-05, "loss": 0.0, "step": 59038 }, { "epoch": 1.15, "learning_rate": 3.0908526289230054e-05, "loss": 0.0, "step": 59040 }, { "epoch": 1.15, "learning_rate": 3.090787932897282e-05, "loss": 0.0002, "step": 59042 }, { "epoch": 1.15, "learning_rate": 3.090723236871559e-05, "loss": 0.0001, "step": 59044 }, { "epoch": 1.15, "learning_rate": 3.090658540845836e-05, "loss": 0.0001, "step": 59046 }, { "epoch": 1.15, "learning_rate": 3.090593844820113e-05, "loss": 0.0, "step": 59048 }, { "epoch": 1.15, "learning_rate": 3.09052914879439e-05, "loss": 0.0017, "step": 59050 }, { "epoch": 1.15, "learning_rate": 3.090464452768667e-05, "loss": 0.0, "step": 59052 }, { "epoch": 1.15, "learning_rate": 3.090399756742943e-05, "loss": 0.0011, "step": 59054 }, { "epoch": 1.15, "learning_rate": 3.0903350607172207e-05, "loss": 0.0, "step": 59056 }, { "epoch": 1.15, "learning_rate": 3.090270364691497e-05, "loss": 0.0, "step": 59058 }, { "epoch": 1.15, "learning_rate": 3.090205668665774e-05, "loss": 0.013, "step": 59060 }, { "epoch": 1.15, "learning_rate": 3.090140972640051e-05, "loss": 0.0, "step": 59062 }, { "epoch": 1.15, "learning_rate": 3.0900762766143276e-05, "loss": 0.002, "step": 59064 }, { "epoch": 1.15, "learning_rate": 3.0900115805886045e-05, "loss": 0.0, "step": 59066 }, { "epoch": 1.15, "learning_rate": 3.0899468845628814e-05, "loss": 0.0, "step": 59068 }, { "epoch": 1.15, "learning_rate": 3.089882188537158e-05, "loss": 0.0, "step": 59070 }, { "epoch": 1.15, "learning_rate": 3.089817492511435e-05, "loss": 0.0, "step": 59072 }, { "epoch": 1.15, "learning_rate": 3.089752796485712e-05, "loss": 0.0, "step": 59074 }, { "epoch": 1.15, "learning_rate": 3.0896881004599884e-05, "loss": 0.0, "step": 59076 }, { "epoch": 1.15, "learning_rate": 3.089623404434266e-05, "loss": 0.0, "step": 59078 }, { "epoch": 1.15, "learning_rate": 3.089558708408543e-05, "loss": 0.0, "step": 59080 }, { "epoch": 1.15, "learning_rate": 3.089494012382819e-05, "loss": 0.0, "step": 59082 }, { "epoch": 1.15, "learning_rate": 3.089429316357097e-05, "loss": 0.0, "step": 59084 }, { "epoch": 1.15, "learning_rate": 3.089364620331373e-05, "loss": 0.0, "step": 59086 }, { "epoch": 1.15, "learning_rate": 3.0892999243056505e-05, "loss": 0.0, "step": 59088 }, { "epoch": 1.15, "learning_rate": 3.089235228279927e-05, "loss": 0.0, "step": 59090 }, { "epoch": 1.15, "learning_rate": 3.0891705322542037e-05, "loss": 0.0001, "step": 59092 }, { "epoch": 1.15, "learning_rate": 3.0891058362284806e-05, "loss": 0.0, "step": 59094 }, { "epoch": 1.15, "learning_rate": 3.0890411402027575e-05, "loss": 0.0, "step": 59096 }, { "epoch": 1.15, "learning_rate": 3.0889764441770344e-05, "loss": 0.0011, "step": 59098 }, { "epoch": 1.15, "learning_rate": 3.088911748151311e-05, "loss": 0.0, "step": 59100 }, { "epoch": 1.15, "learning_rate": 3.088847052125588e-05, "loss": 0.0, "step": 59102 }, { "epoch": 1.15, "learning_rate": 3.0887823560998644e-05, "loss": 0.0, "step": 59104 }, { "epoch": 1.15, "learning_rate": 3.088717660074142e-05, "loss": 0.003, "step": 59106 }, { "epoch": 1.15, "learning_rate": 3.088652964048418e-05, "loss": 0.0, "step": 59108 }, { "epoch": 1.15, "learning_rate": 3.088588268022696e-05, "loss": 0.0, "step": 59110 }, { "epoch": 1.15, "learning_rate": 3.088523571996973e-05, "loss": 0.0, "step": 59112 }, { "epoch": 1.15, "learning_rate": 3.088458875971249e-05, "loss": 0.021, "step": 59114 }, { "epoch": 1.15, "learning_rate": 3.0883941799455266e-05, "loss": 0.0, "step": 59116 }, { "epoch": 1.15, "learning_rate": 3.088329483919803e-05, "loss": 0.0, "step": 59118 }, { "epoch": 1.15, "learning_rate": 3.08826478789408e-05, "loss": 0.0, "step": 59120 }, { "epoch": 1.15, "learning_rate": 3.0882000918683566e-05, "loss": 0.0004, "step": 59122 }, { "epoch": 1.15, "learning_rate": 3.0881353958426335e-05, "loss": 0.0, "step": 59124 }, { "epoch": 1.15, "learning_rate": 3.0880706998169104e-05, "loss": 0.0336, "step": 59126 }, { "epoch": 1.15, "learning_rate": 3.088006003791187e-05, "loss": 0.0, "step": 59128 }, { "epoch": 1.15, "learning_rate": 3.087941307765464e-05, "loss": 0.0, "step": 59130 }, { "epoch": 1.15, "learning_rate": 3.087876611739741e-05, "loss": 0.0003, "step": 59132 }, { "epoch": 1.15, "learning_rate": 3.087811915714018e-05, "loss": 0.0, "step": 59134 }, { "epoch": 1.15, "learning_rate": 3.087747219688294e-05, "loss": 0.001, "step": 59136 }, { "epoch": 1.15, "learning_rate": 3.087682523662572e-05, "loss": 0.0001, "step": 59138 }, { "epoch": 1.15, "learning_rate": 3.087617827636848e-05, "loss": 0.0016, "step": 59140 }, { "epoch": 1.15, "learning_rate": 3.087553131611125e-05, "loss": 0.0, "step": 59142 }, { "epoch": 1.15, "learning_rate": 3.087488435585402e-05, "loss": 0.0, "step": 59144 }, { "epoch": 1.15, "learning_rate": 3.087423739559679e-05, "loss": 0.0085, "step": 59146 }, { "epoch": 1.15, "learning_rate": 3.0873590435339564e-05, "loss": 0.0, "step": 59148 }, { "epoch": 1.15, "learning_rate": 3.0872943475082326e-05, "loss": 0.0, "step": 59150 }, { "epoch": 1.15, "learning_rate": 3.0872296514825096e-05, "loss": 0.0, "step": 59152 }, { "epoch": 1.15, "learning_rate": 3.0871649554567865e-05, "loss": 0.0, "step": 59154 }, { "epoch": 1.15, "learning_rate": 3.0871002594310634e-05, "loss": 0.0, "step": 59156 }, { "epoch": 1.15, "learning_rate": 3.08703556340534e-05, "loss": 0.0, "step": 59158 }, { "epoch": 1.15, "learning_rate": 3.086970867379617e-05, "loss": 0.0004, "step": 59160 }, { "epoch": 1.15, "learning_rate": 3.086906171353894e-05, "loss": 0.0001, "step": 59162 }, { "epoch": 1.15, "learning_rate": 3.08684147532817e-05, "loss": 0.0, "step": 59164 }, { "epoch": 1.15, "learning_rate": 3.086776779302448e-05, "loss": 0.0, "step": 59166 }, { "epoch": 1.15, "learning_rate": 3.086712083276724e-05, "loss": 0.0, "step": 59168 }, { "epoch": 1.15, "learning_rate": 3.086647387251002e-05, "loss": 0.0, "step": 59170 }, { "epoch": 1.15, "learning_rate": 3.086582691225278e-05, "loss": 0.0, "step": 59172 }, { "epoch": 1.15, "learning_rate": 3.086517995199555e-05, "loss": 0.0059, "step": 59174 }, { "epoch": 1.15, "learning_rate": 3.086453299173832e-05, "loss": 0.0, "step": 59176 }, { "epoch": 1.15, "learning_rate": 3.086388603148109e-05, "loss": 0.0, "step": 59178 }, { "epoch": 1.15, "learning_rate": 3.0863239071223856e-05, "loss": 0.0009, "step": 59180 }, { "epoch": 1.15, "learning_rate": 3.0862592110966625e-05, "loss": 0.0, "step": 59182 }, { "epoch": 1.15, "learning_rate": 3.0861945150709394e-05, "loss": 0.0001, "step": 59184 }, { "epoch": 1.15, "learning_rate": 3.086129819045216e-05, "loss": 0.0042, "step": 59186 }, { "epoch": 1.15, "learning_rate": 3.086065123019493e-05, "loss": 0.0356, "step": 59188 }, { "epoch": 1.15, "learning_rate": 3.0860004269937695e-05, "loss": 0.0, "step": 59190 }, { "epoch": 1.15, "learning_rate": 3.085935730968047e-05, "loss": 0.0, "step": 59192 }, { "epoch": 1.15, "learning_rate": 3.085871034942324e-05, "loss": 0.0013, "step": 59194 }, { "epoch": 1.15, "learning_rate": 3.0858063389166e-05, "loss": 0.0, "step": 59196 }, { "epoch": 1.15, "learning_rate": 3.085741642890878e-05, "loss": 0.0, "step": 59198 }, { "epoch": 1.15, "learning_rate": 3.085676946865154e-05, "loss": 0.0009, "step": 59200 }, { "epoch": 1.15, "learning_rate": 3.085612250839431e-05, "loss": 0.0, "step": 59202 }, { "epoch": 1.15, "learning_rate": 3.085547554813708e-05, "loss": 0.0006, "step": 59204 }, { "epoch": 1.15, "learning_rate": 3.085482858787985e-05, "loss": 0.0, "step": 59206 }, { "epoch": 1.15, "learning_rate": 3.0854181627622616e-05, "loss": 0.0, "step": 59208 }, { "epoch": 1.15, "learning_rate": 3.0853534667365385e-05, "loss": 0.0, "step": 59210 }, { "epoch": 1.15, "learning_rate": 3.0852887707108154e-05, "loss": 0.0, "step": 59212 }, { "epoch": 1.15, "learning_rate": 3.0852240746850924e-05, "loss": 0.0, "step": 59214 }, { "epoch": 1.15, "learning_rate": 3.085159378659369e-05, "loss": 0.0001, "step": 59216 }, { "epoch": 1.15, "learning_rate": 3.0850946826336455e-05, "loss": 0.0, "step": 59218 }, { "epoch": 1.15, "learning_rate": 3.085029986607923e-05, "loss": 0.0013, "step": 59220 }, { "epoch": 1.15, "learning_rate": 3.084965290582199e-05, "loss": 0.0019, "step": 59222 }, { "epoch": 1.15, "learning_rate": 3.084900594556476e-05, "loss": 0.0, "step": 59224 }, { "epoch": 1.15, "learning_rate": 3.084835898530754e-05, "loss": 0.0, "step": 59226 }, { "epoch": 1.15, "learning_rate": 3.08477120250503e-05, "loss": 0.0, "step": 59228 }, { "epoch": 1.15, "learning_rate": 3.0847065064793076e-05, "loss": 0.0003, "step": 59230 }, { "epoch": 1.15, "learning_rate": 3.084641810453584e-05, "loss": 0.0, "step": 59232 }, { "epoch": 1.15, "learning_rate": 3.084577114427861e-05, "loss": 0.0001, "step": 59234 }, { "epoch": 1.15, "learning_rate": 3.084512418402138e-05, "loss": 0.0003, "step": 59236 }, { "epoch": 1.15, "learning_rate": 3.0844477223764146e-05, "loss": 0.0001, "step": 59238 }, { "epoch": 1.15, "learning_rate": 3.0843830263506915e-05, "loss": 0.0, "step": 59240 }, { "epoch": 1.15, "learning_rate": 3.0843183303249684e-05, "loss": 0.0, "step": 59242 }, { "epoch": 1.15, "learning_rate": 3.084253634299245e-05, "loss": 0.0, "step": 59244 }, { "epoch": 1.15, "learning_rate": 3.0841889382735215e-05, "loss": 0.0, "step": 59246 }, { "epoch": 1.15, "learning_rate": 3.084124242247799e-05, "loss": 0.0, "step": 59248 }, { "epoch": 1.15, "learning_rate": 3.0840595462220754e-05, "loss": 0.0049, "step": 59250 }, { "epoch": 1.15, "learning_rate": 3.083994850196353e-05, "loss": 0.0, "step": 59252 }, { "epoch": 1.15, "learning_rate": 3.083930154170629e-05, "loss": 0.0043, "step": 59254 }, { "epoch": 1.15, "learning_rate": 3.083865458144906e-05, "loss": 0.0, "step": 59256 }, { "epoch": 1.15, "learning_rate": 3.083800762119184e-05, "loss": 0.0, "step": 59258 }, { "epoch": 1.15, "learning_rate": 3.08373606609346e-05, "loss": 0.0046, "step": 59260 }, { "epoch": 1.15, "learning_rate": 3.083671370067737e-05, "loss": 0.0023, "step": 59262 }, { "epoch": 1.15, "learning_rate": 3.083606674042014e-05, "loss": 0.0001, "step": 59264 }, { "epoch": 1.15, "learning_rate": 3.0835419780162906e-05, "loss": 0.0, "step": 59266 }, { "epoch": 1.15, "learning_rate": 3.0834772819905675e-05, "loss": 0.0, "step": 59268 }, { "epoch": 1.15, "learning_rate": 3.0834125859648444e-05, "loss": 0.0001, "step": 59270 }, { "epoch": 1.15, "learning_rate": 3.0833478899391213e-05, "loss": 0.0, "step": 59272 }, { "epoch": 1.15, "learning_rate": 3.083283193913398e-05, "loss": 0.0, "step": 59274 }, { "epoch": 1.15, "learning_rate": 3.083218497887675e-05, "loss": 0.0, "step": 59276 }, { "epoch": 1.15, "learning_rate": 3.0831538018619514e-05, "loss": 0.0001, "step": 59278 }, { "epoch": 1.15, "learning_rate": 3.083089105836229e-05, "loss": 0.0, "step": 59280 }, { "epoch": 1.15, "learning_rate": 3.083024409810505e-05, "loss": 0.0, "step": 59282 }, { "epoch": 1.15, "learning_rate": 3.082959713784782e-05, "loss": 0.0007, "step": 59284 }, { "epoch": 1.15, "learning_rate": 3.082895017759059e-05, "loss": 0.0, "step": 59286 }, { "epoch": 1.15, "learning_rate": 3.082830321733336e-05, "loss": 0.0, "step": 59288 }, { "epoch": 1.15, "learning_rate": 3.082765625707613e-05, "loss": 0.0, "step": 59290 }, { "epoch": 1.15, "learning_rate": 3.08270092968189e-05, "loss": 0.0006, "step": 59292 }, { "epoch": 1.15, "learning_rate": 3.082636233656167e-05, "loss": 0.0037, "step": 59294 }, { "epoch": 1.15, "learning_rate": 3.0825715376304436e-05, "loss": 0.0, "step": 59296 }, { "epoch": 1.15, "learning_rate": 3.0825068416047205e-05, "loss": 0.0, "step": 59298 }, { "epoch": 1.15, "learning_rate": 3.082442145578997e-05, "loss": 0.0, "step": 59300 }, { "epoch": 1.15, "learning_rate": 3.082377449553274e-05, "loss": 0.0, "step": 59302 }, { "epoch": 1.15, "learning_rate": 3.082312753527551e-05, "loss": 0.0, "step": 59304 }, { "epoch": 1.15, "learning_rate": 3.0822480575018274e-05, "loss": 0.0, "step": 59306 }, { "epoch": 1.15, "learning_rate": 3.082183361476105e-05, "loss": 0.0, "step": 59308 }, { "epoch": 1.15, "learning_rate": 3.082118665450381e-05, "loss": 0.0, "step": 59310 }, { "epoch": 1.15, "learning_rate": 3.082053969424659e-05, "loss": 0.0003, "step": 59312 }, { "epoch": 1.15, "learning_rate": 3.081989273398935e-05, "loss": 0.0, "step": 59314 }, { "epoch": 1.15, "learning_rate": 3.081924577373212e-05, "loss": 0.0, "step": 59316 }, { "epoch": 1.15, "learning_rate": 3.081859881347489e-05, "loss": 0.0, "step": 59318 }, { "epoch": 1.15, "learning_rate": 3.081795185321766e-05, "loss": 0.0, "step": 59320 }, { "epoch": 1.15, "learning_rate": 3.081730489296043e-05, "loss": 0.0002, "step": 59322 }, { "epoch": 1.15, "learning_rate": 3.0816657932703196e-05, "loss": 0.0003, "step": 59324 }, { "epoch": 1.15, "learning_rate": 3.0816010972445965e-05, "loss": 0.0, "step": 59326 }, { "epoch": 1.15, "learning_rate": 3.0815364012188734e-05, "loss": 0.0127, "step": 59328 }, { "epoch": 1.15, "learning_rate": 3.08147170519315e-05, "loss": 0.0, "step": 59330 }, { "epoch": 1.15, "learning_rate": 3.0814070091674266e-05, "loss": 0.0, "step": 59332 }, { "epoch": 1.15, "learning_rate": 3.081342313141704e-05, "loss": 0.0021, "step": 59334 }, { "epoch": 1.15, "learning_rate": 3.081277617115981e-05, "loss": 0.0, "step": 59336 }, { "epoch": 1.15, "learning_rate": 3.081212921090257e-05, "loss": 0.0, "step": 59338 }, { "epoch": 1.15, "learning_rate": 3.081148225064535e-05, "loss": 0.0024, "step": 59340 }, { "epoch": 1.15, "learning_rate": 3.081083529038811e-05, "loss": 0.0001, "step": 59342 }, { "epoch": 1.15, "learning_rate": 3.081018833013088e-05, "loss": 0.0, "step": 59344 }, { "epoch": 1.15, "learning_rate": 3.080954136987365e-05, "loss": 0.0013, "step": 59346 }, { "epoch": 1.15, "learning_rate": 3.080889440961642e-05, "loss": 0.0, "step": 59348 }, { "epoch": 1.15, "learning_rate": 3.080824744935919e-05, "loss": 0.0, "step": 59350 }, { "epoch": 1.15, "learning_rate": 3.0807600489101957e-05, "loss": 0.0084, "step": 59352 }, { "epoch": 1.15, "learning_rate": 3.0806953528844726e-05, "loss": 0.0, "step": 59354 }, { "epoch": 1.15, "learning_rate": 3.0806306568587495e-05, "loss": 0.0, "step": 59356 }, { "epoch": 1.15, "learning_rate": 3.0805659608330264e-05, "loss": 0.0001, "step": 59358 }, { "epoch": 1.15, "learning_rate": 3.0805012648073026e-05, "loss": 0.0, "step": 59360 }, { "epoch": 1.15, "learning_rate": 3.08043656878158e-05, "loss": 0.0067, "step": 59362 }, { "epoch": 1.15, "learning_rate": 3.0803718727558564e-05, "loss": 0.0, "step": 59364 }, { "epoch": 1.15, "learning_rate": 3.080307176730133e-05, "loss": 0.0001, "step": 59366 }, { "epoch": 1.15, "learning_rate": 3.08024248070441e-05, "loss": 0.0004, "step": 59368 }, { "epoch": 1.15, "learning_rate": 3.080177784678687e-05, "loss": 0.0, "step": 59370 }, { "epoch": 1.15, "learning_rate": 3.080113088652965e-05, "loss": 0.0001, "step": 59372 }, { "epoch": 1.15, "learning_rate": 3.080048392627241e-05, "loss": 0.0, "step": 59374 }, { "epoch": 1.15, "learning_rate": 3.079983696601518e-05, "loss": 0.0002, "step": 59376 }, { "epoch": 1.15, "learning_rate": 3.079919000575795e-05, "loss": 0.0, "step": 59378 }, { "epoch": 1.15, "learning_rate": 3.079854304550072e-05, "loss": 0.0054, "step": 59380 }, { "epoch": 1.15, "learning_rate": 3.0797896085243486e-05, "loss": 0.0001, "step": 59382 }, { "epoch": 1.15, "learning_rate": 3.0797249124986255e-05, "loss": 0.0001, "step": 59384 }, { "epoch": 1.15, "learning_rate": 3.0796602164729024e-05, "loss": 0.0, "step": 59386 }, { "epoch": 1.15, "learning_rate": 3.0795955204471786e-05, "loss": 0.0, "step": 59388 }, { "epoch": 1.15, "learning_rate": 3.079530824421456e-05, "loss": 0.0, "step": 59390 }, { "epoch": 1.15, "learning_rate": 3.0794661283957325e-05, "loss": 0.0, "step": 59392 }, { "epoch": 1.15, "learning_rate": 3.07940143237001e-05, "loss": 0.0, "step": 59394 }, { "epoch": 1.15, "learning_rate": 3.079336736344286e-05, "loss": 0.0, "step": 59396 }, { "epoch": 1.15, "learning_rate": 3.079272040318563e-05, "loss": 0.0, "step": 59398 }, { "epoch": 1.15, "learning_rate": 3.07920734429284e-05, "loss": 0.0082, "step": 59400 }, { "epoch": 1.15, "learning_rate": 3.079142648267117e-05, "loss": 0.0012, "step": 59402 }, { "epoch": 1.15, "learning_rate": 3.079077952241394e-05, "loss": 0.024, "step": 59404 }, { "epoch": 1.15, "learning_rate": 3.079013256215671e-05, "loss": 0.0, "step": 59406 }, { "epoch": 1.15, "learning_rate": 3.078948560189948e-05, "loss": 0.0, "step": 59408 }, { "epoch": 1.15, "learning_rate": 3.0788838641642246e-05, "loss": 0.0, "step": 59410 }, { "epoch": 1.15, "learning_rate": 3.0788191681385015e-05, "loss": 0.0001, "step": 59412 }, { "epoch": 1.15, "learning_rate": 3.0787544721127785e-05, "loss": 0.0, "step": 59414 }, { "epoch": 1.15, "learning_rate": 3.0786897760870554e-05, "loss": 0.0, "step": 59416 }, { "epoch": 1.15, "learning_rate": 3.078625080061332e-05, "loss": 0.0002, "step": 59418 }, { "epoch": 1.15, "learning_rate": 3.0785603840356085e-05, "loss": 0.0, "step": 59420 }, { "epoch": 1.15, "learning_rate": 3.078495688009886e-05, "loss": 0.0, "step": 59422 }, { "epoch": 1.15, "learning_rate": 3.078430991984162e-05, "loss": 0.0, "step": 59424 }, { "epoch": 1.15, "learning_rate": 3.078366295958439e-05, "loss": 0.0, "step": 59426 }, { "epoch": 1.15, "learning_rate": 3.078301599932716e-05, "loss": 0.0, "step": 59428 }, { "epoch": 1.15, "learning_rate": 3.078236903906993e-05, "loss": 0.0, "step": 59430 }, { "epoch": 1.15, "learning_rate": 3.07817220788127e-05, "loss": 0.0, "step": 59432 }, { "epoch": 1.15, "learning_rate": 3.078107511855547e-05, "loss": 0.0, "step": 59434 }, { "epoch": 1.15, "learning_rate": 3.078042815829824e-05, "loss": 0.0, "step": 59436 }, { "epoch": 1.15, "learning_rate": 3.077978119804101e-05, "loss": 0.0, "step": 59438 }, { "epoch": 1.15, "learning_rate": 3.0779134237783776e-05, "loss": 0.0001, "step": 59440 }, { "epoch": 1.15, "learning_rate": 3.077848727752654e-05, "loss": 0.0016, "step": 59442 }, { "epoch": 1.15, "learning_rate": 3.0777840317269314e-05, "loss": 0.0, "step": 59444 }, { "epoch": 1.15, "learning_rate": 3.0777193357012076e-05, "loss": 0.0, "step": 59446 }, { "epoch": 1.15, "learning_rate": 3.0776546396754845e-05, "loss": 0.0001, "step": 59448 }, { "epoch": 1.15, "learning_rate": 3.077589943649762e-05, "loss": 0.0, "step": 59450 }, { "epoch": 1.15, "learning_rate": 3.0775252476240384e-05, "loss": 0.0, "step": 59452 }, { "epoch": 1.15, "learning_rate": 3.077460551598316e-05, "loss": 0.0001, "step": 59454 }, { "epoch": 1.15, "learning_rate": 3.077395855572592e-05, "loss": 0.0103, "step": 59456 }, { "epoch": 1.15, "learning_rate": 3.077331159546869e-05, "loss": 0.0, "step": 59458 }, { "epoch": 1.15, "learning_rate": 3.077266463521146e-05, "loss": 0.0001, "step": 59460 }, { "epoch": 1.15, "learning_rate": 3.077201767495423e-05, "loss": 0.0001, "step": 59462 }, { "epoch": 1.15, "learning_rate": 3.0771370714697e-05, "loss": 0.0001, "step": 59464 }, { "epoch": 1.15, "learning_rate": 3.077072375443977e-05, "loss": 0.0, "step": 59466 }, { "epoch": 1.15, "learning_rate": 3.0770076794182536e-05, "loss": 0.0001, "step": 59468 }, { "epoch": 1.15, "learning_rate": 3.0769429833925305e-05, "loss": 0.0, "step": 59470 }, { "epoch": 1.15, "learning_rate": 3.0768782873668074e-05, "loss": 0.0, "step": 59472 }, { "epoch": 1.15, "learning_rate": 3.076813591341084e-05, "loss": 0.0, "step": 59474 }, { "epoch": 1.15, "learning_rate": 3.076748895315361e-05, "loss": 0.0, "step": 59476 }, { "epoch": 1.15, "learning_rate": 3.0766841992896375e-05, "loss": 0.0, "step": 59478 }, { "epoch": 1.15, "learning_rate": 3.0766195032639144e-05, "loss": 0.0187, "step": 59480 }, { "epoch": 1.15, "learning_rate": 3.076554807238192e-05, "loss": 0.0, "step": 59482 }, { "epoch": 1.15, "learning_rate": 3.076490111212468e-05, "loss": 0.0, "step": 59484 }, { "epoch": 1.15, "learning_rate": 3.076425415186745e-05, "loss": 0.0, "step": 59486 }, { "epoch": 1.15, "learning_rate": 3.076360719161022e-05, "loss": 0.0, "step": 59488 }, { "epoch": 1.15, "learning_rate": 3.076296023135299e-05, "loss": 0.0014, "step": 59490 }, { "epoch": 1.15, "learning_rate": 3.076231327109576e-05, "loss": 0.0, "step": 59492 }, { "epoch": 1.15, "learning_rate": 3.076166631083853e-05, "loss": 0.0, "step": 59494 }, { "epoch": 1.15, "learning_rate": 3.07610193505813e-05, "loss": 0.0, "step": 59496 }, { "epoch": 1.15, "learning_rate": 3.0760372390324066e-05, "loss": 0.0, "step": 59498 }, { "epoch": 1.15, "learning_rate": 3.0759725430066835e-05, "loss": 0.0, "step": 59500 }, { "epoch": 1.15, "learning_rate": 3.07590784698096e-05, "loss": 0.0, "step": 59502 }, { "epoch": 1.15, "learning_rate": 3.075843150955237e-05, "loss": 0.0, "step": 59504 }, { "epoch": 1.15, "learning_rate": 3.0757784549295135e-05, "loss": 0.0003, "step": 59506 }, { "epoch": 1.15, "learning_rate": 3.0757137589037904e-05, "loss": 0.0, "step": 59508 }, { "epoch": 1.16, "learning_rate": 3.0756490628780674e-05, "loss": 0.0003, "step": 59510 }, { "epoch": 1.16, "learning_rate": 3.075584366852344e-05, "loss": 0.0001, "step": 59512 }, { "epoch": 1.16, "learning_rate": 3.075519670826622e-05, "loss": 0.0, "step": 59514 }, { "epoch": 1.16, "learning_rate": 3.075454974800898e-05, "loss": 0.0, "step": 59516 }, { "epoch": 1.16, "learning_rate": 3.075390278775175e-05, "loss": 0.0001, "step": 59518 }, { "epoch": 1.16, "learning_rate": 3.075325582749452e-05, "loss": 0.0, "step": 59520 }, { "epoch": 1.16, "learning_rate": 3.075260886723729e-05, "loss": 0.0, "step": 59522 }, { "epoch": 1.16, "learning_rate": 3.075196190698005e-05, "loss": 0.0006, "step": 59524 }, { "epoch": 1.16, "learning_rate": 3.0751314946722826e-05, "loss": 0.0052, "step": 59526 }, { "epoch": 1.16, "learning_rate": 3.0750667986465595e-05, "loss": 0.0009, "step": 59528 }, { "epoch": 1.16, "learning_rate": 3.075002102620836e-05, "loss": 0.0, "step": 59530 }, { "epoch": 1.16, "learning_rate": 3.0749374065951133e-05, "loss": 0.0, "step": 59532 }, { "epoch": 1.16, "learning_rate": 3.0748727105693896e-05, "loss": 0.0, "step": 59534 }, { "epoch": 1.16, "learning_rate": 3.074808014543667e-05, "loss": 0.0, "step": 59536 }, { "epoch": 1.16, "learning_rate": 3.0747433185179434e-05, "loss": 0.0, "step": 59538 }, { "epoch": 1.16, "learning_rate": 3.07467862249222e-05, "loss": 0.0, "step": 59540 }, { "epoch": 1.16, "learning_rate": 3.074613926466497e-05, "loss": 0.0, "step": 59542 }, { "epoch": 1.16, "learning_rate": 3.074549230440774e-05, "loss": 0.0, "step": 59544 }, { "epoch": 1.16, "learning_rate": 3.074484534415051e-05, "loss": 0.0, "step": 59546 }, { "epoch": 1.16, "learning_rate": 3.074419838389328e-05, "loss": 0.0, "step": 59548 }, { "epoch": 1.16, "learning_rate": 3.074355142363605e-05, "loss": 0.0, "step": 59550 }, { "epoch": 1.16, "learning_rate": 3.074290446337882e-05, "loss": 0.0, "step": 59552 }, { "epoch": 1.16, "learning_rate": 3.0742257503121587e-05, "loss": 0.0, "step": 59554 }, { "epoch": 1.16, "learning_rate": 3.074161054286435e-05, "loss": 0.0, "step": 59556 }, { "epoch": 1.16, "learning_rate": 3.0740963582607125e-05, "loss": 0.0, "step": 59558 }, { "epoch": 1.16, "learning_rate": 3.0740316622349894e-05, "loss": 0.0, "step": 59560 }, { "epoch": 1.16, "learning_rate": 3.0739669662092656e-05, "loss": 0.0, "step": 59562 }, { "epoch": 1.16, "learning_rate": 3.073902270183543e-05, "loss": 0.0001, "step": 59564 }, { "epoch": 1.16, "learning_rate": 3.0738375741578194e-05, "loss": 0.0066, "step": 59566 }, { "epoch": 1.16, "learning_rate": 3.0737728781320963e-05, "loss": 0.0, "step": 59568 }, { "epoch": 1.16, "learning_rate": 3.073708182106373e-05, "loss": 0.0002, "step": 59570 }, { "epoch": 1.16, "learning_rate": 3.07364348608065e-05, "loss": 0.002, "step": 59572 }, { "epoch": 1.16, "learning_rate": 3.073578790054927e-05, "loss": 0.0002, "step": 59574 }, { "epoch": 1.16, "learning_rate": 3.073514094029204e-05, "loss": 0.0, "step": 59576 }, { "epoch": 1.16, "learning_rate": 3.073449398003481e-05, "loss": 0.0, "step": 59578 }, { "epoch": 1.16, "learning_rate": 3.073384701977758e-05, "loss": 0.0, "step": 59580 }, { "epoch": 1.16, "learning_rate": 3.073320005952035e-05, "loss": 0.0, "step": 59582 }, { "epoch": 1.16, "learning_rate": 3.073255309926311e-05, "loss": 0.0001, "step": 59584 }, { "epoch": 1.16, "learning_rate": 3.0731906139005885e-05, "loss": 0.0, "step": 59586 }, { "epoch": 1.16, "learning_rate": 3.073125917874865e-05, "loss": 0.0, "step": 59588 }, { "epoch": 1.16, "learning_rate": 3.0730612218491417e-05, "loss": 0.0, "step": 59590 }, { "epoch": 1.16, "learning_rate": 3.072996525823419e-05, "loss": 0.0, "step": 59592 }, { "epoch": 1.16, "learning_rate": 3.0729318297976955e-05, "loss": 0.0873, "step": 59594 }, { "epoch": 1.16, "learning_rate": 3.072867133771973e-05, "loss": 0.0001, "step": 59596 }, { "epoch": 1.16, "learning_rate": 3.072802437746249e-05, "loss": 0.0, "step": 59598 }, { "epoch": 1.16, "learning_rate": 3.072737741720526e-05, "loss": 0.0033, "step": 59600 }, { "epoch": 1.16, "learning_rate": 3.072673045694803e-05, "loss": 0.0001, "step": 59602 }, { "epoch": 1.16, "learning_rate": 3.07260834966908e-05, "loss": 0.0, "step": 59604 }, { "epoch": 1.16, "learning_rate": 3.072543653643357e-05, "loss": 0.0, "step": 59606 }, { "epoch": 1.16, "learning_rate": 3.072478957617634e-05, "loss": 0.0, "step": 59608 }, { "epoch": 1.16, "learning_rate": 3.072414261591911e-05, "loss": 0.0, "step": 59610 }, { "epoch": 1.16, "learning_rate": 3.0723495655661876e-05, "loss": 0.0, "step": 59612 }, { "epoch": 1.16, "learning_rate": 3.0722848695404646e-05, "loss": 0.0, "step": 59614 }, { "epoch": 1.16, "learning_rate": 3.072220173514741e-05, "loss": 0.0, "step": 59616 }, { "epoch": 1.16, "learning_rate": 3.0721554774890184e-05, "loss": 0.0, "step": 59618 }, { "epoch": 1.16, "learning_rate": 3.0720907814632946e-05, "loss": 0.0001, "step": 59620 }, { "epoch": 1.16, "learning_rate": 3.0720260854375715e-05, "loss": 0.0, "step": 59622 }, { "epoch": 1.16, "learning_rate": 3.0719613894118484e-05, "loss": 0.0, "step": 59624 }, { "epoch": 1.16, "learning_rate": 3.071896693386125e-05, "loss": 0.0, "step": 59626 }, { "epoch": 1.16, "learning_rate": 3.071831997360402e-05, "loss": 0.0, "step": 59628 }, { "epoch": 1.16, "learning_rate": 3.071767301334679e-05, "loss": 0.0, "step": 59630 }, { "epoch": 1.16, "learning_rate": 3.071702605308956e-05, "loss": 0.0001, "step": 59632 }, { "epoch": 1.16, "learning_rate": 3.071637909283233e-05, "loss": 0.0, "step": 59634 }, { "epoch": 1.16, "learning_rate": 3.07157321325751e-05, "loss": 0.0, "step": 59636 }, { "epoch": 1.16, "learning_rate": 3.071508517231787e-05, "loss": 0.0, "step": 59638 }, { "epoch": 1.16, "learning_rate": 3.071443821206064e-05, "loss": 0.0, "step": 59640 }, { "epoch": 1.16, "learning_rate": 3.0713791251803406e-05, "loss": 0.0, "step": 59642 }, { "epoch": 1.16, "learning_rate": 3.071314429154617e-05, "loss": 0.0, "step": 59644 }, { "epoch": 1.16, "learning_rate": 3.0712497331288944e-05, "loss": 0.0, "step": 59646 }, { "epoch": 1.16, "learning_rate": 3.0711850371031706e-05, "loss": 0.0, "step": 59648 }, { "epoch": 1.16, "learning_rate": 3.0711203410774476e-05, "loss": 0.0, "step": 59650 }, { "epoch": 1.16, "learning_rate": 3.0710556450517245e-05, "loss": 0.0, "step": 59652 }, { "epoch": 1.16, "learning_rate": 3.0709909490260014e-05, "loss": 0.0001, "step": 59654 }, { "epoch": 1.16, "learning_rate": 3.070926253000278e-05, "loss": 0.0, "step": 59656 }, { "epoch": 1.16, "learning_rate": 3.070861556974555e-05, "loss": 0.0, "step": 59658 }, { "epoch": 1.16, "learning_rate": 3.070796860948832e-05, "loss": 0.0001, "step": 59660 }, { "epoch": 1.16, "learning_rate": 3.070732164923109e-05, "loss": 0.0, "step": 59662 }, { "epoch": 1.16, "learning_rate": 3.070667468897386e-05, "loss": 0.0001, "step": 59664 }, { "epoch": 1.16, "learning_rate": 3.070602772871662e-05, "loss": 0.0, "step": 59666 }, { "epoch": 1.16, "learning_rate": 3.07053807684594e-05, "loss": 0.0, "step": 59668 }, { "epoch": 1.16, "learning_rate": 3.070473380820216e-05, "loss": 0.003, "step": 59670 }, { "epoch": 1.16, "learning_rate": 3.070408684794493e-05, "loss": 0.0, "step": 59672 }, { "epoch": 1.16, "learning_rate": 3.0703439887687705e-05, "loss": 0.0088, "step": 59674 }, { "epoch": 1.16, "learning_rate": 3.070279292743047e-05, "loss": 0.0, "step": 59676 }, { "epoch": 1.16, "learning_rate": 3.070214596717324e-05, "loss": 0.0005, "step": 59678 }, { "epoch": 1.16, "learning_rate": 3.0701499006916005e-05, "loss": 0.0, "step": 59680 }, { "epoch": 1.16, "learning_rate": 3.0700852046658774e-05, "loss": 0.0, "step": 59682 }, { "epoch": 1.16, "learning_rate": 3.070020508640154e-05, "loss": 0.0, "step": 59684 }, { "epoch": 1.16, "learning_rate": 3.069955812614431e-05, "loss": 0.0, "step": 59686 }, { "epoch": 1.16, "learning_rate": 3.069891116588708e-05, "loss": 0.0, "step": 59688 }, { "epoch": 1.16, "learning_rate": 3.069826420562985e-05, "loss": 0.0, "step": 59690 }, { "epoch": 1.16, "learning_rate": 3.069761724537262e-05, "loss": 0.0, "step": 59692 }, { "epoch": 1.16, "learning_rate": 3.069697028511539e-05, "loss": 0.0, "step": 59694 }, { "epoch": 1.16, "learning_rate": 3.069632332485816e-05, "loss": 0.0, "step": 59696 }, { "epoch": 1.16, "learning_rate": 3.069567636460092e-05, "loss": 0.0, "step": 59698 }, { "epoch": 1.16, "learning_rate": 3.0695029404343696e-05, "loss": 0.0, "step": 59700 }, { "epoch": 1.16, "learning_rate": 3.069438244408646e-05, "loss": 0.0, "step": 59702 }, { "epoch": 1.16, "learning_rate": 3.069373548382923e-05, "loss": 0.0001, "step": 59704 }, { "epoch": 1.16, "learning_rate": 3.0693088523572e-05, "loss": 0.0, "step": 59706 }, { "epoch": 1.16, "learning_rate": 3.0692441563314765e-05, "loss": 0.0, "step": 59708 }, { "epoch": 1.16, "learning_rate": 3.0691794603057535e-05, "loss": 0.0, "step": 59710 }, { "epoch": 1.16, "learning_rate": 3.0691147642800304e-05, "loss": 0.0019, "step": 59712 }, { "epoch": 1.16, "learning_rate": 3.069050068254307e-05, "loss": 0.0001, "step": 59714 }, { "epoch": 1.16, "learning_rate": 3.068985372228584e-05, "loss": 0.0024, "step": 59716 }, { "epoch": 1.16, "learning_rate": 3.068920676202861e-05, "loss": 0.0, "step": 59718 }, { "epoch": 1.16, "learning_rate": 3.068855980177138e-05, "loss": 0.0, "step": 59720 }, { "epoch": 1.16, "learning_rate": 3.068791284151415e-05, "loss": 0.0, "step": 59722 }, { "epoch": 1.16, "learning_rate": 3.068726588125692e-05, "loss": 0.0, "step": 59724 }, { "epoch": 1.16, "learning_rate": 3.068661892099968e-05, "loss": 0.0001, "step": 59726 }, { "epoch": 1.16, "learning_rate": 3.0685971960742456e-05, "loss": 0.0, "step": 59728 }, { "epoch": 1.16, "learning_rate": 3.068532500048522e-05, "loss": 0.0062, "step": 59730 }, { "epoch": 1.16, "learning_rate": 3.068467804022799e-05, "loss": 0.0, "step": 59732 }, { "epoch": 1.16, "learning_rate": 3.068403107997076e-05, "loss": 0.0, "step": 59734 }, { "epoch": 1.16, "learning_rate": 3.0683384119713526e-05, "loss": 0.0, "step": 59736 }, { "epoch": 1.16, "learning_rate": 3.06827371594563e-05, "loss": 0.0001, "step": 59738 }, { "epoch": 1.16, "learning_rate": 3.0682090199199064e-05, "loss": 0.0008, "step": 59740 }, { "epoch": 1.16, "learning_rate": 3.068144323894183e-05, "loss": 0.0024, "step": 59742 }, { "epoch": 1.16, "learning_rate": 3.06807962786846e-05, "loss": 0.0, "step": 59744 }, { "epoch": 1.16, "learning_rate": 3.068014931842737e-05, "loss": 0.0, "step": 59746 }, { "epoch": 1.16, "learning_rate": 3.0679502358170134e-05, "loss": 0.0, "step": 59748 }, { "epoch": 1.16, "learning_rate": 3.067885539791291e-05, "loss": 0.0, "step": 59750 }, { "epoch": 1.16, "learning_rate": 3.067820843765568e-05, "loss": 0.0, "step": 59752 }, { "epoch": 1.16, "learning_rate": 3.067756147739845e-05, "loss": 0.0, "step": 59754 }, { "epoch": 1.16, "learning_rate": 3.067691451714122e-05, "loss": 0.0027, "step": 59756 }, { "epoch": 1.16, "learning_rate": 3.067626755688398e-05, "loss": 0.0, "step": 59758 }, { "epoch": 1.16, "learning_rate": 3.0675620596626755e-05, "loss": 0.0002, "step": 59760 }, { "epoch": 1.16, "learning_rate": 3.067497363636952e-05, "loss": 0.0002, "step": 59762 }, { "epoch": 1.16, "learning_rate": 3.0674326676112286e-05, "loss": 0.0005, "step": 59764 }, { "epoch": 1.16, "learning_rate": 3.0673679715855055e-05, "loss": 0.0, "step": 59766 }, { "epoch": 1.16, "learning_rate": 3.0673032755597824e-05, "loss": 0.0022, "step": 59768 }, { "epoch": 1.16, "learning_rate": 3.0672385795340593e-05, "loss": 0.0025, "step": 59770 }, { "epoch": 1.16, "learning_rate": 3.067173883508336e-05, "loss": 0.0, "step": 59772 }, { "epoch": 1.16, "learning_rate": 3.067109187482613e-05, "loss": 0.0001, "step": 59774 }, { "epoch": 1.16, "learning_rate": 3.06704449145689e-05, "loss": 0.0, "step": 59776 }, { "epoch": 1.16, "learning_rate": 3.066979795431167e-05, "loss": 0.0008, "step": 59778 }, { "epoch": 1.16, "learning_rate": 3.066915099405443e-05, "loss": 0.0, "step": 59780 }, { "epoch": 1.16, "learning_rate": 3.066850403379721e-05, "loss": 0.0, "step": 59782 }, { "epoch": 1.16, "learning_rate": 3.066785707353998e-05, "loss": 0.0, "step": 59784 }, { "epoch": 1.16, "learning_rate": 3.066721011328274e-05, "loss": 0.0148, "step": 59786 }, { "epoch": 1.16, "learning_rate": 3.0666563153025515e-05, "loss": 0.0079, "step": 59788 }, { "epoch": 1.16, "learning_rate": 3.066591619276828e-05, "loss": 0.0, "step": 59790 }, { "epoch": 1.16, "learning_rate": 3.066526923251105e-05, "loss": 0.0003, "step": 59792 }, { "epoch": 1.16, "learning_rate": 3.0664622272253816e-05, "loss": 0.0, "step": 59794 }, { "epoch": 1.16, "learning_rate": 3.0663975311996585e-05, "loss": 0.0, "step": 59796 }, { "epoch": 1.16, "learning_rate": 3.0663328351739354e-05, "loss": 0.0002, "step": 59798 }, { "epoch": 1.16, "learning_rate": 3.066268139148212e-05, "loss": 0.0, "step": 59800 }, { "epoch": 1.16, "learning_rate": 3.066203443122489e-05, "loss": 0.0, "step": 59802 }, { "epoch": 1.16, "learning_rate": 3.066138747096766e-05, "loss": 0.0, "step": 59804 }, { "epoch": 1.16, "learning_rate": 3.066074051071043e-05, "loss": 0.0, "step": 59806 }, { "epoch": 1.16, "learning_rate": 3.066009355045319e-05, "loss": 0.0013, "step": 59808 }, { "epoch": 1.16, "learning_rate": 3.065944659019597e-05, "loss": 0.0, "step": 59810 }, { "epoch": 1.16, "learning_rate": 3.065879962993873e-05, "loss": 0.0, "step": 59812 }, { "epoch": 1.16, "learning_rate": 3.06581526696815e-05, "loss": 0.0, "step": 59814 }, { "epoch": 1.16, "learning_rate": 3.0657505709424276e-05, "loss": 0.0, "step": 59816 }, { "epoch": 1.16, "learning_rate": 3.065685874916704e-05, "loss": 0.0001, "step": 59818 }, { "epoch": 1.16, "learning_rate": 3.0656211788909814e-05, "loss": 0.0006, "step": 59820 }, { "epoch": 1.16, "learning_rate": 3.0655564828652576e-05, "loss": 0.0, "step": 59822 }, { "epoch": 1.16, "learning_rate": 3.0654917868395345e-05, "loss": 0.0, "step": 59824 }, { "epoch": 1.16, "learning_rate": 3.0654270908138114e-05, "loss": 0.0, "step": 59826 }, { "epoch": 1.16, "learning_rate": 3.0653623947880883e-05, "loss": 0.0002, "step": 59828 }, { "epoch": 1.16, "learning_rate": 3.065297698762365e-05, "loss": 0.0, "step": 59830 }, { "epoch": 1.16, "learning_rate": 3.065233002736642e-05, "loss": 0.0, "step": 59832 }, { "epoch": 1.16, "learning_rate": 3.065168306710919e-05, "loss": 0.0, "step": 59834 }, { "epoch": 1.16, "learning_rate": 3.065103610685196e-05, "loss": 0.0, "step": 59836 }, { "epoch": 1.16, "learning_rate": 3.065038914659473e-05, "loss": 0.0, "step": 59838 }, { "epoch": 1.16, "learning_rate": 3.064974218633749e-05, "loss": 0.0, "step": 59840 }, { "epoch": 1.16, "learning_rate": 3.064909522608027e-05, "loss": 0.001, "step": 59842 }, { "epoch": 1.16, "learning_rate": 3.064844826582303e-05, "loss": 0.0, "step": 59844 }, { "epoch": 1.16, "learning_rate": 3.06478013055658e-05, "loss": 0.0001, "step": 59846 }, { "epoch": 1.16, "learning_rate": 3.064715434530857e-05, "loss": 0.0, "step": 59848 }, { "epoch": 1.16, "learning_rate": 3.0646507385051337e-05, "loss": 0.0009, "step": 59850 }, { "epoch": 1.16, "learning_rate": 3.0645860424794106e-05, "loss": 0.0, "step": 59852 }, { "epoch": 1.16, "learning_rate": 3.0645213464536875e-05, "loss": 0.0, "step": 59854 }, { "epoch": 1.16, "learning_rate": 3.0644566504279644e-05, "loss": 0.0, "step": 59856 }, { "epoch": 1.16, "learning_rate": 3.064391954402241e-05, "loss": 0.0, "step": 59858 }, { "epoch": 1.16, "learning_rate": 3.064327258376518e-05, "loss": 0.01, "step": 59860 }, { "epoch": 1.16, "learning_rate": 3.064262562350795e-05, "loss": 0.0, "step": 59862 }, { "epoch": 1.16, "learning_rate": 3.064197866325072e-05, "loss": 0.0, "step": 59864 }, { "epoch": 1.16, "learning_rate": 3.064133170299349e-05, "loss": 0.0, "step": 59866 }, { "epoch": 1.16, "learning_rate": 3.064068474273625e-05, "loss": 0.0, "step": 59868 }, { "epoch": 1.16, "learning_rate": 3.064003778247903e-05, "loss": 0.0007, "step": 59870 }, { "epoch": 1.16, "learning_rate": 3.063939082222179e-05, "loss": 0.0, "step": 59872 }, { "epoch": 1.16, "learning_rate": 3.063874386196456e-05, "loss": 0.0, "step": 59874 }, { "epoch": 1.16, "learning_rate": 3.063809690170733e-05, "loss": 0.0003, "step": 59876 }, { "epoch": 1.16, "learning_rate": 3.06374499414501e-05, "loss": 0.0, "step": 59878 }, { "epoch": 1.16, "learning_rate": 3.0636802981192866e-05, "loss": 0.0, "step": 59880 }, { "epoch": 1.16, "learning_rate": 3.0636156020935635e-05, "loss": 0.0001, "step": 59882 }, { "epoch": 1.16, "learning_rate": 3.0635509060678404e-05, "loss": 0.0001, "step": 59884 }, { "epoch": 1.16, "learning_rate": 3.063486210042117e-05, "loss": 0.0794, "step": 59886 }, { "epoch": 1.16, "learning_rate": 3.063421514016394e-05, "loss": 0.0, "step": 59888 }, { "epoch": 1.16, "learning_rate": 3.0633568179906705e-05, "loss": 0.0074, "step": 59890 }, { "epoch": 1.16, "learning_rate": 3.063292121964948e-05, "loss": 0.0, "step": 59892 }, { "epoch": 1.16, "learning_rate": 3.063227425939225e-05, "loss": 0.0001, "step": 59894 }, { "epoch": 1.16, "learning_rate": 3.063162729913501e-05, "loss": 0.0, "step": 59896 }, { "epoch": 1.16, "learning_rate": 3.063098033887779e-05, "loss": 0.0, "step": 59898 }, { "epoch": 1.16, "learning_rate": 3.063033337862055e-05, "loss": 0.0038, "step": 59900 }, { "epoch": 1.16, "learning_rate": 3.0629686418363326e-05, "loss": 0.0, "step": 59902 }, { "epoch": 1.16, "learning_rate": 3.062903945810609e-05, "loss": 0.0, "step": 59904 }, { "epoch": 1.16, "learning_rate": 3.062839249784886e-05, "loss": 0.0001, "step": 59906 }, { "epoch": 1.16, "learning_rate": 3.0627745537591626e-05, "loss": 0.0, "step": 59908 }, { "epoch": 1.16, "learning_rate": 3.0627098577334396e-05, "loss": 0.0, "step": 59910 }, { "epoch": 1.16, "learning_rate": 3.0626451617077165e-05, "loss": 0.0, "step": 59912 }, { "epoch": 1.16, "learning_rate": 3.0625804656819934e-05, "loss": 0.0002, "step": 59914 }, { "epoch": 1.16, "learning_rate": 3.06251576965627e-05, "loss": 0.0, "step": 59916 }, { "epoch": 1.16, "learning_rate": 3.062451073630547e-05, "loss": 0.0, "step": 59918 }, { "epoch": 1.16, "learning_rate": 3.062386377604824e-05, "loss": 0.0, "step": 59920 }, { "epoch": 1.16, "learning_rate": 3.0623216815791e-05, "loss": 0.0, "step": 59922 }, { "epoch": 1.16, "learning_rate": 3.062256985553378e-05, "loss": 0.0, "step": 59924 }, { "epoch": 1.16, "learning_rate": 3.062192289527654e-05, "loss": 0.0, "step": 59926 }, { "epoch": 1.16, "learning_rate": 3.062127593501931e-05, "loss": 0.0, "step": 59928 }, { "epoch": 1.16, "learning_rate": 3.0620628974762086e-05, "loss": 0.0, "step": 59930 }, { "epoch": 1.16, "learning_rate": 3.061998201450485e-05, "loss": 0.0001, "step": 59932 }, { "epoch": 1.16, "learning_rate": 3.061933505424762e-05, "loss": 0.0, "step": 59934 }, { "epoch": 1.16, "learning_rate": 3.061868809399039e-05, "loss": 0.0, "step": 59936 }, { "epoch": 1.16, "learning_rate": 3.0618041133733156e-05, "loss": 0.0011, "step": 59938 }, { "epoch": 1.16, "learning_rate": 3.0617394173475925e-05, "loss": 0.0197, "step": 59940 }, { "epoch": 1.16, "learning_rate": 3.0616747213218694e-05, "loss": 0.0, "step": 59942 }, { "epoch": 1.16, "learning_rate": 3.061610025296146e-05, "loss": 0.0, "step": 59944 }, { "epoch": 1.16, "learning_rate": 3.061545329270423e-05, "loss": 0.0, "step": 59946 }, { "epoch": 1.16, "learning_rate": 3.0614806332447e-05, "loss": 0.0, "step": 59948 }, { "epoch": 1.16, "learning_rate": 3.0614159372189764e-05, "loss": 0.0, "step": 59950 }, { "epoch": 1.16, "learning_rate": 3.061351241193254e-05, "loss": 0.0, "step": 59952 }, { "epoch": 1.16, "learning_rate": 3.06128654516753e-05, "loss": 0.0005, "step": 59954 }, { "epoch": 1.16, "learning_rate": 3.061221849141807e-05, "loss": 0.0, "step": 59956 }, { "epoch": 1.16, "learning_rate": 3.061157153116084e-05, "loss": 0.0138, "step": 59958 }, { "epoch": 1.16, "learning_rate": 3.061092457090361e-05, "loss": 0.0, "step": 59960 }, { "epoch": 1.16, "learning_rate": 3.0610277610646385e-05, "loss": 0.0017, "step": 59962 }, { "epoch": 1.16, "learning_rate": 3.060963065038915e-05, "loss": 0.0, "step": 59964 }, { "epoch": 1.16, "learning_rate": 3.0608983690131916e-05, "loss": 0.0, "step": 59966 }, { "epoch": 1.16, "learning_rate": 3.0608336729874685e-05, "loss": 0.0, "step": 59968 }, { "epoch": 1.16, "learning_rate": 3.0607689769617454e-05, "loss": 0.0068, "step": 59970 }, { "epoch": 1.16, "learning_rate": 3.060704280936022e-05, "loss": 0.0001, "step": 59972 }, { "epoch": 1.16, "learning_rate": 3.060639584910299e-05, "loss": 0.0027, "step": 59974 }, { "epoch": 1.16, "learning_rate": 3.060574888884576e-05, "loss": 0.0077, "step": 59976 }, { "epoch": 1.16, "learning_rate": 3.060510192858853e-05, "loss": 0.0, "step": 59978 }, { "epoch": 1.16, "learning_rate": 3.06044549683313e-05, "loss": 0.0067, "step": 59980 }, { "epoch": 1.16, "learning_rate": 3.060380800807406e-05, "loss": 0.0, "step": 59982 }, { "epoch": 1.16, "learning_rate": 3.060316104781684e-05, "loss": 0.0, "step": 59984 }, { "epoch": 1.16, "learning_rate": 3.06025140875596e-05, "loss": 0.0, "step": 59986 }, { "epoch": 1.16, "learning_rate": 3.060186712730237e-05, "loss": 0.0002, "step": 59988 }, { "epoch": 1.16, "learning_rate": 3.060122016704514e-05, "loss": 0.0, "step": 59990 }, { "epoch": 1.16, "learning_rate": 3.060057320678791e-05, "loss": 0.0, "step": 59992 }, { "epoch": 1.16, "learning_rate": 3.059992624653068e-05, "loss": 0.0, "step": 59994 }, { "epoch": 1.16, "learning_rate": 3.0599279286273446e-05, "loss": 0.0, "step": 59996 }, { "epoch": 1.16, "learning_rate": 3.0598632326016215e-05, "loss": 0.007, "step": 59998 }, { "epoch": 1.16, "learning_rate": 3.0597985365758984e-05, "loss": 0.0, "step": 60000 }, { "epoch": 1.16, "learning_rate": 3.059733840550175e-05, "loss": 0.0, "step": 60002 }, { "epoch": 1.16, "learning_rate": 3.0596691445244515e-05, "loss": 0.0, "step": 60004 }, { "epoch": 1.16, "learning_rate": 3.059604448498729e-05, "loss": 0.0001, "step": 60006 }, { "epoch": 1.16, "learning_rate": 3.059539752473006e-05, "loss": 0.0, "step": 60008 }, { "epoch": 1.16, "learning_rate": 3.059475056447282e-05, "loss": 0.0, "step": 60010 }, { "epoch": 1.16, "learning_rate": 3.05941036042156e-05, "loss": 0.0, "step": 60012 }, { "epoch": 1.16, "learning_rate": 3.059345664395836e-05, "loss": 0.0113, "step": 60014 }, { "epoch": 1.16, "learning_rate": 3.059280968370113e-05, "loss": 0.0, "step": 60016 }, { "epoch": 1.16, "learning_rate": 3.05921627234439e-05, "loss": 0.0003, "step": 60018 }, { "epoch": 1.16, "learning_rate": 3.059151576318667e-05, "loss": 0.0, "step": 60020 }, { "epoch": 1.16, "learning_rate": 3.059086880292944e-05, "loss": 0.0003, "step": 60022 }, { "epoch": 1.16, "learning_rate": 3.0590221842672206e-05, "loss": 0.0, "step": 60024 }, { "epoch": 1.17, "learning_rate": 3.0589574882414975e-05, "loss": 0.0001, "step": 60026 }, { "epoch": 1.17, "learning_rate": 3.0588927922157744e-05, "loss": 0.0003, "step": 60028 }, { "epoch": 1.17, "learning_rate": 3.0588280961900513e-05, "loss": 0.0, "step": 60030 }, { "epoch": 1.17, "learning_rate": 3.0587634001643276e-05, "loss": 0.0, "step": 60032 }, { "epoch": 1.17, "learning_rate": 3.058698704138605e-05, "loss": 0.0, "step": 60034 }, { "epoch": 1.17, "learning_rate": 3.0586340081128814e-05, "loss": 0.0, "step": 60036 }, { "epoch": 1.17, "learning_rate": 3.058569312087158e-05, "loss": 0.0, "step": 60038 }, { "epoch": 1.17, "learning_rate": 3.058504616061436e-05, "loss": 0.0, "step": 60040 }, { "epoch": 1.17, "learning_rate": 3.058439920035712e-05, "loss": 0.0, "step": 60042 }, { "epoch": 1.17, "learning_rate": 3.05837522400999e-05, "loss": 0.0, "step": 60044 }, { "epoch": 1.17, "learning_rate": 3.058310527984266e-05, "loss": 0.0001, "step": 60046 }, { "epoch": 1.17, "learning_rate": 3.058245831958543e-05, "loss": 0.0, "step": 60048 }, { "epoch": 1.17, "learning_rate": 3.05818113593282e-05, "loss": 0.0001, "step": 60050 }, { "epoch": 1.17, "learning_rate": 3.058116439907097e-05, "loss": 0.0001, "step": 60052 }, { "epoch": 1.17, "learning_rate": 3.0580517438813736e-05, "loss": 0.0, "step": 60054 }, { "epoch": 1.17, "learning_rate": 3.0579870478556505e-05, "loss": 0.0, "step": 60056 }, { "epoch": 1.17, "learning_rate": 3.0579223518299274e-05, "loss": 0.0, "step": 60058 }, { "epoch": 1.17, "learning_rate": 3.057857655804204e-05, "loss": 0.0061, "step": 60060 }, { "epoch": 1.17, "learning_rate": 3.057792959778481e-05, "loss": 0.0, "step": 60062 }, { "epoch": 1.17, "learning_rate": 3.0577282637527574e-05, "loss": 0.0, "step": 60064 }, { "epoch": 1.17, "learning_rate": 3.057663567727035e-05, "loss": 0.0, "step": 60066 }, { "epoch": 1.17, "learning_rate": 3.057598871701311e-05, "loss": 0.0, "step": 60068 }, { "epoch": 1.17, "learning_rate": 3.057534175675588e-05, "loss": 0.0, "step": 60070 }, { "epoch": 1.17, "learning_rate": 3.057469479649866e-05, "loss": 0.0, "step": 60072 }, { "epoch": 1.17, "learning_rate": 3.057404783624142e-05, "loss": 0.0, "step": 60074 }, { "epoch": 1.17, "learning_rate": 3.057340087598419e-05, "loss": 0.0, "step": 60076 }, { "epoch": 1.17, "learning_rate": 3.057275391572696e-05, "loss": 0.0003, "step": 60078 }, { "epoch": 1.17, "learning_rate": 3.057210695546973e-05, "loss": 0.0011, "step": 60080 }, { "epoch": 1.17, "learning_rate": 3.0571459995212496e-05, "loss": 0.0, "step": 60082 }, { "epoch": 1.17, "learning_rate": 3.0570813034955265e-05, "loss": 0.0, "step": 60084 }, { "epoch": 1.17, "learning_rate": 3.0570166074698034e-05, "loss": 0.0, "step": 60086 }, { "epoch": 1.17, "learning_rate": 3.05695191144408e-05, "loss": 0.0, "step": 60088 }, { "epoch": 1.17, "learning_rate": 3.056887215418357e-05, "loss": 0.0003, "step": 60090 }, { "epoch": 1.17, "learning_rate": 3.0568225193926335e-05, "loss": 0.0, "step": 60092 }, { "epoch": 1.17, "learning_rate": 3.056757823366911e-05, "loss": 0.0, "step": 60094 }, { "epoch": 1.17, "learning_rate": 3.056693127341187e-05, "loss": 0.0, "step": 60096 }, { "epoch": 1.17, "learning_rate": 3.056628431315464e-05, "loss": 0.0007, "step": 60098 }, { "epoch": 1.17, "learning_rate": 3.056563735289741e-05, "loss": 0.0, "step": 60100 }, { "epoch": 1.17, "learning_rate": 3.056499039264018e-05, "loss": 0.0, "step": 60102 }, { "epoch": 1.17, "learning_rate": 3.056434343238295e-05, "loss": 0.0, "step": 60104 }, { "epoch": 1.17, "learning_rate": 3.056369647212572e-05, "loss": 0.0, "step": 60106 }, { "epoch": 1.17, "learning_rate": 3.056304951186849e-05, "loss": 0.0137, "step": 60108 }, { "epoch": 1.17, "learning_rate": 3.0562402551611257e-05, "loss": 0.0002, "step": 60110 }, { "epoch": 1.17, "learning_rate": 3.0561755591354026e-05, "loss": 0.0004, "step": 60112 }, { "epoch": 1.17, "learning_rate": 3.056110863109679e-05, "loss": 0.0002, "step": 60114 }, { "epoch": 1.17, "learning_rate": 3.0560461670839564e-05, "loss": 0.0, "step": 60116 }, { "epoch": 1.17, "learning_rate": 3.055981471058233e-05, "loss": 0.0126, "step": 60118 }, { "epoch": 1.17, "learning_rate": 3.05591677503251e-05, "loss": 0.0, "step": 60120 }, { "epoch": 1.17, "learning_rate": 3.055852079006787e-05, "loss": 0.0002, "step": 60122 }, { "epoch": 1.17, "learning_rate": 3.055787382981063e-05, "loss": 0.0, "step": 60124 }, { "epoch": 1.17, "learning_rate": 3.055722686955341e-05, "loss": 0.0046, "step": 60126 }, { "epoch": 1.17, "learning_rate": 3.055657990929617e-05, "loss": 0.0001, "step": 60128 }, { "epoch": 1.17, "learning_rate": 3.055593294903894e-05, "loss": 0.0, "step": 60130 }, { "epoch": 1.17, "learning_rate": 3.055528598878171e-05, "loss": 0.0, "step": 60132 }, { "epoch": 1.17, "learning_rate": 3.055463902852448e-05, "loss": 0.0, "step": 60134 }, { "epoch": 1.17, "learning_rate": 3.055399206826725e-05, "loss": 0.0, "step": 60136 }, { "epoch": 1.17, "learning_rate": 3.055334510801002e-05, "loss": 0.0, "step": 60138 }, { "epoch": 1.17, "learning_rate": 3.0552698147752786e-05, "loss": 0.0, "step": 60140 }, { "epoch": 1.17, "learning_rate": 3.0552051187495555e-05, "loss": 0.0, "step": 60142 }, { "epoch": 1.17, "learning_rate": 3.0551404227238324e-05, "loss": 0.0, "step": 60144 }, { "epoch": 1.17, "learning_rate": 3.0550757266981086e-05, "loss": 0.001, "step": 60146 }, { "epoch": 1.17, "learning_rate": 3.055011030672386e-05, "loss": 0.0, "step": 60148 }, { "epoch": 1.17, "learning_rate": 3.0549463346466625e-05, "loss": 0.0, "step": 60150 }, { "epoch": 1.17, "learning_rate": 3.0548816386209394e-05, "loss": 0.0001, "step": 60152 }, { "epoch": 1.17, "learning_rate": 3.054816942595217e-05, "loss": 0.0, "step": 60154 }, { "epoch": 1.17, "learning_rate": 3.054752246569493e-05, "loss": 0.0159, "step": 60156 }, { "epoch": 1.17, "learning_rate": 3.05468755054377e-05, "loss": 0.0, "step": 60158 }, { "epoch": 1.17, "learning_rate": 3.054622854518047e-05, "loss": 0.0, "step": 60160 }, { "epoch": 1.17, "learning_rate": 3.054558158492324e-05, "loss": 0.0, "step": 60162 }, { "epoch": 1.17, "learning_rate": 3.054493462466601e-05, "loss": 0.0, "step": 60164 }, { "epoch": 1.17, "learning_rate": 3.054428766440878e-05, "loss": 0.0, "step": 60166 }, { "epoch": 1.17, "learning_rate": 3.0543640704151546e-05, "loss": 0.0001, "step": 60168 }, { "epoch": 1.17, "learning_rate": 3.0542993743894316e-05, "loss": 0.0, "step": 60170 }, { "epoch": 1.17, "learning_rate": 3.0542346783637085e-05, "loss": 0.0, "step": 60172 }, { "epoch": 1.17, "learning_rate": 3.054169982337985e-05, "loss": 0.0068, "step": 60174 }, { "epoch": 1.17, "learning_rate": 3.054105286312262e-05, "loss": 0.0, "step": 60176 }, { "epoch": 1.17, "learning_rate": 3.0540405902865385e-05, "loss": 0.0, "step": 60178 }, { "epoch": 1.17, "learning_rate": 3.0539758942608154e-05, "loss": 0.0082, "step": 60180 }, { "epoch": 1.17, "learning_rate": 3.053911198235092e-05, "loss": 0.0, "step": 60182 }, { "epoch": 1.17, "learning_rate": 3.053846502209369e-05, "loss": 0.0, "step": 60184 }, { "epoch": 1.17, "learning_rate": 3.053781806183647e-05, "loss": 0.0, "step": 60186 }, { "epoch": 1.17, "learning_rate": 3.053717110157923e-05, "loss": 0.0, "step": 60188 }, { "epoch": 1.17, "learning_rate": 3.0536524141322e-05, "loss": 0.0, "step": 60190 }, { "epoch": 1.17, "learning_rate": 3.053587718106477e-05, "loss": 0.0, "step": 60192 }, { "epoch": 1.17, "learning_rate": 3.053523022080754e-05, "loss": 0.0001, "step": 60194 }, { "epoch": 1.17, "learning_rate": 3.053458326055031e-05, "loss": 0.0, "step": 60196 }, { "epoch": 1.17, "learning_rate": 3.0533936300293076e-05, "loss": 0.0, "step": 60198 }, { "epoch": 1.17, "learning_rate": 3.0533289340035845e-05, "loss": 0.0, "step": 60200 }, { "epoch": 1.17, "learning_rate": 3.0532642379778614e-05, "loss": 0.0, "step": 60202 }, { "epoch": 1.17, "learning_rate": 3.053199541952138e-05, "loss": 0.0004, "step": 60204 }, { "epoch": 1.17, "learning_rate": 3.0531348459264145e-05, "loss": 0.0109, "step": 60206 }, { "epoch": 1.17, "learning_rate": 3.053070149900692e-05, "loss": 0.0, "step": 60208 }, { "epoch": 1.17, "learning_rate": 3.0530054538749684e-05, "loss": 0.0002, "step": 60210 }, { "epoch": 1.17, "learning_rate": 3.052940757849245e-05, "loss": 0.0086, "step": 60212 }, { "epoch": 1.17, "learning_rate": 3.052876061823522e-05, "loss": 0.0, "step": 60214 }, { "epoch": 1.17, "learning_rate": 3.052811365797799e-05, "loss": 0.0, "step": 60216 }, { "epoch": 1.17, "learning_rate": 3.052746669772076e-05, "loss": 0.0, "step": 60218 }, { "epoch": 1.17, "learning_rate": 3.052681973746353e-05, "loss": 0.0, "step": 60220 }, { "epoch": 1.17, "learning_rate": 3.05261727772063e-05, "loss": 0.0001, "step": 60222 }, { "epoch": 1.17, "learning_rate": 3.052552581694907e-05, "loss": 0.0, "step": 60224 }, { "epoch": 1.17, "learning_rate": 3.0524878856691836e-05, "loss": 0.0, "step": 60226 }, { "epoch": 1.17, "learning_rate": 3.05242318964346e-05, "loss": 0.0001, "step": 60228 }, { "epoch": 1.17, "learning_rate": 3.0523584936177374e-05, "loss": 0.0179, "step": 60230 }, { "epoch": 1.17, "learning_rate": 3.0522937975920144e-05, "loss": 0.0013, "step": 60232 }, { "epoch": 1.17, "learning_rate": 3.0522291015662906e-05, "loss": 0.0, "step": 60234 }, { "epoch": 1.17, "learning_rate": 3.052164405540568e-05, "loss": 0.0017, "step": 60236 }, { "epoch": 1.17, "learning_rate": 3.0520997095148444e-05, "loss": 0.0, "step": 60238 }, { "epoch": 1.17, "learning_rate": 3.052035013489121e-05, "loss": 0.0, "step": 60240 }, { "epoch": 1.17, "learning_rate": 3.051970317463398e-05, "loss": 0.0, "step": 60242 }, { "epoch": 1.17, "learning_rate": 3.051905621437675e-05, "loss": 0.0, "step": 60244 }, { "epoch": 1.17, "learning_rate": 3.051840925411952e-05, "loss": 0.0001, "step": 60246 }, { "epoch": 1.17, "learning_rate": 3.051776229386229e-05, "loss": 0.0, "step": 60248 }, { "epoch": 1.17, "learning_rate": 3.0517115333605055e-05, "loss": 0.0002, "step": 60250 }, { "epoch": 1.17, "learning_rate": 3.0516468373347828e-05, "loss": 0.0, "step": 60252 }, { "epoch": 1.17, "learning_rate": 3.0515821413090597e-05, "loss": 0.0, "step": 60254 }, { "epoch": 1.17, "learning_rate": 3.0515174452833362e-05, "loss": 0.0, "step": 60256 }, { "epoch": 1.17, "learning_rate": 3.0514527492576135e-05, "loss": 0.0, "step": 60258 }, { "epoch": 1.17, "learning_rate": 3.05138805323189e-05, "loss": 0.0, "step": 60260 }, { "epoch": 1.17, "learning_rate": 3.0513233572061673e-05, "loss": 0.0, "step": 60262 }, { "epoch": 1.17, "learning_rate": 3.051258661180444e-05, "loss": 0.0, "step": 60264 }, { "epoch": 1.17, "learning_rate": 3.0511939651547204e-05, "loss": 0.0, "step": 60266 }, { "epoch": 1.17, "learning_rate": 3.0511292691289977e-05, "loss": 0.0, "step": 60268 }, { "epoch": 1.17, "learning_rate": 3.0510645731032743e-05, "loss": 0.0, "step": 60270 }, { "epoch": 1.17, "learning_rate": 3.0509998770775512e-05, "loss": 0.0, "step": 60272 }, { "epoch": 1.17, "learning_rate": 3.0509351810518284e-05, "loss": 0.0, "step": 60274 }, { "epoch": 1.17, "learning_rate": 3.050870485026105e-05, "loss": 0.0, "step": 60276 }, { "epoch": 1.17, "learning_rate": 3.0508057890003816e-05, "loss": 0.0, "step": 60278 }, { "epoch": 1.17, "learning_rate": 3.0507410929746588e-05, "loss": 0.0, "step": 60280 }, { "epoch": 1.17, "learning_rate": 3.0506763969489354e-05, "loss": 0.0, "step": 60282 }, { "epoch": 1.17, "learning_rate": 3.0506117009232126e-05, "loss": 0.0085, "step": 60284 }, { "epoch": 1.17, "learning_rate": 3.0505470048974892e-05, "loss": 0.0, "step": 60286 }, { "epoch": 1.17, "learning_rate": 3.050482308871766e-05, "loss": 0.0089, "step": 60288 }, { "epoch": 1.17, "learning_rate": 3.0504176128460433e-05, "loss": 0.0, "step": 60290 }, { "epoch": 1.17, "learning_rate": 3.05035291682032e-05, "loss": 0.0022, "step": 60292 }, { "epoch": 1.17, "learning_rate": 3.0502882207945965e-05, "loss": 0.0, "step": 60294 }, { "epoch": 1.17, "learning_rate": 3.0502235247688737e-05, "loss": 0.0, "step": 60296 }, { "epoch": 1.17, "learning_rate": 3.0501588287431503e-05, "loss": 0.0, "step": 60298 }, { "epoch": 1.17, "learning_rate": 3.0500941327174272e-05, "loss": 0.0063, "step": 60300 }, { "epoch": 1.17, "learning_rate": 3.050029436691704e-05, "loss": 0.001, "step": 60302 }, { "epoch": 1.17, "learning_rate": 3.049964740665981e-05, "loss": 0.0001, "step": 60304 }, { "epoch": 1.17, "learning_rate": 3.0499000446402583e-05, "loss": 0.0, "step": 60306 }, { "epoch": 1.17, "learning_rate": 3.049835348614535e-05, "loss": 0.0002, "step": 60308 }, { "epoch": 1.17, "learning_rate": 3.0497706525888114e-05, "loss": 0.0, "step": 60310 }, { "epoch": 1.17, "learning_rate": 3.0497059565630887e-05, "loss": 0.0001, "step": 60312 }, { "epoch": 1.17, "learning_rate": 3.0496412605373652e-05, "loss": 0.0015, "step": 60314 }, { "epoch": 1.17, "learning_rate": 3.049576564511642e-05, "loss": 0.0, "step": 60316 }, { "epoch": 1.17, "learning_rate": 3.049511868485919e-05, "loss": 0.0, "step": 60318 }, { "epoch": 1.17, "learning_rate": 3.049447172460196e-05, "loss": 0.0, "step": 60320 }, { "epoch": 1.17, "learning_rate": 3.0493824764344725e-05, "loss": 0.0, "step": 60322 }, { "epoch": 1.17, "learning_rate": 3.0493177804087498e-05, "loss": 0.0, "step": 60324 }, { "epoch": 1.17, "learning_rate": 3.0492530843830263e-05, "loss": 0.0, "step": 60326 }, { "epoch": 1.17, "learning_rate": 3.0491883883573036e-05, "loss": 0.0, "step": 60328 }, { "epoch": 1.17, "learning_rate": 3.04912369233158e-05, "loss": 0.0, "step": 60330 }, { "epoch": 1.17, "learning_rate": 3.0490589963058567e-05, "loss": 0.0, "step": 60332 }, { "epoch": 1.17, "learning_rate": 3.048994300280134e-05, "loss": 0.0, "step": 60334 }, { "epoch": 1.17, "learning_rate": 3.048929604254411e-05, "loss": 0.0002, "step": 60336 }, { "epoch": 1.17, "learning_rate": 3.0488649082286875e-05, "loss": 0.0, "step": 60338 }, { "epoch": 1.17, "learning_rate": 3.0488002122029647e-05, "loss": 0.0, "step": 60340 }, { "epoch": 1.17, "learning_rate": 3.0487355161772413e-05, "loss": 0.0, "step": 60342 }, { "epoch": 1.17, "learning_rate": 3.0486708201515185e-05, "loss": 0.0, "step": 60344 }, { "epoch": 1.17, "learning_rate": 3.048606124125795e-05, "loss": 0.0008, "step": 60346 }, { "epoch": 1.17, "learning_rate": 3.0485414281000717e-05, "loss": 0.0, "step": 60348 }, { "epoch": 1.17, "learning_rate": 3.048476732074349e-05, "loss": 0.0008, "step": 60350 }, { "epoch": 1.17, "learning_rate": 3.0484120360486258e-05, "loss": 0.0, "step": 60352 }, { "epoch": 1.17, "learning_rate": 3.0483473400229024e-05, "loss": 0.0004, "step": 60354 }, { "epoch": 1.17, "learning_rate": 3.0482826439971796e-05, "loss": 0.0, "step": 60356 }, { "epoch": 1.17, "learning_rate": 3.0482179479714562e-05, "loss": 0.0002, "step": 60358 }, { "epoch": 1.17, "learning_rate": 3.0481532519457328e-05, "loss": 0.0, "step": 60360 }, { "epoch": 1.17, "learning_rate": 3.04808855592001e-05, "loss": 0.0, "step": 60362 }, { "epoch": 1.17, "learning_rate": 3.0480238598942866e-05, "loss": 0.0, "step": 60364 }, { "epoch": 1.17, "learning_rate": 3.047959163868564e-05, "loss": 0.0, "step": 60366 }, { "epoch": 1.17, "learning_rate": 3.0478944678428407e-05, "loss": 0.0, "step": 60368 }, { "epoch": 1.17, "learning_rate": 3.0478297718171173e-05, "loss": 0.0, "step": 60370 }, { "epoch": 1.17, "learning_rate": 3.0477650757913946e-05, "loss": 0.0, "step": 60372 }, { "epoch": 1.17, "learning_rate": 3.047700379765671e-05, "loss": 0.0082, "step": 60374 }, { "epoch": 1.17, "learning_rate": 3.0476356837399477e-05, "loss": 0.0, "step": 60376 }, { "epoch": 1.17, "learning_rate": 3.047570987714225e-05, "loss": 0.0, "step": 60378 }, { "epoch": 1.17, "learning_rate": 3.0475062916885015e-05, "loss": 0.0, "step": 60380 }, { "epoch": 1.17, "learning_rate": 3.0474415956627784e-05, "loss": 0.0009, "step": 60382 }, { "epoch": 1.17, "learning_rate": 3.0473768996370557e-05, "loss": 0.0005, "step": 60384 }, { "epoch": 1.17, "learning_rate": 3.0473122036113322e-05, "loss": 0.0, "step": 60386 }, { "epoch": 1.17, "learning_rate": 3.0472475075856095e-05, "loss": 0.0, "step": 60388 }, { "epoch": 1.17, "learning_rate": 3.047182811559886e-05, "loss": 0.0, "step": 60390 }, { "epoch": 1.17, "learning_rate": 3.0471181155341626e-05, "loss": 0.0, "step": 60392 }, { "epoch": 1.17, "learning_rate": 3.04705341950844e-05, "loss": 0.0019, "step": 60394 }, { "epoch": 1.17, "learning_rate": 3.0469887234827164e-05, "loss": 0.0, "step": 60396 }, { "epoch": 1.17, "learning_rate": 3.0469240274569934e-05, "loss": 0.0, "step": 60398 }, { "epoch": 1.17, "learning_rate": 3.0468593314312706e-05, "loss": 0.0, "step": 60400 }, { "epoch": 1.17, "learning_rate": 3.046794635405547e-05, "loss": 0.0, "step": 60402 }, { "epoch": 1.17, "learning_rate": 3.0467299393798244e-05, "loss": 0.0, "step": 60404 }, { "epoch": 1.17, "learning_rate": 3.046665243354101e-05, "loss": 0.0001, "step": 60406 }, { "epoch": 1.17, "learning_rate": 3.0466005473283776e-05, "loss": 0.0, "step": 60408 }, { "epoch": 1.17, "learning_rate": 3.0465358513026548e-05, "loss": 0.0, "step": 60410 }, { "epoch": 1.17, "learning_rate": 3.0464711552769314e-05, "loss": 0.0, "step": 60412 }, { "epoch": 1.17, "learning_rate": 3.0464064592512083e-05, "loss": 0.0, "step": 60414 }, { "epoch": 1.17, "learning_rate": 3.0463417632254855e-05, "loss": 0.0, "step": 60416 }, { "epoch": 1.17, "learning_rate": 3.046277067199762e-05, "loss": 0.0002, "step": 60418 }, { "epoch": 1.17, "learning_rate": 3.0462123711740387e-05, "loss": 0.0001, "step": 60420 }, { "epoch": 1.17, "learning_rate": 3.046147675148316e-05, "loss": 0.0014, "step": 60422 }, { "epoch": 1.17, "learning_rate": 3.0460829791225925e-05, "loss": 0.0, "step": 60424 }, { "epoch": 1.17, "learning_rate": 3.0460182830968697e-05, "loss": 0.0, "step": 60426 }, { "epoch": 1.17, "learning_rate": 3.0459535870711463e-05, "loss": 0.0, "step": 60428 }, { "epoch": 1.17, "learning_rate": 3.0458888910454232e-05, "loss": 0.0052, "step": 60430 }, { "epoch": 1.17, "learning_rate": 3.0458241950197005e-05, "loss": 0.0, "step": 60432 }, { "epoch": 1.17, "learning_rate": 3.045759498993977e-05, "loss": 0.0, "step": 60434 }, { "epoch": 1.17, "learning_rate": 3.0456948029682536e-05, "loss": 0.0, "step": 60436 }, { "epoch": 1.17, "learning_rate": 3.045630106942531e-05, "loss": 0.0, "step": 60438 }, { "epoch": 1.17, "learning_rate": 3.0455654109168074e-05, "loss": 0.012, "step": 60440 }, { "epoch": 1.17, "learning_rate": 3.045500714891084e-05, "loss": 0.0, "step": 60442 }, { "epoch": 1.17, "learning_rate": 3.0454360188653612e-05, "loss": 0.0009, "step": 60444 }, { "epoch": 1.17, "learning_rate": 3.045371322839638e-05, "loss": 0.0001, "step": 60446 }, { "epoch": 1.17, "learning_rate": 3.045306626813915e-05, "loss": 0.0, "step": 60448 }, { "epoch": 1.17, "learning_rate": 3.045241930788192e-05, "loss": 0.0013, "step": 60450 }, { "epoch": 1.17, "learning_rate": 3.0451772347624685e-05, "loss": 0.0, "step": 60452 }, { "epoch": 1.17, "learning_rate": 3.0451125387367458e-05, "loss": 0.0023, "step": 60454 }, { "epoch": 1.17, "learning_rate": 3.0450478427110223e-05, "loss": 0.0013, "step": 60456 }, { "epoch": 1.17, "learning_rate": 3.044983146685299e-05, "loss": 0.0, "step": 60458 }, { "epoch": 1.17, "learning_rate": 3.044918450659576e-05, "loss": 0.0, "step": 60460 }, { "epoch": 1.17, "learning_rate": 3.044853754633853e-05, "loss": 0.0, "step": 60462 }, { "epoch": 1.17, "learning_rate": 3.0447890586081296e-05, "loss": 0.0003, "step": 60464 }, { "epoch": 1.17, "learning_rate": 3.044724362582407e-05, "loss": 0.0, "step": 60466 }, { "epoch": 1.17, "learning_rate": 3.0446596665566835e-05, "loss": 0.0, "step": 60468 }, { "epoch": 1.17, "learning_rate": 3.0445949705309607e-05, "loss": 0.0032, "step": 60470 }, { "epoch": 1.17, "learning_rate": 3.0445302745052373e-05, "loss": 0.0, "step": 60472 }, { "epoch": 1.17, "learning_rate": 3.044465578479514e-05, "loss": 0.0, "step": 60474 }, { "epoch": 1.17, "learning_rate": 3.044400882453791e-05, "loss": 0.0001, "step": 60476 }, { "epoch": 1.17, "learning_rate": 3.044336186428068e-05, "loss": 0.0, "step": 60478 }, { "epoch": 1.17, "learning_rate": 3.0442714904023446e-05, "loss": 0.0, "step": 60480 }, { "epoch": 1.17, "learning_rate": 3.0442067943766218e-05, "loss": 0.0, "step": 60482 }, { "epoch": 1.17, "learning_rate": 3.0441420983508984e-05, "loss": 0.0, "step": 60484 }, { "epoch": 1.17, "learning_rate": 3.0440774023251756e-05, "loss": 0.0, "step": 60486 }, { "epoch": 1.17, "learning_rate": 3.0440127062994522e-05, "loss": 0.0, "step": 60488 }, { "epoch": 1.17, "learning_rate": 3.0439480102737288e-05, "loss": 0.0, "step": 60490 }, { "epoch": 1.17, "learning_rate": 3.043883314248006e-05, "loss": 0.0, "step": 60492 }, { "epoch": 1.17, "learning_rate": 3.043818618222283e-05, "loss": 0.0, "step": 60494 }, { "epoch": 1.17, "learning_rate": 3.0437539221965595e-05, "loss": 0.0, "step": 60496 }, { "epoch": 1.17, "learning_rate": 3.0436892261708367e-05, "loss": 0.0, "step": 60498 }, { "epoch": 1.17, "learning_rate": 3.0436245301451133e-05, "loss": 0.0, "step": 60500 }, { "epoch": 1.17, "learning_rate": 3.04355983411939e-05, "loss": 0.0, "step": 60502 }, { "epoch": 1.17, "learning_rate": 3.043495138093667e-05, "loss": 0.0, "step": 60504 }, { "epoch": 1.17, "learning_rate": 3.0434304420679437e-05, "loss": 0.0, "step": 60506 }, { "epoch": 1.17, "learning_rate": 3.043365746042221e-05, "loss": 0.0, "step": 60508 }, { "epoch": 1.17, "learning_rate": 3.0433010500164975e-05, "loss": 0.0, "step": 60510 }, { "epoch": 1.17, "learning_rate": 3.0432363539907744e-05, "loss": 0.0001, "step": 60512 }, { "epoch": 1.17, "learning_rate": 3.0431716579650517e-05, "loss": 0.002, "step": 60514 }, { "epoch": 1.17, "learning_rate": 3.0431069619393282e-05, "loss": 0.0001, "step": 60516 }, { "epoch": 1.17, "learning_rate": 3.0430422659136048e-05, "loss": 0.0, "step": 60518 }, { "epoch": 1.17, "learning_rate": 3.042977569887882e-05, "loss": 0.0, "step": 60520 }, { "epoch": 1.17, "learning_rate": 3.0429128738621586e-05, "loss": 0.0, "step": 60522 }, { "epoch": 1.17, "learning_rate": 3.0428481778364355e-05, "loss": 0.0, "step": 60524 }, { "epoch": 1.17, "learning_rate": 3.0427834818107124e-05, "loss": 0.0, "step": 60526 }, { "epoch": 1.17, "learning_rate": 3.0427187857849894e-05, "loss": 0.0, "step": 60528 }, { "epoch": 1.17, "learning_rate": 3.0426540897592666e-05, "loss": 0.0, "step": 60530 }, { "epoch": 1.17, "learning_rate": 3.042589393733543e-05, "loss": 0.0, "step": 60532 }, { "epoch": 1.17, "learning_rate": 3.0425246977078197e-05, "loss": 0.0, "step": 60534 }, { "epoch": 1.17, "learning_rate": 3.042460001682097e-05, "loss": 0.0, "step": 60536 }, { "epoch": 1.17, "learning_rate": 3.0423953056563736e-05, "loss": 0.0, "step": 60538 }, { "epoch": 1.18, "learning_rate": 3.0423306096306505e-05, "loss": 0.0, "step": 60540 }, { "epoch": 1.18, "learning_rate": 3.0422659136049274e-05, "loss": 0.0003, "step": 60542 }, { "epoch": 1.18, "learning_rate": 3.0422012175792043e-05, "loss": 0.0, "step": 60544 }, { "epoch": 1.18, "learning_rate": 3.0421365215534815e-05, "loss": 0.0004, "step": 60546 }, { "epoch": 1.18, "learning_rate": 3.042071825527758e-05, "loss": 0.0027, "step": 60548 }, { "epoch": 1.18, "learning_rate": 3.0420071295020347e-05, "loss": 0.0, "step": 60550 }, { "epoch": 1.18, "learning_rate": 3.041942433476312e-05, "loss": 0.0, "step": 60552 }, { "epoch": 1.18, "learning_rate": 3.0418777374505885e-05, "loss": 0.0, "step": 60554 }, { "epoch": 1.18, "learning_rate": 3.0418130414248654e-05, "loss": 0.0, "step": 60556 }, { "epoch": 1.18, "learning_rate": 3.0417483453991423e-05, "loss": 0.0, "step": 60558 }, { "epoch": 1.18, "learning_rate": 3.0416836493734192e-05, "loss": 0.0, "step": 60560 }, { "epoch": 1.18, "learning_rate": 3.0416189533476958e-05, "loss": 0.0, "step": 60562 }, { "epoch": 1.18, "learning_rate": 3.041554257321973e-05, "loss": 0.0, "step": 60564 }, { "epoch": 1.18, "learning_rate": 3.0414895612962496e-05, "loss": 0.0, "step": 60566 }, { "epoch": 1.18, "learning_rate": 3.041424865270527e-05, "loss": 0.0002, "step": 60568 }, { "epoch": 1.18, "learning_rate": 3.0413601692448034e-05, "loss": 0.0171, "step": 60570 }, { "epoch": 1.18, "learning_rate": 3.04129547321908e-05, "loss": 0.0018, "step": 60572 }, { "epoch": 1.18, "learning_rate": 3.0412307771933572e-05, "loss": 0.0001, "step": 60574 }, { "epoch": 1.18, "learning_rate": 3.041166081167634e-05, "loss": 0.0, "step": 60576 }, { "epoch": 1.18, "learning_rate": 3.0411013851419107e-05, "loss": 0.0014, "step": 60578 }, { "epoch": 1.18, "learning_rate": 3.041036689116188e-05, "loss": 0.0, "step": 60580 }, { "epoch": 1.18, "learning_rate": 3.0409719930904645e-05, "loss": 0.0002, "step": 60582 }, { "epoch": 1.18, "learning_rate": 3.040907297064741e-05, "loss": 0.0, "step": 60584 }, { "epoch": 1.18, "learning_rate": 3.0408426010390183e-05, "loss": 0.0, "step": 60586 }, { "epoch": 1.18, "learning_rate": 3.040777905013295e-05, "loss": 0.0, "step": 60588 }, { "epoch": 1.18, "learning_rate": 3.040713208987572e-05, "loss": 0.0077, "step": 60590 }, { "epoch": 1.18, "learning_rate": 3.040648512961849e-05, "loss": 0.0, "step": 60592 }, { "epoch": 1.18, "learning_rate": 3.0405838169361256e-05, "loss": 0.0, "step": 60594 }, { "epoch": 1.18, "learning_rate": 3.040519120910403e-05, "loss": 0.0013, "step": 60596 }, { "epoch": 1.18, "learning_rate": 3.0404544248846795e-05, "loss": 0.0004, "step": 60598 }, { "epoch": 1.18, "learning_rate": 3.040389728858956e-05, "loss": 0.0004, "step": 60600 }, { "epoch": 1.18, "learning_rate": 3.0403250328332333e-05, "loss": 0.0003, "step": 60602 }, { "epoch": 1.18, "learning_rate": 3.04026033680751e-05, "loss": 0.0, "step": 60604 }, { "epoch": 1.18, "learning_rate": 3.0401956407817867e-05, "loss": 0.0, "step": 60606 }, { "epoch": 1.18, "learning_rate": 3.040130944756064e-05, "loss": 0.0, "step": 60608 }, { "epoch": 1.18, "learning_rate": 3.0400662487303406e-05, "loss": 0.0, "step": 60610 }, { "epoch": 1.18, "learning_rate": 3.0400015527046178e-05, "loss": 0.0659, "step": 60612 }, { "epoch": 1.18, "learning_rate": 3.0399368566788944e-05, "loss": 0.0185, "step": 60614 }, { "epoch": 1.18, "learning_rate": 3.039872160653171e-05, "loss": 0.0001, "step": 60616 }, { "epoch": 1.18, "learning_rate": 3.0398074646274482e-05, "loss": 0.0, "step": 60618 }, { "epoch": 1.18, "learning_rate": 3.0397427686017248e-05, "loss": 0.0, "step": 60620 }, { "epoch": 1.18, "learning_rate": 3.0396780725760017e-05, "loss": 0.0, "step": 60622 }, { "epoch": 1.18, "learning_rate": 3.039613376550279e-05, "loss": 0.0, "step": 60624 }, { "epoch": 1.18, "learning_rate": 3.0395486805245555e-05, "loss": 0.008, "step": 60626 }, { "epoch": 1.18, "learning_rate": 3.0394839844988327e-05, "loss": 0.0002, "step": 60628 }, { "epoch": 1.18, "learning_rate": 3.0394192884731093e-05, "loss": 0.0, "step": 60630 }, { "epoch": 1.18, "learning_rate": 3.039354592447386e-05, "loss": 0.0, "step": 60632 }, { "epoch": 1.18, "learning_rate": 3.039289896421663e-05, "loss": 0.0001, "step": 60634 }, { "epoch": 1.18, "learning_rate": 3.0392252003959397e-05, "loss": 0.0, "step": 60636 }, { "epoch": 1.18, "learning_rate": 3.0391605043702166e-05, "loss": 0.0058, "step": 60638 }, { "epoch": 1.18, "learning_rate": 3.039095808344494e-05, "loss": 0.0, "step": 60640 }, { "epoch": 1.18, "learning_rate": 3.0390311123187704e-05, "loss": 0.0, "step": 60642 }, { "epoch": 1.18, "learning_rate": 3.038966416293047e-05, "loss": 0.0011, "step": 60644 }, { "epoch": 1.18, "learning_rate": 3.0389017202673242e-05, "loss": 0.0009, "step": 60646 }, { "epoch": 1.18, "learning_rate": 3.0388370242416008e-05, "loss": 0.0, "step": 60648 }, { "epoch": 1.18, "learning_rate": 3.038772328215878e-05, "loss": 0.0001, "step": 60650 }, { "epoch": 1.18, "learning_rate": 3.0387076321901546e-05, "loss": 0.0, "step": 60652 }, { "epoch": 1.18, "learning_rate": 3.0386429361644315e-05, "loss": 0.0, "step": 60654 }, { "epoch": 1.18, "learning_rate": 3.0385782401387088e-05, "loss": 0.0007, "step": 60656 }, { "epoch": 1.18, "learning_rate": 3.0385135441129854e-05, "loss": 0.0028, "step": 60658 }, { "epoch": 1.18, "learning_rate": 3.038448848087262e-05, "loss": 0.0, "step": 60660 }, { "epoch": 1.18, "learning_rate": 3.038384152061539e-05, "loss": 0.0022, "step": 60662 }, { "epoch": 1.18, "learning_rate": 3.0383194560358157e-05, "loss": 0.0008, "step": 60664 }, { "epoch": 1.18, "learning_rate": 3.0382547600100923e-05, "loss": 0.0, "step": 60666 }, { "epoch": 1.18, "learning_rate": 3.0381900639843696e-05, "loss": 0.0, "step": 60668 }, { "epoch": 1.18, "learning_rate": 3.0381253679586465e-05, "loss": 0.0, "step": 60670 }, { "epoch": 1.18, "learning_rate": 3.0380606719329237e-05, "loss": 0.0, "step": 60672 }, { "epoch": 1.18, "learning_rate": 3.0379959759072003e-05, "loss": 0.0, "step": 60674 }, { "epoch": 1.18, "learning_rate": 3.037931279881477e-05, "loss": 0.0, "step": 60676 }, { "epoch": 1.18, "learning_rate": 3.037866583855754e-05, "loss": 0.0003, "step": 60678 }, { "epoch": 1.18, "learning_rate": 3.0378018878300307e-05, "loss": 0.0065, "step": 60680 }, { "epoch": 1.18, "learning_rate": 3.0377371918043072e-05, "loss": 0.0002, "step": 60682 }, { "epoch": 1.18, "learning_rate": 3.0376724957785845e-05, "loss": 0.0, "step": 60684 }, { "epoch": 1.18, "learning_rate": 3.0376077997528614e-05, "loss": 0.0, "step": 60686 }, { "epoch": 1.18, "learning_rate": 3.0375431037271383e-05, "loss": 0.0, "step": 60688 }, { "epoch": 1.18, "learning_rate": 3.0374784077014152e-05, "loss": 0.0121, "step": 60690 }, { "epoch": 1.18, "learning_rate": 3.0374137116756918e-05, "loss": 0.0, "step": 60692 }, { "epoch": 1.18, "learning_rate": 3.037349015649969e-05, "loss": 0.0032, "step": 60694 }, { "epoch": 1.18, "learning_rate": 3.0372843196242456e-05, "loss": 0.0, "step": 60696 }, { "epoch": 1.18, "learning_rate": 3.037219623598522e-05, "loss": 0.0, "step": 60698 }, { "epoch": 1.18, "learning_rate": 3.0371549275727994e-05, "loss": 0.0, "step": 60700 }, { "epoch": 1.18, "learning_rate": 3.0370902315470763e-05, "loss": 0.0001, "step": 60702 }, { "epoch": 1.18, "learning_rate": 3.037025535521353e-05, "loss": 0.0, "step": 60704 }, { "epoch": 1.18, "learning_rate": 3.03696083949563e-05, "loss": 0.0047, "step": 60706 }, { "epoch": 1.18, "learning_rate": 3.0368961434699067e-05, "loss": 0.0, "step": 60708 }, { "epoch": 1.18, "learning_rate": 3.036831447444184e-05, "loss": 0.0001, "step": 60710 }, { "epoch": 1.18, "learning_rate": 3.0367667514184605e-05, "loss": 0.0, "step": 60712 }, { "epoch": 1.18, "learning_rate": 3.036702055392737e-05, "loss": 0.0, "step": 60714 }, { "epoch": 1.18, "learning_rate": 3.0366373593670143e-05, "loss": 0.0, "step": 60716 }, { "epoch": 1.18, "learning_rate": 3.0365726633412912e-05, "loss": 0.0, "step": 60718 }, { "epoch": 1.18, "learning_rate": 3.0365079673155678e-05, "loss": 0.0008, "step": 60720 }, { "epoch": 1.18, "learning_rate": 3.036443271289845e-05, "loss": 0.0, "step": 60722 }, { "epoch": 1.18, "learning_rate": 3.0363785752641216e-05, "loss": 0.0, "step": 60724 }, { "epoch": 1.18, "learning_rate": 3.0363138792383982e-05, "loss": 0.0, "step": 60726 }, { "epoch": 1.18, "learning_rate": 3.0362491832126755e-05, "loss": 0.0, "step": 60728 }, { "epoch": 1.18, "learning_rate": 3.036184487186952e-05, "loss": 0.0001, "step": 60730 }, { "epoch": 1.18, "learning_rate": 3.0361197911612293e-05, "loss": 0.0001, "step": 60732 }, { "epoch": 1.18, "learning_rate": 3.0360550951355062e-05, "loss": 0.0003, "step": 60734 }, { "epoch": 1.18, "learning_rate": 3.0359903991097827e-05, "loss": 0.0, "step": 60736 }, { "epoch": 1.18, "learning_rate": 3.03592570308406e-05, "loss": 0.0, "step": 60738 }, { "epoch": 1.18, "learning_rate": 3.0358610070583366e-05, "loss": 0.0, "step": 60740 }, { "epoch": 1.18, "learning_rate": 3.035796311032613e-05, "loss": 0.0, "step": 60742 }, { "epoch": 1.18, "learning_rate": 3.0357316150068904e-05, "loss": 0.005, "step": 60744 }, { "epoch": 1.18, "learning_rate": 3.035666918981167e-05, "loss": 0.0, "step": 60746 }, { "epoch": 1.18, "learning_rate": 3.035602222955444e-05, "loss": 0.0004, "step": 60748 }, { "epoch": 1.18, "learning_rate": 3.0355375269297208e-05, "loss": 0.0001, "step": 60750 }, { "epoch": 1.18, "learning_rate": 3.0354728309039977e-05, "loss": 0.0, "step": 60752 }, { "epoch": 1.18, "learning_rate": 3.035408134878275e-05, "loss": 0.0, "step": 60754 }, { "epoch": 1.18, "learning_rate": 3.0353434388525515e-05, "loss": 0.0, "step": 60756 }, { "epoch": 1.18, "learning_rate": 3.035278742826828e-05, "loss": 0.0, "step": 60758 }, { "epoch": 1.18, "learning_rate": 3.0352140468011053e-05, "loss": 0.0001, "step": 60760 }, { "epoch": 1.18, "learning_rate": 3.035149350775382e-05, "loss": 0.0002, "step": 60762 }, { "epoch": 1.18, "learning_rate": 3.0350846547496588e-05, "loss": 0.0, "step": 60764 }, { "epoch": 1.18, "learning_rate": 3.0350199587239357e-05, "loss": 0.0001, "step": 60766 }, { "epoch": 1.18, "learning_rate": 3.0349552626982126e-05, "loss": 0.0, "step": 60768 }, { "epoch": 1.18, "learning_rate": 3.03489056667249e-05, "loss": 0.0001, "step": 60770 }, { "epoch": 1.18, "learning_rate": 3.0348258706467664e-05, "loss": 0.0, "step": 60772 }, { "epoch": 1.18, "learning_rate": 3.034761174621043e-05, "loss": 0.0, "step": 60774 }, { "epoch": 1.18, "learning_rate": 3.0346964785953202e-05, "loss": 0.0, "step": 60776 }, { "epoch": 1.18, "learning_rate": 3.0346317825695968e-05, "loss": 0.0021, "step": 60778 }, { "epoch": 1.18, "learning_rate": 3.0345670865438737e-05, "loss": 0.0001, "step": 60780 }, { "epoch": 1.18, "learning_rate": 3.0345023905181506e-05, "loss": 0.0, "step": 60782 }, { "epoch": 1.18, "learning_rate": 3.0344376944924275e-05, "loss": 0.0, "step": 60784 }, { "epoch": 1.18, "learning_rate": 3.034372998466704e-05, "loss": 0.0007, "step": 60786 }, { "epoch": 1.18, "learning_rate": 3.0343083024409813e-05, "loss": 0.0, "step": 60788 }, { "epoch": 1.18, "learning_rate": 3.034243606415258e-05, "loss": 0.0001, "step": 60790 }, { "epoch": 1.18, "learning_rate": 3.034178910389535e-05, "loss": 0.0, "step": 60792 }, { "epoch": 1.18, "learning_rate": 3.0341142143638117e-05, "loss": 0.0, "step": 60794 }, { "epoch": 1.18, "learning_rate": 3.0340495183380886e-05, "loss": 0.0, "step": 60796 }, { "epoch": 1.18, "learning_rate": 3.0339848223123656e-05, "loss": 0.0, "step": 60798 }, { "epoch": 1.18, "learning_rate": 3.0339201262866425e-05, "loss": 0.0, "step": 60800 }, { "epoch": 1.18, "learning_rate": 3.033855430260919e-05, "loss": 0.0, "step": 60802 }, { "epoch": 1.18, "learning_rate": 3.0337907342351963e-05, "loss": 0.0, "step": 60804 }, { "epoch": 1.18, "learning_rate": 3.033726038209473e-05, "loss": 0.0, "step": 60806 }, { "epoch": 1.18, "learning_rate": 3.0336613421837494e-05, "loss": 0.0001, "step": 60808 }, { "epoch": 1.18, "learning_rate": 3.0335966461580267e-05, "loss": 0.0, "step": 60810 }, { "epoch": 1.18, "learning_rate": 3.0335319501323032e-05, "loss": 0.0004, "step": 60812 }, { "epoch": 1.18, "learning_rate": 3.0334672541065805e-05, "loss": 0.0, "step": 60814 }, { "epoch": 1.18, "learning_rate": 3.0334025580808574e-05, "loss": 0.0, "step": 60816 }, { "epoch": 1.18, "learning_rate": 3.033337862055134e-05, "loss": 0.0, "step": 60818 }, { "epoch": 1.18, "learning_rate": 3.0332731660294112e-05, "loss": 0.0, "step": 60820 }, { "epoch": 1.18, "learning_rate": 3.0332084700036878e-05, "loss": 0.0, "step": 60822 }, { "epoch": 1.18, "learning_rate": 3.0331437739779643e-05, "loss": 0.0002, "step": 60824 }, { "epoch": 1.18, "learning_rate": 3.0330790779522416e-05, "loss": 0.0, "step": 60826 }, { "epoch": 1.18, "learning_rate": 3.033014381926518e-05, "loss": 0.0, "step": 60828 }, { "epoch": 1.18, "learning_rate": 3.0329496859007954e-05, "loss": 0.0, "step": 60830 }, { "epoch": 1.18, "learning_rate": 3.0328849898750723e-05, "loss": 0.0, "step": 60832 }, { "epoch": 1.18, "learning_rate": 3.032820293849349e-05, "loss": 0.0, "step": 60834 }, { "epoch": 1.18, "learning_rate": 3.032755597823626e-05, "loss": 0.0001, "step": 60836 }, { "epoch": 1.18, "learning_rate": 3.0326909017979027e-05, "loss": 0.0324, "step": 60838 }, { "epoch": 1.18, "learning_rate": 3.0326262057721793e-05, "loss": 0.0, "step": 60840 }, { "epoch": 1.18, "learning_rate": 3.0325615097464565e-05, "loss": 0.0, "step": 60842 }, { "epoch": 1.18, "learning_rate": 3.032496813720733e-05, "loss": 0.0, "step": 60844 }, { "epoch": 1.18, "learning_rate": 3.03243211769501e-05, "loss": 0.0, "step": 60846 }, { "epoch": 1.18, "learning_rate": 3.0323674216692872e-05, "loss": 0.0, "step": 60848 }, { "epoch": 1.18, "learning_rate": 3.0323027256435638e-05, "loss": 0.0, "step": 60850 }, { "epoch": 1.18, "learning_rate": 3.032238029617841e-05, "loss": 0.0, "step": 60852 }, { "epoch": 1.18, "learning_rate": 3.0321733335921176e-05, "loss": 0.0, "step": 60854 }, { "epoch": 1.18, "learning_rate": 3.0321086375663942e-05, "loss": 0.0019, "step": 60856 }, { "epoch": 1.18, "learning_rate": 3.0320439415406715e-05, "loss": 0.0, "step": 60858 }, { "epoch": 1.18, "learning_rate": 3.031979245514948e-05, "loss": 0.0, "step": 60860 }, { "epoch": 1.18, "learning_rate": 3.031914549489225e-05, "loss": 0.0, "step": 60862 }, { "epoch": 1.18, "learning_rate": 3.0318498534635022e-05, "loss": 0.0, "step": 60864 }, { "epoch": 1.18, "learning_rate": 3.0317851574377787e-05, "loss": 0.0, "step": 60866 }, { "epoch": 1.18, "learning_rate": 3.0317204614120553e-05, "loss": 0.0, "step": 60868 }, { "epoch": 1.18, "learning_rate": 3.0316557653863326e-05, "loss": 0.0, "step": 60870 }, { "epoch": 1.18, "learning_rate": 3.031591069360609e-05, "loss": 0.0, "step": 60872 }, { "epoch": 1.18, "learning_rate": 3.0315263733348864e-05, "loss": 0.0, "step": 60874 }, { "epoch": 1.18, "learning_rate": 3.031461677309163e-05, "loss": 0.0001, "step": 60876 }, { "epoch": 1.18, "learning_rate": 3.03139698128344e-05, "loss": 0.0, "step": 60878 }, { "epoch": 1.18, "learning_rate": 3.031332285257717e-05, "loss": 0.0, "step": 60880 }, { "epoch": 1.18, "learning_rate": 3.0312675892319937e-05, "loss": 0.0, "step": 60882 }, { "epoch": 1.18, "learning_rate": 3.0312028932062702e-05, "loss": 0.0, "step": 60884 }, { "epoch": 1.18, "learning_rate": 3.0311381971805475e-05, "loss": 0.0, "step": 60886 }, { "epoch": 1.18, "learning_rate": 3.031073501154824e-05, "loss": 0.0, "step": 60888 }, { "epoch": 1.18, "learning_rate": 3.0310088051291006e-05, "loss": 0.0, "step": 60890 }, { "epoch": 1.18, "learning_rate": 3.030944109103378e-05, "loss": 0.0, "step": 60892 }, { "epoch": 1.18, "learning_rate": 3.0308794130776548e-05, "loss": 0.0, "step": 60894 }, { "epoch": 1.18, "learning_rate": 3.030814717051932e-05, "loss": 0.0, "step": 60896 }, { "epoch": 1.18, "learning_rate": 3.0307500210262086e-05, "loss": 0.0042, "step": 60898 }, { "epoch": 1.18, "learning_rate": 3.0306853250004852e-05, "loss": 0.0, "step": 60900 }, { "epoch": 1.18, "learning_rate": 3.0306206289747624e-05, "loss": 0.0071, "step": 60902 }, { "epoch": 1.18, "learning_rate": 3.030555932949039e-05, "loss": 0.0032, "step": 60904 }, { "epoch": 1.18, "learning_rate": 3.0304912369233156e-05, "loss": 0.0, "step": 60906 }, { "epoch": 1.18, "learning_rate": 3.0304265408975928e-05, "loss": 0.0001, "step": 60908 }, { "epoch": 1.18, "learning_rate": 3.0303618448718697e-05, "loss": 0.0051, "step": 60910 }, { "epoch": 1.18, "learning_rate": 3.030297148846147e-05, "loss": 0.0, "step": 60912 }, { "epoch": 1.18, "learning_rate": 3.0302324528204235e-05, "loss": 0.0, "step": 60914 }, { "epoch": 1.18, "learning_rate": 3.0301677567947e-05, "loss": 0.0, "step": 60916 }, { "epoch": 1.18, "learning_rate": 3.0301030607689773e-05, "loss": 0.0, "step": 60918 }, { "epoch": 1.18, "learning_rate": 3.030038364743254e-05, "loss": 0.0, "step": 60920 }, { "epoch": 1.18, "learning_rate": 3.0299736687175305e-05, "loss": 0.0004, "step": 60922 }, { "epoch": 1.18, "learning_rate": 3.0299089726918077e-05, "loss": 0.0, "step": 60924 }, { "epoch": 1.18, "learning_rate": 3.0298442766660846e-05, "loss": 0.0, "step": 60926 }, { "epoch": 1.18, "learning_rate": 3.0297795806403612e-05, "loss": 0.0, "step": 60928 }, { "epoch": 1.18, "learning_rate": 3.0297148846146385e-05, "loss": 0.0, "step": 60930 }, { "epoch": 1.18, "learning_rate": 3.029650188588915e-05, "loss": 0.0, "step": 60932 }, { "epoch": 1.18, "learning_rate": 3.0295854925631923e-05, "loss": 0.0001, "step": 60934 }, { "epoch": 1.18, "learning_rate": 3.029520796537469e-05, "loss": 0.0, "step": 60936 }, { "epoch": 1.18, "learning_rate": 3.0294561005117454e-05, "loss": 0.0001, "step": 60938 }, { "epoch": 1.18, "learning_rate": 3.0293914044860227e-05, "loss": 0.0, "step": 60940 }, { "epoch": 1.18, "learning_rate": 3.0293267084602996e-05, "loss": 0.0, "step": 60942 }, { "epoch": 1.18, "learning_rate": 3.029262012434576e-05, "loss": 0.0, "step": 60944 }, { "epoch": 1.18, "learning_rate": 3.0291973164088534e-05, "loss": 0.0, "step": 60946 }, { "epoch": 1.18, "learning_rate": 3.02913262038313e-05, "loss": 0.0001, "step": 60948 }, { "epoch": 1.18, "learning_rate": 3.0290679243574065e-05, "loss": 0.0, "step": 60950 }, { "epoch": 1.18, "learning_rate": 3.0290032283316838e-05, "loss": 0.0, "step": 60952 }, { "epoch": 1.18, "learning_rate": 3.0289385323059603e-05, "loss": 0.0, "step": 60954 }, { "epoch": 1.18, "learning_rate": 3.0288738362802376e-05, "loss": 0.0, "step": 60956 }, { "epoch": 1.18, "learning_rate": 3.0288091402545145e-05, "loss": 0.0, "step": 60958 }, { "epoch": 1.18, "learning_rate": 3.028744444228791e-05, "loss": 0.0001, "step": 60960 }, { "epoch": 1.18, "learning_rate": 3.0286797482030683e-05, "loss": 0.0, "step": 60962 }, { "epoch": 1.18, "learning_rate": 3.028615052177345e-05, "loss": 0.0, "step": 60964 }, { "epoch": 1.18, "learning_rate": 3.0285503561516215e-05, "loss": 0.0, "step": 60966 }, { "epoch": 1.18, "learning_rate": 3.0284856601258987e-05, "loss": 0.0001, "step": 60968 }, { "epoch": 1.18, "learning_rate": 3.0284209641001753e-05, "loss": 0.0004, "step": 60970 }, { "epoch": 1.18, "learning_rate": 3.0283562680744525e-05, "loss": 0.0, "step": 60972 }, { "epoch": 1.18, "learning_rate": 3.0282915720487294e-05, "loss": 0.0, "step": 60974 }, { "epoch": 1.18, "learning_rate": 3.028226876023006e-05, "loss": 0.0, "step": 60976 }, { "epoch": 1.18, "learning_rate": 3.0281621799972832e-05, "loss": 0.0, "step": 60978 }, { "epoch": 1.18, "learning_rate": 3.0280974839715598e-05, "loss": 0.0, "step": 60980 }, { "epoch": 1.18, "learning_rate": 3.0280327879458364e-05, "loss": 0.0, "step": 60982 }, { "epoch": 1.18, "learning_rate": 3.0279680919201136e-05, "loss": 0.0018, "step": 60984 }, { "epoch": 1.18, "learning_rate": 3.0279033958943902e-05, "loss": 0.0, "step": 60986 }, { "epoch": 1.18, "learning_rate": 3.027838699868667e-05, "loss": 0.0146, "step": 60988 }, { "epoch": 1.18, "learning_rate": 3.027774003842944e-05, "loss": 0.0, "step": 60990 }, { "epoch": 1.18, "learning_rate": 3.027709307817221e-05, "loss": 0.0093, "step": 60992 }, { "epoch": 1.18, "learning_rate": 3.0276446117914982e-05, "loss": 0.0, "step": 60994 }, { "epoch": 1.18, "learning_rate": 3.0275799157657747e-05, "loss": 0.0, "step": 60996 }, { "epoch": 1.18, "learning_rate": 3.0275152197400513e-05, "loss": 0.0, "step": 60998 }, { "epoch": 1.18, "learning_rate": 3.0274505237143286e-05, "loss": 0.0002, "step": 61000 }, { "epoch": 1.18, "learning_rate": 3.027385827688605e-05, "loss": 0.0001, "step": 61002 }, { "epoch": 1.18, "learning_rate": 3.027321131662882e-05, "loss": 0.0, "step": 61004 }, { "epoch": 1.18, "learning_rate": 3.027256435637159e-05, "loss": 0.0144, "step": 61006 }, { "epoch": 1.18, "learning_rate": 3.027191739611436e-05, "loss": 0.0, "step": 61008 }, { "epoch": 1.18, "learning_rate": 3.0271270435857124e-05, "loss": 0.0, "step": 61010 }, { "epoch": 1.18, "learning_rate": 3.0270623475599897e-05, "loss": 0.0033, "step": 61012 }, { "epoch": 1.18, "learning_rate": 3.0269976515342662e-05, "loss": 0.0, "step": 61014 }, { "epoch": 1.18, "learning_rate": 3.0269329555085435e-05, "loss": 0.0, "step": 61016 }, { "epoch": 1.18, "learning_rate": 3.02686825948282e-05, "loss": 0.0, "step": 61018 }, { "epoch": 1.18, "learning_rate": 3.026803563457097e-05, "loss": 0.0, "step": 61020 }, { "epoch": 1.18, "learning_rate": 3.026738867431374e-05, "loss": 0.0, "step": 61022 }, { "epoch": 1.18, "learning_rate": 3.0266741714056508e-05, "loss": 0.0, "step": 61024 }, { "epoch": 1.18, "learning_rate": 3.0266094753799274e-05, "loss": 0.0, "step": 61026 }, { "epoch": 1.18, "learning_rate": 3.0265447793542046e-05, "loss": 0.0, "step": 61028 }, { "epoch": 1.18, "learning_rate": 3.0264800833284812e-05, "loss": 0.0, "step": 61030 }, { "epoch": 1.18, "learning_rate": 3.0264153873027577e-05, "loss": 0.0, "step": 61032 }, { "epoch": 1.18, "learning_rate": 3.026350691277035e-05, "loss": 0.0, "step": 61034 }, { "epoch": 1.18, "learning_rate": 3.026285995251312e-05, "loss": 0.0004, "step": 61036 }, { "epoch": 1.18, "learning_rate": 3.0262212992255888e-05, "loss": 0.0, "step": 61038 }, { "epoch": 1.18, "learning_rate": 3.0261566031998657e-05, "loss": 0.0, "step": 61040 }, { "epoch": 1.18, "learning_rate": 3.0260919071741423e-05, "loss": 0.0003, "step": 61042 }, { "epoch": 1.18, "learning_rate": 3.0260272111484195e-05, "loss": 0.0, "step": 61044 }, { "epoch": 1.18, "learning_rate": 3.025962515122696e-05, "loss": 0.0, "step": 61046 }, { "epoch": 1.18, "learning_rate": 3.0258978190969727e-05, "loss": 0.0, "step": 61048 }, { "epoch": 1.18, "learning_rate": 3.02583312307125e-05, "loss": 0.0, "step": 61050 }, { "epoch": 1.18, "learning_rate": 3.0257684270455265e-05, "loss": 0.0, "step": 61052 }, { "epoch": 1.18, "learning_rate": 3.0257037310198037e-05, "loss": 0.0, "step": 61054 }, { "epoch": 1.19, "learning_rate": 3.0256390349940806e-05, "loss": 0.0, "step": 61056 }, { "epoch": 1.19, "learning_rate": 3.0255743389683572e-05, "loss": 0.0, "step": 61058 }, { "epoch": 1.19, "learning_rate": 3.0255096429426345e-05, "loss": 0.0, "step": 61060 }, { "epoch": 1.19, "learning_rate": 3.025444946916911e-05, "loss": 0.0, "step": 61062 }, { "epoch": 1.19, "learning_rate": 3.0253802508911876e-05, "loss": 0.0003, "step": 61064 }, { "epoch": 1.19, "learning_rate": 3.025315554865465e-05, "loss": 0.0, "step": 61066 }, { "epoch": 1.19, "learning_rate": 3.0252508588397414e-05, "loss": 0.0, "step": 61068 }, { "epoch": 1.19, "learning_rate": 3.0251861628140183e-05, "loss": 0.0, "step": 61070 }, { "epoch": 1.19, "learning_rate": 3.0251214667882956e-05, "loss": 0.0019, "step": 61072 }, { "epoch": 1.19, "learning_rate": 3.025056770762572e-05, "loss": 0.0, "step": 61074 }, { "epoch": 1.19, "learning_rate": 3.0249920747368494e-05, "loss": 0.0, "step": 61076 }, { "epoch": 1.19, "learning_rate": 3.024927378711126e-05, "loss": 0.0, "step": 61078 }, { "epoch": 1.19, "learning_rate": 3.0248626826854025e-05, "loss": 0.0039, "step": 61080 }, { "epoch": 1.19, "learning_rate": 3.0247979866596798e-05, "loss": 0.0, "step": 61082 }, { "epoch": 1.19, "learning_rate": 3.0247332906339563e-05, "loss": 0.0, "step": 61084 }, { "epoch": 1.19, "learning_rate": 3.0246685946082333e-05, "loss": 0.0, "step": 61086 }, { "epoch": 1.19, "learning_rate": 3.0246038985825105e-05, "loss": 0.0, "step": 61088 }, { "epoch": 1.19, "learning_rate": 3.024539202556787e-05, "loss": 0.0001, "step": 61090 }, { "epoch": 1.19, "learning_rate": 3.0244745065310636e-05, "loss": 0.0, "step": 61092 }, { "epoch": 1.19, "learning_rate": 3.024409810505341e-05, "loss": 0.0004, "step": 61094 }, { "epoch": 1.19, "learning_rate": 3.0243451144796175e-05, "loss": 0.0005, "step": 61096 }, { "epoch": 1.19, "learning_rate": 3.0242804184538947e-05, "loss": 0.0094, "step": 61098 }, { "epoch": 1.19, "learning_rate": 3.0242157224281713e-05, "loss": 0.0, "step": 61100 }, { "epoch": 1.19, "learning_rate": 3.0241510264024482e-05, "loss": 0.0001, "step": 61102 }, { "epoch": 1.19, "learning_rate": 3.0240863303767254e-05, "loss": 0.0, "step": 61104 }, { "epoch": 1.19, "learning_rate": 3.024021634351002e-05, "loss": 0.0, "step": 61106 }, { "epoch": 1.19, "learning_rate": 3.0239569383252786e-05, "loss": 0.0, "step": 61108 }, { "epoch": 1.19, "learning_rate": 3.0238922422995558e-05, "loss": 0.0, "step": 61110 }, { "epoch": 1.19, "learning_rate": 3.0238275462738324e-05, "loss": 0.0, "step": 61112 }, { "epoch": 1.19, "learning_rate": 3.0237628502481096e-05, "loss": 0.0, "step": 61114 }, { "epoch": 1.19, "learning_rate": 3.0236981542223862e-05, "loss": 0.0011, "step": 61116 }, { "epoch": 1.19, "learning_rate": 3.023633458196663e-05, "loss": 0.0019, "step": 61118 }, { "epoch": 1.19, "learning_rate": 3.0235687621709404e-05, "loss": 0.0002, "step": 61120 }, { "epoch": 1.19, "learning_rate": 3.023504066145217e-05, "loss": 0.0, "step": 61122 }, { "epoch": 1.19, "learning_rate": 3.0234393701194935e-05, "loss": 0.0, "step": 61124 }, { "epoch": 1.19, "learning_rate": 3.0233746740937707e-05, "loss": 0.0, "step": 61126 }, { "epoch": 1.19, "learning_rate": 3.0233099780680473e-05, "loss": 0.0, "step": 61128 }, { "epoch": 1.19, "learning_rate": 3.023245282042324e-05, "loss": 0.0009, "step": 61130 }, { "epoch": 1.19, "learning_rate": 3.023180586016601e-05, "loss": 0.0, "step": 61132 }, { "epoch": 1.19, "learning_rate": 3.023115889990878e-05, "loss": 0.0, "step": 61134 }, { "epoch": 1.19, "learning_rate": 3.0230511939651553e-05, "loss": 0.0, "step": 61136 }, { "epoch": 1.19, "learning_rate": 3.022986497939432e-05, "loss": 0.0006, "step": 61138 }, { "epoch": 1.19, "learning_rate": 3.0229218019137084e-05, "loss": 0.0001, "step": 61140 }, { "epoch": 1.19, "learning_rate": 3.0228571058879857e-05, "loss": 0.0004, "step": 61142 }, { "epoch": 1.19, "learning_rate": 3.0227924098622622e-05, "loss": 0.0, "step": 61144 }, { "epoch": 1.19, "learning_rate": 3.0227277138365388e-05, "loss": 0.0, "step": 61146 }, { "epoch": 1.19, "learning_rate": 3.022663017810816e-05, "loss": 0.0, "step": 61148 }, { "epoch": 1.19, "learning_rate": 3.022598321785093e-05, "loss": 0.0001, "step": 61150 }, { "epoch": 1.19, "learning_rate": 3.0225336257593695e-05, "loss": 0.0, "step": 61152 }, { "epoch": 1.19, "learning_rate": 3.0224689297336468e-05, "loss": 0.0, "step": 61154 }, { "epoch": 1.19, "learning_rate": 3.0224042337079234e-05, "loss": 0.0, "step": 61156 }, { "epoch": 1.19, "learning_rate": 3.0223395376822006e-05, "loss": 0.0, "step": 61158 }, { "epoch": 1.19, "learning_rate": 3.0222748416564772e-05, "loss": 0.0064, "step": 61160 }, { "epoch": 1.19, "learning_rate": 3.0222101456307537e-05, "loss": 0.0013, "step": 61162 }, { "epoch": 1.19, "learning_rate": 3.022145449605031e-05, "loss": 0.0, "step": 61164 }, { "epoch": 1.19, "learning_rate": 3.022080753579308e-05, "loss": 0.0, "step": 61166 }, { "epoch": 1.19, "learning_rate": 3.0220160575535845e-05, "loss": 0.0004, "step": 61168 }, { "epoch": 1.19, "learning_rate": 3.0219513615278617e-05, "loss": 0.0, "step": 61170 }, { "epoch": 1.19, "learning_rate": 3.0218866655021383e-05, "loss": 0.0, "step": 61172 }, { "epoch": 1.19, "learning_rate": 3.021821969476415e-05, "loss": 0.0, "step": 61174 }, { "epoch": 1.19, "learning_rate": 3.021757273450692e-05, "loss": 0.0172, "step": 61176 }, { "epoch": 1.19, "learning_rate": 3.0216925774249687e-05, "loss": 0.0011, "step": 61178 }, { "epoch": 1.19, "learning_rate": 3.021627881399246e-05, "loss": 0.0, "step": 61180 }, { "epoch": 1.19, "learning_rate": 3.0215631853735228e-05, "loss": 0.0, "step": 61182 }, { "epoch": 1.19, "learning_rate": 3.0214984893477994e-05, "loss": 0.0001, "step": 61184 }, { "epoch": 1.19, "learning_rate": 3.0214337933220766e-05, "loss": 0.0, "step": 61186 }, { "epoch": 1.19, "learning_rate": 3.0213690972963532e-05, "loss": 0.0001, "step": 61188 }, { "epoch": 1.19, "learning_rate": 3.0213044012706298e-05, "loss": 0.0, "step": 61190 }, { "epoch": 1.19, "learning_rate": 3.021239705244907e-05, "loss": 0.0, "step": 61192 }, { "epoch": 1.19, "learning_rate": 3.0211750092191836e-05, "loss": 0.0004, "step": 61194 }, { "epoch": 1.19, "learning_rate": 3.021110313193461e-05, "loss": 0.0002, "step": 61196 }, { "epoch": 1.19, "learning_rate": 3.0210456171677378e-05, "loss": 0.0029, "step": 61198 }, { "epoch": 1.19, "learning_rate": 3.0209809211420143e-05, "loss": 0.0025, "step": 61200 }, { "epoch": 1.19, "learning_rate": 3.0209162251162916e-05, "loss": 0.0001, "step": 61202 }, { "epoch": 1.19, "learning_rate": 3.020851529090568e-05, "loss": 0.0069, "step": 61204 }, { "epoch": 1.19, "learning_rate": 3.0207868330648447e-05, "loss": 0.0, "step": 61206 }, { "epoch": 1.19, "learning_rate": 3.020722137039122e-05, "loss": 0.0, "step": 61208 }, { "epoch": 1.19, "learning_rate": 3.0206574410133985e-05, "loss": 0.0002, "step": 61210 }, { "epoch": 1.19, "learning_rate": 3.0205927449876754e-05, "loss": 0.0078, "step": 61212 }, { "epoch": 1.19, "learning_rate": 3.0205280489619527e-05, "loss": 0.0004, "step": 61214 }, { "epoch": 1.19, "learning_rate": 3.0204633529362293e-05, "loss": 0.0, "step": 61216 }, { "epoch": 1.19, "learning_rate": 3.0203986569105065e-05, "loss": 0.0, "step": 61218 }, { "epoch": 1.19, "learning_rate": 3.020333960884783e-05, "loss": 0.0058, "step": 61220 }, { "epoch": 1.19, "learning_rate": 3.0202692648590596e-05, "loss": 0.0, "step": 61222 }, { "epoch": 1.19, "learning_rate": 3.020204568833337e-05, "loss": 0.0, "step": 61224 }, { "epoch": 1.19, "learning_rate": 3.0201398728076135e-05, "loss": 0.0, "step": 61226 }, { "epoch": 1.19, "learning_rate": 3.0200751767818904e-05, "loss": 0.0, "step": 61228 }, { "epoch": 1.19, "learning_rate": 3.0200104807561673e-05, "loss": 0.0, "step": 61230 }, { "epoch": 1.19, "learning_rate": 3.0199457847304442e-05, "loss": 0.0, "step": 61232 }, { "epoch": 1.19, "learning_rate": 3.0198810887047208e-05, "loss": 0.0, "step": 61234 }, { "epoch": 1.19, "learning_rate": 3.019816392678998e-05, "loss": 0.0003, "step": 61236 }, { "epoch": 1.19, "learning_rate": 3.0197516966532746e-05, "loss": 0.0, "step": 61238 }, { "epoch": 1.19, "learning_rate": 3.0196870006275518e-05, "loss": 0.0, "step": 61240 }, { "epoch": 1.19, "learning_rate": 3.0196223046018284e-05, "loss": 0.0001, "step": 61242 }, { "epoch": 1.19, "learning_rate": 3.0195576085761053e-05, "loss": 0.0, "step": 61244 }, { "epoch": 1.19, "learning_rate": 3.0194929125503822e-05, "loss": 0.0, "step": 61246 }, { "epoch": 1.19, "learning_rate": 3.019428216524659e-05, "loss": 0.0, "step": 61248 }, { "epoch": 1.19, "learning_rate": 3.0193635204989357e-05, "loss": 0.0, "step": 61250 }, { "epoch": 1.19, "learning_rate": 3.019298824473213e-05, "loss": 0.0, "step": 61252 }, { "epoch": 1.19, "learning_rate": 3.0192341284474895e-05, "loss": 0.0, "step": 61254 }, { "epoch": 1.19, "learning_rate": 3.0191694324217667e-05, "loss": 0.0, "step": 61256 }, { "epoch": 1.19, "learning_rate": 3.0191047363960433e-05, "loss": 0.0026, "step": 61258 }, { "epoch": 1.19, "learning_rate": 3.0190400403703202e-05, "loss": 0.0, "step": 61260 }, { "epoch": 1.19, "learning_rate": 3.018975344344597e-05, "loss": 0.0, "step": 61262 }, { "epoch": 1.19, "learning_rate": 3.018910648318874e-05, "loss": 0.0, "step": 61264 }, { "epoch": 1.19, "learning_rate": 3.0188459522931506e-05, "loss": 0.0006, "step": 61266 }, { "epoch": 1.19, "learning_rate": 3.018781256267428e-05, "loss": 0.0014, "step": 61268 }, { "epoch": 1.19, "learning_rate": 3.0187165602417044e-05, "loss": 0.0, "step": 61270 }, { "epoch": 1.19, "learning_rate": 3.018651864215981e-05, "loss": 0.0, "step": 61272 }, { "epoch": 1.19, "learning_rate": 3.0185871681902582e-05, "loss": 0.0, "step": 61274 }, { "epoch": 1.19, "learning_rate": 3.018522472164535e-05, "loss": 0.0, "step": 61276 }, { "epoch": 1.19, "learning_rate": 3.018457776138812e-05, "loss": 0.0144, "step": 61278 }, { "epoch": 1.19, "learning_rate": 3.018393080113089e-05, "loss": 0.0001, "step": 61280 }, { "epoch": 1.19, "learning_rate": 3.0183283840873655e-05, "loss": 0.0001, "step": 61282 }, { "epoch": 1.19, "learning_rate": 3.0182636880616428e-05, "loss": 0.0, "step": 61284 }, { "epoch": 1.19, "learning_rate": 3.0181989920359194e-05, "loss": 0.0, "step": 61286 }, { "epoch": 1.19, "learning_rate": 3.018134296010196e-05, "loss": 0.0, "step": 61288 }, { "epoch": 1.19, "learning_rate": 3.0180695999844732e-05, "loss": 0.0, "step": 61290 }, { "epoch": 1.19, "learning_rate": 3.0180049039587497e-05, "loss": 0.0, "step": 61292 }, { "epoch": 1.19, "learning_rate": 3.0179402079330266e-05, "loss": 0.0, "step": 61294 }, { "epoch": 1.19, "learning_rate": 3.017875511907304e-05, "loss": 0.0, "step": 61296 }, { "epoch": 1.19, "learning_rate": 3.0178108158815805e-05, "loss": 0.0, "step": 61298 }, { "epoch": 1.19, "learning_rate": 3.0177461198558577e-05, "loss": 0.0, "step": 61300 }, { "epoch": 1.19, "learning_rate": 3.0176814238301343e-05, "loss": 0.0, "step": 61302 }, { "epoch": 1.19, "learning_rate": 3.017616727804411e-05, "loss": 0.0, "step": 61304 }, { "epoch": 1.19, "learning_rate": 3.017552031778688e-05, "loss": 0.0, "step": 61306 }, { "epoch": 1.19, "learning_rate": 3.0174873357529647e-05, "loss": 0.0, "step": 61308 }, { "epoch": 1.19, "learning_rate": 3.0174226397272416e-05, "loss": 0.0, "step": 61310 }, { "epoch": 1.19, "learning_rate": 3.0173579437015188e-05, "loss": 0.0, "step": 61312 }, { "epoch": 1.19, "learning_rate": 3.0172932476757954e-05, "loss": 0.0, "step": 61314 }, { "epoch": 1.19, "learning_rate": 3.017228551650072e-05, "loss": 0.0, "step": 61316 }, { "epoch": 1.19, "learning_rate": 3.0171638556243492e-05, "loss": 0.0, "step": 61318 }, { "epoch": 1.19, "learning_rate": 3.0170991595986258e-05, "loss": 0.0003, "step": 61320 }, { "epoch": 1.19, "learning_rate": 3.017034463572903e-05, "loss": 0.0003, "step": 61322 }, { "epoch": 1.19, "learning_rate": 3.0169697675471796e-05, "loss": 0.0016, "step": 61324 }, { "epoch": 1.19, "learning_rate": 3.0169050715214565e-05, "loss": 0.0, "step": 61326 }, { "epoch": 1.19, "learning_rate": 3.0168403754957338e-05, "loss": 0.0, "step": 61328 }, { "epoch": 1.19, "learning_rate": 3.0167756794700103e-05, "loss": 0.0, "step": 61330 }, { "epoch": 1.19, "learning_rate": 3.016710983444287e-05, "loss": 0.0, "step": 61332 }, { "epoch": 1.19, "learning_rate": 3.016646287418564e-05, "loss": 0.0, "step": 61334 }, { "epoch": 1.19, "learning_rate": 3.0165815913928407e-05, "loss": 0.0, "step": 61336 }, { "epoch": 1.19, "learning_rate": 3.016516895367118e-05, "loss": 0.0001, "step": 61338 }, { "epoch": 1.19, "learning_rate": 3.0164521993413945e-05, "loss": 0.0, "step": 61340 }, { "epoch": 1.19, "learning_rate": 3.0163875033156714e-05, "loss": 0.0, "step": 61342 }, { "epoch": 1.19, "learning_rate": 3.0163228072899487e-05, "loss": 0.0, "step": 61344 }, { "epoch": 1.19, "learning_rate": 3.0162581112642253e-05, "loss": 0.0, "step": 61346 }, { "epoch": 1.19, "learning_rate": 3.0161934152385018e-05, "loss": 0.0, "step": 61348 }, { "epoch": 1.19, "learning_rate": 3.016128719212779e-05, "loss": 0.0003, "step": 61350 }, { "epoch": 1.19, "learning_rate": 3.0160640231870556e-05, "loss": 0.0029, "step": 61352 }, { "epoch": 1.19, "learning_rate": 3.0159993271613325e-05, "loss": 0.0, "step": 61354 }, { "epoch": 1.19, "learning_rate": 3.0159346311356095e-05, "loss": 0.0, "step": 61356 }, { "epoch": 1.19, "learning_rate": 3.0158699351098864e-05, "loss": 0.0, "step": 61358 }, { "epoch": 1.19, "learning_rate": 3.0158052390841636e-05, "loss": 0.0, "step": 61360 }, { "epoch": 1.19, "learning_rate": 3.0157405430584402e-05, "loss": 0.0, "step": 61362 }, { "epoch": 1.19, "learning_rate": 3.0156758470327167e-05, "loss": 0.0, "step": 61364 }, { "epoch": 1.19, "learning_rate": 3.015611151006994e-05, "loss": 0.0, "step": 61366 }, { "epoch": 1.19, "learning_rate": 3.0155464549812706e-05, "loss": 0.0, "step": 61368 }, { "epoch": 1.19, "learning_rate": 3.015481758955547e-05, "loss": 0.0021, "step": 61370 }, { "epoch": 1.19, "learning_rate": 3.0154170629298244e-05, "loss": 0.0017, "step": 61372 }, { "epoch": 1.19, "learning_rate": 3.0153523669041013e-05, "loss": 0.0, "step": 61374 }, { "epoch": 1.19, "learning_rate": 3.015287670878378e-05, "loss": 0.0001, "step": 61376 }, { "epoch": 1.19, "learning_rate": 3.015222974852655e-05, "loss": 0.0, "step": 61378 }, { "epoch": 1.19, "learning_rate": 3.0151582788269317e-05, "loss": 0.0064, "step": 61380 }, { "epoch": 1.19, "learning_rate": 3.015093582801209e-05, "loss": 0.0, "step": 61382 }, { "epoch": 1.19, "learning_rate": 3.0150288867754855e-05, "loss": 0.0, "step": 61384 }, { "epoch": 1.19, "learning_rate": 3.014964190749762e-05, "loss": 0.0, "step": 61386 }, { "epoch": 1.19, "learning_rate": 3.0148994947240393e-05, "loss": 0.0, "step": 61388 }, { "epoch": 1.19, "learning_rate": 3.0148347986983162e-05, "loss": 0.0, "step": 61390 }, { "epoch": 1.19, "learning_rate": 3.0147701026725928e-05, "loss": 0.0, "step": 61392 }, { "epoch": 1.19, "learning_rate": 3.01470540664687e-05, "loss": 0.0002, "step": 61394 }, { "epoch": 1.19, "learning_rate": 3.0146407106211466e-05, "loss": 0.0, "step": 61396 }, { "epoch": 1.19, "learning_rate": 3.0145760145954232e-05, "loss": 0.0, "step": 61398 }, { "epoch": 1.19, "learning_rate": 3.0145113185697004e-05, "loss": 0.0, "step": 61400 }, { "epoch": 1.19, "learning_rate": 3.014446622543977e-05, "loss": 0.0004, "step": 61402 }, { "epoch": 1.19, "learning_rate": 3.0143819265182542e-05, "loss": 0.0027, "step": 61404 }, { "epoch": 1.19, "learning_rate": 3.014317230492531e-05, "loss": 0.0, "step": 61406 }, { "epoch": 1.19, "learning_rate": 3.0142525344668077e-05, "loss": 0.0, "step": 61408 }, { "epoch": 1.19, "learning_rate": 3.014187838441085e-05, "loss": 0.0, "step": 61410 }, { "epoch": 1.19, "learning_rate": 3.0141231424153615e-05, "loss": 0.0, "step": 61412 }, { "epoch": 1.19, "learning_rate": 3.014058446389638e-05, "loss": 0.0346, "step": 61414 }, { "epoch": 1.19, "learning_rate": 3.0139937503639154e-05, "loss": 0.0001, "step": 61416 }, { "epoch": 1.19, "learning_rate": 3.013929054338192e-05, "loss": 0.0, "step": 61418 }, { "epoch": 1.19, "learning_rate": 3.013864358312469e-05, "loss": 0.0, "step": 61420 }, { "epoch": 1.19, "learning_rate": 3.013799662286746e-05, "loss": 0.0, "step": 61422 }, { "epoch": 1.19, "learning_rate": 3.0137349662610226e-05, "loss": 0.0001, "step": 61424 }, { "epoch": 1.19, "learning_rate": 3.0136702702353e-05, "loss": 0.0056, "step": 61426 }, { "epoch": 1.19, "learning_rate": 3.0136055742095765e-05, "loss": 0.0005, "step": 61428 }, { "epoch": 1.19, "learning_rate": 3.013540878183853e-05, "loss": 0.0, "step": 61430 }, { "epoch": 1.19, "learning_rate": 3.0134761821581303e-05, "loss": 0.0001, "step": 61432 }, { "epoch": 1.19, "learning_rate": 3.013411486132407e-05, "loss": 0.0057, "step": 61434 }, { "epoch": 1.19, "learning_rate": 3.0133467901066838e-05, "loss": 0.0001, "step": 61436 }, { "epoch": 1.19, "learning_rate": 3.013282094080961e-05, "loss": 0.0, "step": 61438 }, { "epoch": 1.19, "learning_rate": 3.0132173980552376e-05, "loss": 0.0, "step": 61440 }, { "epoch": 1.19, "learning_rate": 3.0131527020295148e-05, "loss": 0.0, "step": 61442 }, { "epoch": 1.19, "learning_rate": 3.0130880060037914e-05, "loss": 0.0, "step": 61444 }, { "epoch": 1.19, "learning_rate": 3.013023309978068e-05, "loss": 0.0, "step": 61446 }, { "epoch": 1.19, "learning_rate": 3.0129586139523452e-05, "loss": 0.0, "step": 61448 }, { "epoch": 1.19, "learning_rate": 3.0128939179266218e-05, "loss": 0.0002, "step": 61450 }, { "epoch": 1.19, "learning_rate": 3.0128292219008987e-05, "loss": 0.0, "step": 61452 }, { "epoch": 1.19, "learning_rate": 3.012764525875176e-05, "loss": 0.0, "step": 61454 }, { "epoch": 1.19, "learning_rate": 3.0126998298494525e-05, "loss": 0.0001, "step": 61456 }, { "epoch": 1.19, "learning_rate": 3.012635133823729e-05, "loss": 0.0, "step": 61458 }, { "epoch": 1.19, "learning_rate": 3.0125704377980063e-05, "loss": 0.0, "step": 61460 }, { "epoch": 1.19, "learning_rate": 3.012505741772283e-05, "loss": 0.0042, "step": 61462 }, { "epoch": 1.19, "learning_rate": 3.01244104574656e-05, "loss": 0.0, "step": 61464 }, { "epoch": 1.19, "learning_rate": 3.0123763497208367e-05, "loss": 0.0, "step": 61466 }, { "epoch": 1.19, "learning_rate": 3.0123116536951136e-05, "loss": 0.0, "step": 61468 }, { "epoch": 1.19, "learning_rate": 3.0122469576693905e-05, "loss": 0.0, "step": 61470 }, { "epoch": 1.19, "learning_rate": 3.0121822616436674e-05, "loss": 0.0, "step": 61472 }, { "epoch": 1.19, "learning_rate": 3.0121499136308055e-05, "loss": 0.0762, "step": 61474 }, { "epoch": 1.19, "learning_rate": 3.0120852176050828e-05, "loss": 0.0, "step": 61476 }, { "epoch": 1.19, "learning_rate": 3.0120205215793597e-05, "loss": 0.0, "step": 61478 }, { "epoch": 1.19, "learning_rate": 3.0119558255536363e-05, "loss": 0.0, "step": 61480 }, { "epoch": 1.19, "learning_rate": 3.0118911295279135e-05, "loss": 0.0, "step": 61482 }, { "epoch": 1.19, "learning_rate": 3.01182643350219e-05, "loss": 0.0, "step": 61484 }, { "epoch": 1.19, "learning_rate": 3.0117617374764667e-05, "loss": 0.0, "step": 61486 }, { "epoch": 1.19, "learning_rate": 3.011697041450744e-05, "loss": 0.0, "step": 61488 }, { "epoch": 1.19, "learning_rate": 3.0116323454250205e-05, "loss": 0.0, "step": 61490 }, { "epoch": 1.19, "learning_rate": 3.0115676493992977e-05, "loss": 0.0, "step": 61492 }, { "epoch": 1.19, "learning_rate": 3.0115029533735746e-05, "loss": 0.0, "step": 61494 }, { "epoch": 1.19, "learning_rate": 3.0114382573478512e-05, "loss": 0.0002, "step": 61496 }, { "epoch": 1.19, "learning_rate": 3.0113735613221284e-05, "loss": 0.0, "step": 61498 }, { "epoch": 1.19, "learning_rate": 3.011308865296405e-05, "loss": 0.0, "step": 61500 }, { "epoch": 1.19, "learning_rate": 3.0112441692706816e-05, "loss": 0.0, "step": 61502 }, { "epoch": 1.19, "learning_rate": 3.011179473244959e-05, "loss": 0.0, "step": 61504 }, { "epoch": 1.19, "learning_rate": 3.0111147772192354e-05, "loss": 0.0, "step": 61506 }, { "epoch": 1.19, "learning_rate": 3.0110500811935123e-05, "loss": 0.0, "step": 61508 }, { "epoch": 1.19, "learning_rate": 3.0109853851677892e-05, "loss": 0.0001, "step": 61510 }, { "epoch": 1.19, "learning_rate": 3.010920689142066e-05, "loss": 0.0, "step": 61512 }, { "epoch": 1.19, "learning_rate": 3.0108559931163434e-05, "loss": 0.0, "step": 61514 }, { "epoch": 1.19, "learning_rate": 3.01079129709062e-05, "loss": 0.0, "step": 61516 }, { "epoch": 1.19, "learning_rate": 3.0107266010648965e-05, "loss": 0.0, "step": 61518 }, { "epoch": 1.19, "learning_rate": 3.0106619050391738e-05, "loss": 0.0, "step": 61520 }, { "epoch": 1.19, "learning_rate": 3.0105972090134503e-05, "loss": 0.0006, "step": 61522 }, { "epoch": 1.19, "learning_rate": 3.0105325129877272e-05, "loss": 0.0, "step": 61524 }, { "epoch": 1.19, "learning_rate": 3.010467816962004e-05, "loss": 0.0, "step": 61526 }, { "epoch": 1.19, "learning_rate": 3.010403120936281e-05, "loss": 0.0, "step": 61528 }, { "epoch": 1.19, "learning_rate": 3.0103384249105576e-05, "loss": 0.0, "step": 61530 }, { "epoch": 1.19, "learning_rate": 3.010273728884835e-05, "loss": 0.0, "step": 61532 }, { "epoch": 1.19, "learning_rate": 3.0102090328591114e-05, "loss": 0.0, "step": 61534 }, { "epoch": 1.19, "learning_rate": 3.0101443368333887e-05, "loss": 0.0, "step": 61536 }, { "epoch": 1.19, "learning_rate": 3.0100796408076653e-05, "loss": 0.0, "step": 61538 }, { "epoch": 1.19, "learning_rate": 3.0100149447819422e-05, "loss": 0.0, "step": 61540 }, { "epoch": 1.19, "learning_rate": 3.009950248756219e-05, "loss": 0.0, "step": 61542 }, { "epoch": 1.19, "learning_rate": 3.009885552730496e-05, "loss": 0.0, "step": 61544 }, { "epoch": 1.19, "learning_rate": 3.0098208567047726e-05, "loss": 0.0, "step": 61546 }, { "epoch": 1.19, "learning_rate": 3.0097561606790498e-05, "loss": 0.0, "step": 61548 }, { "epoch": 1.19, "learning_rate": 3.0096914646533264e-05, "loss": 0.0, "step": 61550 }, { "epoch": 1.19, "learning_rate": 3.0096267686276036e-05, "loss": 0.0, "step": 61552 }, { "epoch": 1.19, "learning_rate": 3.0095620726018802e-05, "loss": 0.0, "step": 61554 }, { "epoch": 1.19, "learning_rate": 3.009497376576157e-05, "loss": 0.0, "step": 61556 }, { "epoch": 1.19, "learning_rate": 3.009432680550434e-05, "loss": 0.0, "step": 61558 }, { "epoch": 1.19, "learning_rate": 3.009367984524711e-05, "loss": 0.0, "step": 61560 }, { "epoch": 1.19, "learning_rate": 3.0093032884989875e-05, "loss": 0.0001, "step": 61562 }, { "epoch": 1.19, "learning_rate": 3.0092385924732647e-05, "loss": 0.0, "step": 61564 }, { "epoch": 1.19, "learning_rate": 3.0091738964475413e-05, "loss": 0.0, "step": 61566 }, { "epoch": 1.19, "learning_rate": 3.009109200421818e-05, "loss": 0.0, "step": 61568 }, { "epoch": 1.2, "learning_rate": 3.009044504396095e-05, "loss": 0.0, "step": 61570 }, { "epoch": 1.2, "learning_rate": 3.0089798083703717e-05, "loss": 0.0, "step": 61572 }, { "epoch": 1.2, "learning_rate": 3.008915112344649e-05, "loss": 0.0, "step": 61574 }, { "epoch": 1.2, "learning_rate": 3.008850416318926e-05, "loss": 0.0001, "step": 61576 }, { "epoch": 1.2, "learning_rate": 3.0087857202932024e-05, "loss": 0.0, "step": 61578 }, { "epoch": 1.2, "learning_rate": 3.0087210242674797e-05, "loss": 0.0, "step": 61580 }, { "epoch": 1.2, "learning_rate": 3.0086563282417562e-05, "loss": 0.0082, "step": 61582 }, { "epoch": 1.2, "learning_rate": 3.0085916322160328e-05, "loss": 0.0006, "step": 61584 }, { "epoch": 1.2, "learning_rate": 3.00852693619031e-05, "loss": 0.0016, "step": 61586 }, { "epoch": 1.2, "learning_rate": 3.0084622401645866e-05, "loss": 0.0, "step": 61588 }, { "epoch": 1.2, "learning_rate": 3.0083975441388635e-05, "loss": 0.0, "step": 61590 }, { "epoch": 1.2, "learning_rate": 3.0083328481131408e-05, "loss": 0.0, "step": 61592 }, { "epoch": 1.2, "learning_rate": 3.0082681520874173e-05, "loss": 0.0, "step": 61594 }, { "epoch": 1.2, "learning_rate": 3.0082034560616946e-05, "loss": 0.0, "step": 61596 }, { "epoch": 1.2, "learning_rate": 3.008138760035971e-05, "loss": 0.0, "step": 61598 }, { "epoch": 1.2, "learning_rate": 3.0080740640102477e-05, "loss": 0.0, "step": 61600 }, { "epoch": 1.2, "learning_rate": 3.008009367984525e-05, "loss": 0.0004, "step": 61602 }, { "epoch": 1.2, "learning_rate": 3.0079446719588015e-05, "loss": 0.0, "step": 61604 }, { "epoch": 1.2, "learning_rate": 3.0078799759330785e-05, "loss": 0.0, "step": 61606 }, { "epoch": 1.2, "learning_rate": 3.0078152799073557e-05, "loss": 0.0, "step": 61608 }, { "epoch": 1.2, "learning_rate": 3.0077505838816323e-05, "loss": 0.0012, "step": 61610 }, { "epoch": 1.2, "learning_rate": 3.007685887855909e-05, "loss": 0.0001, "step": 61612 }, { "epoch": 1.2, "learning_rate": 3.007621191830186e-05, "loss": 0.0004, "step": 61614 }, { "epoch": 1.2, "learning_rate": 3.0075564958044627e-05, "loss": 0.0, "step": 61616 }, { "epoch": 1.2, "learning_rate": 3.00749179977874e-05, "loss": 0.0, "step": 61618 }, { "epoch": 1.2, "learning_rate": 3.0074271037530165e-05, "loss": 0.0, "step": 61620 }, { "epoch": 1.2, "learning_rate": 3.0073624077272934e-05, "loss": 0.0, "step": 61622 }, { "epoch": 1.2, "learning_rate": 3.0072977117015706e-05, "loss": 0.0004, "step": 61624 }, { "epoch": 1.2, "learning_rate": 3.0072330156758472e-05, "loss": 0.0073, "step": 61626 }, { "epoch": 1.2, "learning_rate": 3.0071683196501238e-05, "loss": 0.0005, "step": 61628 }, { "epoch": 1.2, "learning_rate": 3.007103623624401e-05, "loss": 0.0, "step": 61630 }, { "epoch": 1.2, "learning_rate": 3.0070389275986776e-05, "loss": 0.0, "step": 61632 }, { "epoch": 1.2, "learning_rate": 3.006974231572955e-05, "loss": 0.0, "step": 61634 }, { "epoch": 1.2, "learning_rate": 3.0069095355472314e-05, "loss": 0.0, "step": 61636 }, { "epoch": 1.2, "learning_rate": 3.0068448395215083e-05, "loss": 0.0, "step": 61638 }, { "epoch": 1.2, "learning_rate": 3.0067801434957856e-05, "loss": 0.0002, "step": 61640 }, { "epoch": 1.2, "learning_rate": 3.006715447470062e-05, "loss": 0.0, "step": 61642 }, { "epoch": 1.2, "learning_rate": 3.0066507514443387e-05, "loss": 0.0, "step": 61644 }, { "epoch": 1.2, "learning_rate": 3.006586055418616e-05, "loss": 0.0, "step": 61646 }, { "epoch": 1.2, "learning_rate": 3.0065213593928925e-05, "loss": 0.0, "step": 61648 }, { "epoch": 1.2, "learning_rate": 3.006456663367169e-05, "loss": 0.0028, "step": 61650 }, { "epoch": 1.2, "learning_rate": 3.0063919673414463e-05, "loss": 0.0, "step": 61652 }, { "epoch": 1.2, "learning_rate": 3.0063272713157232e-05, "loss": 0.0, "step": 61654 }, { "epoch": 1.2, "learning_rate": 3.0062625752900005e-05, "loss": 0.0, "step": 61656 }, { "epoch": 1.2, "learning_rate": 3.006197879264277e-05, "loss": 0.0001, "step": 61658 }, { "epoch": 1.2, "learning_rate": 3.0061331832385536e-05, "loss": 0.0012, "step": 61660 }, { "epoch": 1.2, "learning_rate": 3.006068487212831e-05, "loss": 0.0, "step": 61662 }, { "epoch": 1.2, "learning_rate": 3.0060037911871074e-05, "loss": 0.0001, "step": 61664 }, { "epoch": 1.2, "learning_rate": 3.005939095161384e-05, "loss": 0.0, "step": 61666 }, { "epoch": 1.2, "learning_rate": 3.0058743991356613e-05, "loss": 0.0033, "step": 61668 }, { "epoch": 1.2, "learning_rate": 3.0058097031099382e-05, "loss": 0.0, "step": 61670 }, { "epoch": 1.2, "learning_rate": 3.0057450070842147e-05, "loss": 0.0, "step": 61672 }, { "epoch": 1.2, "learning_rate": 3.005680311058492e-05, "loss": 0.0, "step": 61674 }, { "epoch": 1.2, "learning_rate": 3.0056156150327686e-05, "loss": 0.0, "step": 61676 }, { "epoch": 1.2, "learning_rate": 3.0055509190070458e-05, "loss": 0.0, "step": 61678 }, { "epoch": 1.2, "learning_rate": 3.0054862229813224e-05, "loss": 0.0001, "step": 61680 }, { "epoch": 1.2, "learning_rate": 3.005421526955599e-05, "loss": 0.0, "step": 61682 }, { "epoch": 1.2, "learning_rate": 3.0053568309298762e-05, "loss": 0.0, "step": 61684 }, { "epoch": 1.2, "learning_rate": 3.005292134904153e-05, "loss": 0.0004, "step": 61686 }, { "epoch": 1.2, "learning_rate": 3.0052274388784297e-05, "loss": 0.0, "step": 61688 }, { "epoch": 1.2, "learning_rate": 3.005162742852707e-05, "loss": 0.0, "step": 61690 }, { "epoch": 1.2, "learning_rate": 3.0050980468269835e-05, "loss": 0.0092, "step": 61692 }, { "epoch": 1.2, "learning_rate": 3.0050333508012607e-05, "loss": 0.0018, "step": 61694 }, { "epoch": 1.2, "learning_rate": 3.0049686547755373e-05, "loss": 0.0, "step": 61696 }, { "epoch": 1.2, "learning_rate": 3.004903958749814e-05, "loss": 0.0, "step": 61698 }, { "epoch": 1.2, "learning_rate": 3.004839262724091e-05, "loss": 0.0, "step": 61700 }, { "epoch": 1.2, "learning_rate": 3.004774566698368e-05, "loss": 0.0, "step": 61702 }, { "epoch": 1.2, "learning_rate": 3.0047098706726446e-05, "loss": 0.0, "step": 61704 }, { "epoch": 1.2, "learning_rate": 3.004645174646922e-05, "loss": 0.0, "step": 61706 }, { "epoch": 1.2, "learning_rate": 3.0045804786211984e-05, "loss": 0.0, "step": 61708 }, { "epoch": 1.2, "learning_rate": 3.004515782595475e-05, "loss": 0.0, "step": 61710 }, { "epoch": 1.2, "learning_rate": 3.0044510865697522e-05, "loss": 0.0, "step": 61712 }, { "epoch": 1.2, "learning_rate": 3.0043863905440288e-05, "loss": 0.0004, "step": 61714 }, { "epoch": 1.2, "learning_rate": 3.004321694518306e-05, "loss": 0.0, "step": 61716 }, { "epoch": 1.2, "learning_rate": 3.004256998492583e-05, "loss": 0.0, "step": 61718 }, { "epoch": 1.2, "learning_rate": 3.0041923024668595e-05, "loss": 0.0, "step": 61720 }, { "epoch": 1.2, "learning_rate": 3.0041276064411368e-05, "loss": 0.0, "step": 61722 }, { "epoch": 1.2, "learning_rate": 3.0040629104154133e-05, "loss": 0.0, "step": 61724 }, { "epoch": 1.2, "learning_rate": 3.00399821438969e-05, "loss": 0.0, "step": 61726 }, { "epoch": 1.2, "learning_rate": 3.003933518363967e-05, "loss": 0.0, "step": 61728 }, { "epoch": 1.2, "learning_rate": 3.0038688223382437e-05, "loss": 0.0003, "step": 61730 }, { "epoch": 1.2, "learning_rate": 3.0038041263125206e-05, "loss": 0.0, "step": 61732 }, { "epoch": 1.2, "learning_rate": 3.003739430286798e-05, "loss": 0.0014, "step": 61734 }, { "epoch": 1.2, "learning_rate": 3.0036747342610745e-05, "loss": 0.0, "step": 61736 }, { "epoch": 1.2, "learning_rate": 3.0036100382353517e-05, "loss": 0.0, "step": 61738 }, { "epoch": 1.2, "learning_rate": 3.0035453422096283e-05, "loss": 0.0, "step": 61740 }, { "epoch": 1.2, "learning_rate": 3.003480646183905e-05, "loss": 0.0036, "step": 61742 }, { "epoch": 1.2, "learning_rate": 3.003415950158182e-05, "loss": 0.0, "step": 61744 }, { "epoch": 1.2, "learning_rate": 3.0033512541324587e-05, "loss": 0.0, "step": 61746 }, { "epoch": 1.2, "learning_rate": 3.0032865581067356e-05, "loss": 0.0, "step": 61748 }, { "epoch": 1.2, "learning_rate": 3.0032218620810125e-05, "loss": 0.0, "step": 61750 }, { "epoch": 1.2, "learning_rate": 3.0031571660552894e-05, "loss": 0.0, "step": 61752 }, { "epoch": 1.2, "learning_rate": 3.003092470029566e-05, "loss": 0.0, "step": 61754 }, { "epoch": 1.2, "learning_rate": 3.0030277740038432e-05, "loss": 0.0, "step": 61756 }, { "epoch": 1.2, "learning_rate": 3.0029630779781198e-05, "loss": 0.0, "step": 61758 }, { "epoch": 1.2, "learning_rate": 3.002898381952397e-05, "loss": 0.0009, "step": 61760 }, { "epoch": 1.2, "learning_rate": 3.0028336859266736e-05, "loss": 0.0, "step": 61762 }, { "epoch": 1.2, "learning_rate": 3.0027689899009505e-05, "loss": 0.0, "step": 61764 }, { "epoch": 1.2, "learning_rate": 3.0027042938752274e-05, "loss": 0.0, "step": 61766 }, { "epoch": 1.2, "learning_rate": 3.0026395978495043e-05, "loss": 0.0, "step": 61768 }, { "epoch": 1.2, "learning_rate": 3.002574901823781e-05, "loss": 0.0, "step": 61770 }, { "epoch": 1.2, "learning_rate": 3.002510205798058e-05, "loss": 0.0014, "step": 61772 }, { "epoch": 1.2, "learning_rate": 3.0024455097723347e-05, "loss": 0.0008, "step": 61774 }, { "epoch": 1.2, "learning_rate": 3.002380813746612e-05, "loss": 0.0, "step": 61776 }, { "epoch": 1.2, "learning_rate": 3.0023161177208885e-05, "loss": 0.0, "step": 61778 }, { "epoch": 1.2, "learning_rate": 3.0022514216951654e-05, "loss": 0.0, "step": 61780 }, { "epoch": 1.2, "learning_rate": 3.0021867256694423e-05, "loss": 0.0001, "step": 61782 }, { "epoch": 1.2, "learning_rate": 3.0021220296437192e-05, "loss": 0.0, "step": 61784 }, { "epoch": 1.2, "learning_rate": 3.0020573336179958e-05, "loss": 0.0, "step": 61786 }, { "epoch": 1.2, "learning_rate": 3.001992637592273e-05, "loss": 0.0, "step": 61788 }, { "epoch": 1.2, "learning_rate": 3.0019279415665496e-05, "loss": 0.0, "step": 61790 }, { "epoch": 1.2, "learning_rate": 3.0018632455408262e-05, "loss": 0.0041, "step": 61792 }, { "epoch": 1.2, "learning_rate": 3.0017985495151034e-05, "loss": 0.0, "step": 61794 }, { "epoch": 1.2, "learning_rate": 3.0017338534893804e-05, "loss": 0.0, "step": 61796 }, { "epoch": 1.2, "learning_rate": 3.0016691574636573e-05, "loss": 0.0, "step": 61798 }, { "epoch": 1.2, "learning_rate": 3.0016044614379342e-05, "loss": 0.0, "step": 61800 }, { "epoch": 1.2, "learning_rate": 3.0015397654122107e-05, "loss": 0.0, "step": 61802 }, { "epoch": 1.2, "learning_rate": 3.001475069386488e-05, "loss": 0.0, "step": 61804 }, { "epoch": 1.2, "learning_rate": 3.0014103733607646e-05, "loss": 0.0027, "step": 61806 }, { "epoch": 1.2, "learning_rate": 3.001345677335041e-05, "loss": 0.0, "step": 61808 }, { "epoch": 1.2, "learning_rate": 3.0012809813093184e-05, "loss": 0.0, "step": 61810 }, { "epoch": 1.2, "learning_rate": 3.001216285283595e-05, "loss": 0.0, "step": 61812 }, { "epoch": 1.2, "learning_rate": 3.001151589257872e-05, "loss": 0.0, "step": 61814 }, { "epoch": 1.2, "learning_rate": 3.001086893232149e-05, "loss": 0.0002, "step": 61816 }, { "epoch": 1.2, "learning_rate": 3.0010221972064257e-05, "loss": 0.0, "step": 61818 }, { "epoch": 1.2, "learning_rate": 3.000957501180703e-05, "loss": 0.0, "step": 61820 }, { "epoch": 1.2, "learning_rate": 3.0008928051549795e-05, "loss": 0.0009, "step": 61822 }, { "epoch": 1.2, "learning_rate": 3.000828109129256e-05, "loss": 0.0003, "step": 61824 }, { "epoch": 1.2, "learning_rate": 3.0007634131035333e-05, "loss": 0.0001, "step": 61826 }, { "epoch": 1.2, "learning_rate": 3.00069871707781e-05, "loss": 0.0, "step": 61828 }, { "epoch": 1.2, "learning_rate": 3.0006340210520868e-05, "loss": 0.0, "step": 61830 }, { "epoch": 1.2, "learning_rate": 3.000569325026364e-05, "loss": 0.0, "step": 61832 }, { "epoch": 1.2, "learning_rate": 3.0005046290006406e-05, "loss": 0.0, "step": 61834 }, { "epoch": 1.2, "learning_rate": 3.000439932974918e-05, "loss": 0.0031, "step": 61836 }, { "epoch": 1.2, "learning_rate": 3.0003752369491944e-05, "loss": 0.0, "step": 61838 }, { "epoch": 1.2, "learning_rate": 3.000310540923471e-05, "loss": 0.0028, "step": 61840 }, { "epoch": 1.2, "learning_rate": 3.0002458448977482e-05, "loss": 0.0, "step": 61842 }, { "epoch": 1.2, "learning_rate": 3.0001811488720248e-05, "loss": 0.0, "step": 61844 }, { "epoch": 1.2, "learning_rate": 3.0001164528463017e-05, "loss": 0.0, "step": 61846 }, { "epoch": 1.2, "learning_rate": 3.000051756820579e-05, "loss": 0.0002, "step": 61848 }, { "epoch": 1.2, "learning_rate": 2.9999870607948555e-05, "loss": 0.0, "step": 61850 }, { "epoch": 1.2, "learning_rate": 2.999922364769132e-05, "loss": 0.0, "step": 61852 }, { "epoch": 1.2, "learning_rate": 2.9998576687434093e-05, "loss": 0.0009, "step": 61854 }, { "epoch": 1.2, "learning_rate": 2.999792972717686e-05, "loss": 0.0, "step": 61856 }, { "epoch": 1.2, "learning_rate": 2.999728276691963e-05, "loss": 0.0099, "step": 61858 }, { "epoch": 1.2, "learning_rate": 2.9996635806662397e-05, "loss": 0.0, "step": 61860 }, { "epoch": 1.2, "learning_rate": 2.9995988846405166e-05, "loss": 0.0, "step": 61862 }, { "epoch": 1.2, "learning_rate": 2.999534188614794e-05, "loss": 0.0, "step": 61864 }, { "epoch": 1.2, "learning_rate": 2.9994694925890705e-05, "loss": 0.0, "step": 61866 }, { "epoch": 1.2, "learning_rate": 2.999404796563347e-05, "loss": 0.0, "step": 61868 }, { "epoch": 1.2, "learning_rate": 2.9993401005376243e-05, "loss": 0.0116, "step": 61870 }, { "epoch": 1.2, "learning_rate": 2.999275404511901e-05, "loss": 0.0, "step": 61872 }, { "epoch": 1.2, "learning_rate": 2.9992107084861774e-05, "loss": 0.0, "step": 61874 }, { "epoch": 1.2, "learning_rate": 2.9991460124604547e-05, "loss": 0.0, "step": 61876 }, { "epoch": 1.2, "learning_rate": 2.9990813164347316e-05, "loss": 0.0058, "step": 61878 }, { "epoch": 1.2, "learning_rate": 2.9990166204090088e-05, "loss": 0.0016, "step": 61880 }, { "epoch": 1.2, "learning_rate": 2.9989519243832854e-05, "loss": 0.0, "step": 61882 }, { "epoch": 1.2, "learning_rate": 2.998887228357562e-05, "loss": 0.0191, "step": 61884 }, { "epoch": 1.2, "learning_rate": 2.9988225323318392e-05, "loss": 0.0001, "step": 61886 }, { "epoch": 1.2, "learning_rate": 2.9987578363061158e-05, "loss": 0.0293, "step": 61888 }, { "epoch": 1.2, "learning_rate": 2.9986931402803923e-05, "loss": 0.0, "step": 61890 }, { "epoch": 1.2, "learning_rate": 2.9986284442546696e-05, "loss": 0.0, "step": 61892 }, { "epoch": 1.2, "learning_rate": 2.9985637482289465e-05, "loss": 0.0, "step": 61894 }, { "epoch": 1.2, "learning_rate": 2.998499052203223e-05, "loss": 0.0, "step": 61896 }, { "epoch": 1.2, "learning_rate": 2.9984343561775003e-05, "loss": 0.0002, "step": 61898 }, { "epoch": 1.2, "learning_rate": 2.998369660151777e-05, "loss": 0.0, "step": 61900 }, { "epoch": 1.2, "learning_rate": 2.998304964126054e-05, "loss": 0.0, "step": 61902 }, { "epoch": 1.2, "learning_rate": 2.9982402681003307e-05, "loss": 0.0, "step": 61904 }, { "epoch": 1.2, "learning_rate": 2.9981755720746073e-05, "loss": 0.0, "step": 61906 }, { "epoch": 1.2, "learning_rate": 2.9981108760488845e-05, "loss": 0.0, "step": 61908 }, { "epoch": 1.2, "learning_rate": 2.9980461800231614e-05, "loss": 0.0, "step": 61910 }, { "epoch": 1.2, "learning_rate": 2.997981483997438e-05, "loss": 0.0, "step": 61912 }, { "epoch": 1.2, "learning_rate": 2.9979167879717152e-05, "loss": 0.0, "step": 61914 }, { "epoch": 1.2, "learning_rate": 2.9978520919459918e-05, "loss": 0.0, "step": 61916 }, { "epoch": 1.2, "learning_rate": 2.997787395920269e-05, "loss": 0.001, "step": 61918 }, { "epoch": 1.2, "learning_rate": 2.9977226998945456e-05, "loss": 0.0, "step": 61920 }, { "epoch": 1.2, "learning_rate": 2.9976580038688222e-05, "loss": 0.0, "step": 61922 }, { "epoch": 1.2, "learning_rate": 2.9975933078430994e-05, "loss": 0.0001, "step": 61924 }, { "epoch": 1.2, "learning_rate": 2.9975286118173764e-05, "loss": 0.0006, "step": 61926 }, { "epoch": 1.2, "learning_rate": 2.997463915791653e-05, "loss": 0.0, "step": 61928 }, { "epoch": 1.2, "learning_rate": 2.99739921976593e-05, "loss": 0.0, "step": 61930 }, { "epoch": 1.2, "learning_rate": 2.9973345237402067e-05, "loss": 0.0, "step": 61932 }, { "epoch": 1.2, "learning_rate": 2.9972698277144833e-05, "loss": 0.0047, "step": 61934 }, { "epoch": 1.2, "learning_rate": 2.9972051316887606e-05, "loss": 0.0003, "step": 61936 }, { "epoch": 1.2, "learning_rate": 2.997140435663037e-05, "loss": 0.0011, "step": 61938 }, { "epoch": 1.2, "learning_rate": 2.9970757396373144e-05, "loss": 0.0, "step": 61940 }, { "epoch": 1.2, "learning_rate": 2.9970110436115913e-05, "loss": 0.0, "step": 61942 }, { "epoch": 1.2, "learning_rate": 2.996946347585868e-05, "loss": 0.0, "step": 61944 }, { "epoch": 1.2, "learning_rate": 2.996881651560145e-05, "loss": 0.0, "step": 61946 }, { "epoch": 1.2, "learning_rate": 2.9968169555344217e-05, "loss": 0.0, "step": 61948 }, { "epoch": 1.2, "learning_rate": 2.9967522595086982e-05, "loss": 0.0, "step": 61950 }, { "epoch": 1.2, "learning_rate": 2.9966875634829755e-05, "loss": 0.0, "step": 61952 }, { "epoch": 1.2, "learning_rate": 2.996622867457252e-05, "loss": 0.0, "step": 61954 }, { "epoch": 1.2, "learning_rate": 2.996558171431529e-05, "loss": 0.0, "step": 61956 }, { "epoch": 1.2, "learning_rate": 2.9964934754058062e-05, "loss": 0.0, "step": 61958 }, { "epoch": 1.2, "learning_rate": 2.9964287793800828e-05, "loss": 0.0, "step": 61960 }, { "epoch": 1.2, "learning_rate": 2.99636408335436e-05, "loss": 0.0, "step": 61962 }, { "epoch": 1.2, "learning_rate": 2.9962993873286366e-05, "loss": 0.0023, "step": 61964 }, { "epoch": 1.2, "learning_rate": 2.996234691302913e-05, "loss": 0.0, "step": 61966 }, { "epoch": 1.2, "learning_rate": 2.9961699952771904e-05, "loss": 0.0, "step": 61968 }, { "epoch": 1.2, "learning_rate": 2.996105299251467e-05, "loss": 0.0022, "step": 61970 }, { "epoch": 1.2, "learning_rate": 2.996040603225744e-05, "loss": 0.0063, "step": 61972 }, { "epoch": 1.2, "learning_rate": 2.995975907200021e-05, "loss": 0.0004, "step": 61974 }, { "epoch": 1.2, "learning_rate": 2.9959112111742977e-05, "loss": 0.0, "step": 61976 }, { "epoch": 1.2, "learning_rate": 2.995846515148575e-05, "loss": 0.0, "step": 61978 }, { "epoch": 1.2, "learning_rate": 2.9957818191228515e-05, "loss": 0.0, "step": 61980 }, { "epoch": 1.2, "learning_rate": 2.995717123097128e-05, "loss": 0.0, "step": 61982 }, { "epoch": 1.2, "learning_rate": 2.9956524270714053e-05, "loss": 0.0, "step": 61984 }, { "epoch": 1.2, "learning_rate": 2.995587731045682e-05, "loss": 0.0, "step": 61986 }, { "epoch": 1.2, "learning_rate": 2.9955230350199588e-05, "loss": 0.0, "step": 61988 }, { "epoch": 1.2, "learning_rate": 2.9954583389942357e-05, "loss": 0.0, "step": 61990 }, { "epoch": 1.2, "learning_rate": 2.9953936429685126e-05, "loss": 0.0, "step": 61992 }, { "epoch": 1.2, "learning_rate": 2.9953289469427892e-05, "loss": 0.0, "step": 61994 }, { "epoch": 1.2, "learning_rate": 2.9952642509170665e-05, "loss": 0.0, "step": 61996 }, { "epoch": 1.2, "learning_rate": 2.995199554891343e-05, "loss": 0.0, "step": 61998 }, { "epoch": 1.2, "learning_rate": 2.9951348588656203e-05, "loss": 0.0, "step": 62000 }, { "epoch": 1.2, "learning_rate": 2.995070162839897e-05, "loss": 0.0, "step": 62002 }, { "epoch": 1.2, "learning_rate": 2.9950054668141737e-05, "loss": 0.0, "step": 62004 }, { "epoch": 1.2, "learning_rate": 2.9949407707884507e-05, "loss": 0.0, "step": 62006 }, { "epoch": 1.2, "learning_rate": 2.9948760747627276e-05, "loss": 0.0, "step": 62008 }, { "epoch": 1.2, "learning_rate": 2.994811378737004e-05, "loss": 0.0, "step": 62010 }, { "epoch": 1.2, "learning_rate": 2.9947466827112814e-05, "loss": 0.0, "step": 62012 }, { "epoch": 1.2, "learning_rate": 2.994681986685558e-05, "loss": 0.0002, "step": 62014 }, { "epoch": 1.2, "learning_rate": 2.9946172906598345e-05, "loss": 0.0002, "step": 62016 }, { "epoch": 1.2, "learning_rate": 2.9945525946341118e-05, "loss": 0.0006, "step": 62018 }, { "epoch": 1.2, "learning_rate": 2.9944878986083887e-05, "loss": 0.0, "step": 62020 }, { "epoch": 1.2, "learning_rate": 2.9944232025826656e-05, "loss": 0.0001, "step": 62022 }, { "epoch": 1.2, "learning_rate": 2.9943585065569425e-05, "loss": 0.0003, "step": 62024 }, { "epoch": 1.2, "learning_rate": 2.994293810531219e-05, "loss": 0.0, "step": 62026 }, { "epoch": 1.2, "learning_rate": 2.9942291145054963e-05, "loss": 0.0, "step": 62028 }, { "epoch": 1.2, "learning_rate": 2.994164418479773e-05, "loss": 0.0, "step": 62030 }, { "epoch": 1.2, "learning_rate": 2.9940997224540494e-05, "loss": 0.0, "step": 62032 }, { "epoch": 1.2, "learning_rate": 2.9940350264283267e-05, "loss": 0.0, "step": 62034 }, { "epoch": 1.2, "learning_rate": 2.9939703304026036e-05, "loss": 0.0, "step": 62036 }, { "epoch": 1.2, "learning_rate": 2.9939056343768802e-05, "loss": 0.0, "step": 62038 }, { "epoch": 1.2, "learning_rate": 2.9938409383511574e-05, "loss": 0.0, "step": 62040 }, { "epoch": 1.2, "learning_rate": 2.993776242325434e-05, "loss": 0.0, "step": 62042 }, { "epoch": 1.2, "learning_rate": 2.9937115462997112e-05, "loss": 0.0001, "step": 62044 }, { "epoch": 1.2, "learning_rate": 2.9936468502739878e-05, "loss": 0.0002, "step": 62046 }, { "epoch": 1.2, "learning_rate": 2.9935821542482644e-05, "loss": 0.0001, "step": 62048 }, { "epoch": 1.2, "learning_rate": 2.9935174582225416e-05, "loss": 0.0, "step": 62050 }, { "epoch": 1.2, "learning_rate": 2.9934527621968182e-05, "loss": 0.0025, "step": 62052 }, { "epoch": 1.2, "learning_rate": 2.993388066171095e-05, "loss": 0.0, "step": 62054 }, { "epoch": 1.2, "learning_rate": 2.9933233701453724e-05, "loss": 0.0001, "step": 62056 }, { "epoch": 1.2, "learning_rate": 2.993258674119649e-05, "loss": 0.0, "step": 62058 }, { "epoch": 1.2, "learning_rate": 2.993193978093926e-05, "loss": 0.0077, "step": 62060 }, { "epoch": 1.2, "learning_rate": 2.9931292820682027e-05, "loss": 0.0001, "step": 62062 }, { "epoch": 1.2, "learning_rate": 2.9930645860424793e-05, "loss": 0.0, "step": 62064 }, { "epoch": 1.2, "learning_rate": 2.9929998900167566e-05, "loss": 0.0003, "step": 62066 }, { "epoch": 1.2, "learning_rate": 2.992935193991033e-05, "loss": 0.0, "step": 62068 }, { "epoch": 1.2, "learning_rate": 2.99287049796531e-05, "loss": 0.0, "step": 62070 }, { "epoch": 1.2, "learning_rate": 2.9928058019395873e-05, "loss": 0.001, "step": 62072 }, { "epoch": 1.2, "learning_rate": 2.992741105913864e-05, "loss": 0.0014, "step": 62074 }, { "epoch": 1.2, "learning_rate": 2.9926764098881404e-05, "loss": 0.0001, "step": 62076 }, { "epoch": 1.2, "learning_rate": 2.9926117138624177e-05, "loss": 0.0039, "step": 62078 }, { "epoch": 1.2, "learning_rate": 2.9925470178366942e-05, "loss": 0.0, "step": 62080 }, { "epoch": 1.2, "learning_rate": 2.9924823218109715e-05, "loss": 0.0, "step": 62082 }, { "epoch": 1.2, "learning_rate": 2.992417625785248e-05, "loss": 0.0, "step": 62084 }, { "epoch": 1.21, "learning_rate": 2.992352929759525e-05, "loss": 0.0009, "step": 62086 }, { "epoch": 1.21, "learning_rate": 2.9922882337338022e-05, "loss": 0.0, "step": 62088 }, { "epoch": 1.21, "learning_rate": 2.9922235377080788e-05, "loss": 0.0, "step": 62090 }, { "epoch": 1.21, "learning_rate": 2.9921588416823553e-05, "loss": 0.0384, "step": 62092 }, { "epoch": 1.21, "learning_rate": 2.9920941456566326e-05, "loss": 0.0, "step": 62094 }, { "epoch": 1.21, "learning_rate": 2.992029449630909e-05, "loss": 0.0, "step": 62096 }, { "epoch": 1.21, "learning_rate": 2.991964753605186e-05, "loss": 0.0, "step": 62098 }, { "epoch": 1.21, "learning_rate": 2.991900057579463e-05, "loss": 0.0, "step": 62100 }, { "epoch": 1.21, "learning_rate": 2.99183536155374e-05, "loss": 0.0, "step": 62102 }, { "epoch": 1.21, "learning_rate": 2.991770665528017e-05, "loss": 0.0001, "step": 62104 }, { "epoch": 1.21, "learning_rate": 2.9917059695022937e-05, "loss": 0.0, "step": 62106 }, { "epoch": 1.21, "learning_rate": 2.9916412734765703e-05, "loss": 0.0, "step": 62108 }, { "epoch": 1.21, "learning_rate": 2.9915765774508475e-05, "loss": 0.0, "step": 62110 }, { "epoch": 1.21, "learning_rate": 2.991511881425124e-05, "loss": 0.0, "step": 62112 }, { "epoch": 1.21, "learning_rate": 2.9914471853994007e-05, "loss": 0.0, "step": 62114 }, { "epoch": 1.21, "learning_rate": 2.991382489373678e-05, "loss": 0.0, "step": 62116 }, { "epoch": 1.21, "learning_rate": 2.9913177933479548e-05, "loss": 0.0064, "step": 62118 }, { "epoch": 1.21, "learning_rate": 2.991253097322232e-05, "loss": 0.0, "step": 62120 }, { "epoch": 1.21, "learning_rate": 2.9911884012965086e-05, "loss": 0.0, "step": 62122 }, { "epoch": 1.21, "learning_rate": 2.9911237052707852e-05, "loss": 0.0, "step": 62124 }, { "epoch": 1.21, "learning_rate": 2.9910590092450625e-05, "loss": 0.0, "step": 62126 }, { "epoch": 1.21, "learning_rate": 2.990994313219339e-05, "loss": 0.0, "step": 62128 }, { "epoch": 1.21, "learning_rate": 2.9909296171936156e-05, "loss": 0.0, "step": 62130 }, { "epoch": 1.21, "learning_rate": 2.990864921167893e-05, "loss": 0.0, "step": 62132 }, { "epoch": 1.21, "learning_rate": 2.9908002251421697e-05, "loss": 0.0002, "step": 62134 }, { "epoch": 1.21, "learning_rate": 2.9907355291164463e-05, "loss": 0.0001, "step": 62136 }, { "epoch": 1.21, "learning_rate": 2.9906708330907236e-05, "loss": 0.0016, "step": 62138 }, { "epoch": 1.21, "learning_rate": 2.990606137065e-05, "loss": 0.0, "step": 62140 }, { "epoch": 1.21, "learning_rate": 2.9905414410392774e-05, "loss": 0.0, "step": 62142 }, { "epoch": 1.21, "learning_rate": 2.990476745013554e-05, "loss": 0.0, "step": 62144 }, { "epoch": 1.21, "learning_rate": 2.9904120489878305e-05, "loss": 0.0, "step": 62146 }, { "epoch": 1.21, "learning_rate": 2.9903473529621078e-05, "loss": 0.0, "step": 62148 }, { "epoch": 1.21, "learning_rate": 2.9902826569363847e-05, "loss": 0.0, "step": 62150 }, { "epoch": 1.21, "learning_rate": 2.9902179609106612e-05, "loss": 0.0, "step": 62152 }, { "epoch": 1.21, "learning_rate": 2.9901532648849385e-05, "loss": 0.0, "step": 62154 }, { "epoch": 1.21, "learning_rate": 2.990088568859215e-05, "loss": 0.0, "step": 62156 }, { "epoch": 1.21, "learning_rate": 2.9900238728334916e-05, "loss": 0.0001, "step": 62158 }, { "epoch": 1.21, "learning_rate": 2.989959176807769e-05, "loss": 0.0, "step": 62160 }, { "epoch": 1.21, "learning_rate": 2.9898944807820454e-05, "loss": 0.0, "step": 62162 }, { "epoch": 1.21, "learning_rate": 2.9898297847563227e-05, "loss": 0.0001, "step": 62164 }, { "epoch": 1.21, "learning_rate": 2.9897650887305996e-05, "loss": 0.0, "step": 62166 }, { "epoch": 1.21, "learning_rate": 2.9897003927048762e-05, "loss": 0.0, "step": 62168 }, { "epoch": 1.21, "learning_rate": 2.9896356966791534e-05, "loss": 0.0028, "step": 62170 }, { "epoch": 1.21, "learning_rate": 2.98957100065343e-05, "loss": 0.0001, "step": 62172 }, { "epoch": 1.21, "learning_rate": 2.9895063046277066e-05, "loss": 0.0013, "step": 62174 }, { "epoch": 1.21, "learning_rate": 2.9894416086019838e-05, "loss": 0.0, "step": 62176 }, { "epoch": 1.21, "learning_rate": 2.9893769125762604e-05, "loss": 0.0, "step": 62178 }, { "epoch": 1.21, "learning_rate": 2.9893122165505373e-05, "loss": 0.0, "step": 62180 }, { "epoch": 1.21, "learning_rate": 2.9892475205248145e-05, "loss": 0.0, "step": 62182 }, { "epoch": 1.21, "learning_rate": 2.989182824499091e-05, "loss": 0.0, "step": 62184 }, { "epoch": 1.21, "learning_rate": 2.9891181284733684e-05, "loss": 0.0016, "step": 62186 }, { "epoch": 1.21, "learning_rate": 2.989053432447645e-05, "loss": 0.0042, "step": 62188 }, { "epoch": 1.21, "learning_rate": 2.9889887364219215e-05, "loss": 0.0142, "step": 62190 }, { "epoch": 1.21, "learning_rate": 2.9889240403961987e-05, "loss": 0.0, "step": 62192 }, { "epoch": 1.21, "learning_rate": 2.9888593443704753e-05, "loss": 0.0001, "step": 62194 }, { "epoch": 1.21, "learning_rate": 2.9887946483447522e-05, "loss": 0.0, "step": 62196 }, { "epoch": 1.21, "learning_rate": 2.9887299523190295e-05, "loss": 0.0, "step": 62198 }, { "epoch": 1.21, "learning_rate": 2.988665256293306e-05, "loss": 0.0, "step": 62200 }, { "epoch": 1.21, "learning_rate": 2.9886005602675833e-05, "loss": 0.0, "step": 62202 }, { "epoch": 1.21, "learning_rate": 2.98853586424186e-05, "loss": 0.0077, "step": 62204 }, { "epoch": 1.21, "learning_rate": 2.9884711682161364e-05, "loss": 0.0, "step": 62206 }, { "epoch": 1.21, "learning_rate": 2.9884064721904137e-05, "loss": 0.0, "step": 62208 }, { "epoch": 1.21, "learning_rate": 2.9883417761646902e-05, "loss": 0.0, "step": 62210 }, { "epoch": 1.21, "learning_rate": 2.988277080138967e-05, "loss": 0.0, "step": 62212 }, { "epoch": 1.21, "learning_rate": 2.9882123841132444e-05, "loss": 0.0, "step": 62214 }, { "epoch": 1.21, "learning_rate": 2.988147688087521e-05, "loss": 0.0007, "step": 62216 }, { "epoch": 1.21, "learning_rate": 2.9880829920617975e-05, "loss": 0.0, "step": 62218 }, { "epoch": 1.21, "learning_rate": 2.9880182960360748e-05, "loss": 0.0019, "step": 62220 }, { "epoch": 1.21, "learning_rate": 2.9879536000103513e-05, "loss": 0.0001, "step": 62222 }, { "epoch": 1.21, "learning_rate": 2.9878889039846286e-05, "loss": 0.0059, "step": 62224 }, { "epoch": 1.21, "learning_rate": 2.987824207958905e-05, "loss": 0.0001, "step": 62226 }, { "epoch": 1.21, "learning_rate": 2.987759511933182e-05, "loss": 0.0, "step": 62228 }, { "epoch": 1.21, "learning_rate": 2.987694815907459e-05, "loss": 0.0116, "step": 62230 }, { "epoch": 1.21, "learning_rate": 2.987630119881736e-05, "loss": 0.0, "step": 62232 }, { "epoch": 1.21, "learning_rate": 2.9875654238560125e-05, "loss": 0.0001, "step": 62234 }, { "epoch": 1.21, "learning_rate": 2.9875007278302897e-05, "loss": 0.0, "step": 62236 }, { "epoch": 1.21, "learning_rate": 2.9874360318045663e-05, "loss": 0.0038, "step": 62238 }, { "epoch": 1.21, "learning_rate": 2.987371335778843e-05, "loss": 0.0, "step": 62240 }, { "epoch": 1.21, "learning_rate": 2.98730663975312e-05, "loss": 0.0001, "step": 62242 }, { "epoch": 1.21, "learning_rate": 2.987241943727397e-05, "loss": 0.0118, "step": 62244 }, { "epoch": 1.21, "learning_rate": 2.987177247701674e-05, "loss": 0.0, "step": 62246 }, { "epoch": 1.21, "learning_rate": 2.9871125516759508e-05, "loss": 0.0, "step": 62248 }, { "epoch": 1.21, "learning_rate": 2.9870478556502274e-05, "loss": 0.0, "step": 62250 }, { "epoch": 1.21, "learning_rate": 2.9869831596245046e-05, "loss": 0.0003, "step": 62252 }, { "epoch": 1.21, "learning_rate": 2.9869184635987812e-05, "loss": 0.0, "step": 62254 }, { "epoch": 1.21, "learning_rate": 2.9868537675730578e-05, "loss": 0.0001, "step": 62256 }, { "epoch": 1.21, "learning_rate": 2.986789071547335e-05, "loss": 0.0, "step": 62258 }, { "epoch": 1.21, "learning_rate": 2.986724375521612e-05, "loss": 0.0, "step": 62260 }, { "epoch": 1.21, "learning_rate": 2.986659679495889e-05, "loss": 0.001, "step": 62262 }, { "epoch": 1.21, "learning_rate": 2.9865949834701657e-05, "loss": 0.0, "step": 62264 }, { "epoch": 1.21, "learning_rate": 2.9865302874444423e-05, "loss": 0.0, "step": 62266 }, { "epoch": 1.21, "learning_rate": 2.9864655914187196e-05, "loss": 0.0, "step": 62268 }, { "epoch": 1.21, "learning_rate": 2.986400895392996e-05, "loss": 0.0, "step": 62270 }, { "epoch": 1.21, "learning_rate": 2.9863361993672727e-05, "loss": 0.0, "step": 62272 }, { "epoch": 1.21, "learning_rate": 2.98627150334155e-05, "loss": 0.0, "step": 62274 }, { "epoch": 1.21, "learning_rate": 2.986206807315827e-05, "loss": 0.0, "step": 62276 }, { "epoch": 1.21, "learning_rate": 2.9861421112901034e-05, "loss": 0.0009, "step": 62278 }, { "epoch": 1.21, "learning_rate": 2.9860774152643807e-05, "loss": 0.0, "step": 62280 }, { "epoch": 1.21, "learning_rate": 2.9860127192386572e-05, "loss": 0.0, "step": 62282 }, { "epoch": 1.21, "learning_rate": 2.9859480232129345e-05, "loss": 0.0054, "step": 62284 }, { "epoch": 1.21, "learning_rate": 2.985883327187211e-05, "loss": 0.0, "step": 62286 }, { "epoch": 1.21, "learning_rate": 2.9858186311614876e-05, "loss": 0.0, "step": 62288 }, { "epoch": 1.21, "learning_rate": 2.985753935135765e-05, "loss": 0.0, "step": 62290 }, { "epoch": 1.21, "learning_rate": 2.9856892391100414e-05, "loss": 0.0, "step": 62292 }, { "epoch": 1.21, "learning_rate": 2.9856245430843184e-05, "loss": 0.0, "step": 62294 }, { "epoch": 1.21, "learning_rate": 2.9855598470585956e-05, "loss": 0.0, "step": 62296 }, { "epoch": 1.21, "learning_rate": 2.9854951510328722e-05, "loss": 0.0, "step": 62298 }, { "epoch": 1.21, "learning_rate": 2.9854304550071487e-05, "loss": 0.0, "step": 62300 }, { "epoch": 1.21, "learning_rate": 2.985365758981426e-05, "loss": 0.0, "step": 62302 }, { "epoch": 1.21, "learning_rate": 2.9853010629557026e-05, "loss": 0.0, "step": 62304 }, { "epoch": 1.21, "learning_rate": 2.9852363669299798e-05, "loss": 0.0001, "step": 62306 }, { "epoch": 1.21, "learning_rate": 2.9851716709042564e-05, "loss": 0.0, "step": 62308 }, { "epoch": 1.21, "learning_rate": 2.9851069748785333e-05, "loss": 0.0001, "step": 62310 }, { "epoch": 1.21, "learning_rate": 2.9850422788528105e-05, "loss": 0.0015, "step": 62312 }, { "epoch": 1.21, "learning_rate": 2.984977582827087e-05, "loss": 0.0, "step": 62314 }, { "epoch": 1.21, "learning_rate": 2.9849128868013637e-05, "loss": 0.001, "step": 62316 }, { "epoch": 1.21, "learning_rate": 2.984848190775641e-05, "loss": 0.0, "step": 62318 }, { "epoch": 1.21, "learning_rate": 2.9847834947499175e-05, "loss": 0.0, "step": 62320 }, { "epoch": 1.21, "learning_rate": 2.9847187987241944e-05, "loss": 0.0, "step": 62322 }, { "epoch": 1.21, "learning_rate": 2.9846541026984713e-05, "loss": 0.0, "step": 62324 }, { "epoch": 1.21, "learning_rate": 2.9845894066727482e-05, "loss": 0.0, "step": 62326 }, { "epoch": 1.21, "learning_rate": 2.9845247106470255e-05, "loss": 0.0, "step": 62328 }, { "epoch": 1.21, "learning_rate": 2.984460014621302e-05, "loss": 0.0025, "step": 62330 }, { "epoch": 1.21, "learning_rate": 2.9843953185955786e-05, "loss": 0.0, "step": 62332 }, { "epoch": 1.21, "learning_rate": 2.984330622569856e-05, "loss": 0.0, "step": 62334 }, { "epoch": 1.21, "learning_rate": 2.9842659265441324e-05, "loss": 0.0, "step": 62336 }, { "epoch": 1.21, "learning_rate": 2.9842012305184093e-05, "loss": 0.0, "step": 62338 }, { "epoch": 1.21, "learning_rate": 2.9841365344926862e-05, "loss": 0.0, "step": 62340 }, { "epoch": 1.21, "learning_rate": 2.984071838466963e-05, "loss": 0.0, "step": 62342 }, { "epoch": 1.21, "learning_rate": 2.9840071424412404e-05, "loss": 0.0, "step": 62344 }, { "epoch": 1.21, "learning_rate": 2.983942446415517e-05, "loss": 0.0001, "step": 62346 }, { "epoch": 1.21, "learning_rate": 2.9838777503897935e-05, "loss": 0.0016, "step": 62348 }, { "epoch": 1.21, "learning_rate": 2.9838130543640708e-05, "loss": 0.0, "step": 62350 }, { "epoch": 1.21, "learning_rate": 2.9837483583383473e-05, "loss": 0.0, "step": 62352 }, { "epoch": 1.21, "learning_rate": 2.983683662312624e-05, "loss": 0.0, "step": 62354 }, { "epoch": 1.21, "learning_rate": 2.983618966286901e-05, "loss": 0.0, "step": 62356 }, { "epoch": 1.21, "learning_rate": 2.983554270261178e-05, "loss": 0.0, "step": 62358 }, { "epoch": 1.21, "learning_rate": 2.9834895742354546e-05, "loss": 0.0, "step": 62360 }, { "epoch": 1.21, "learning_rate": 2.983424878209732e-05, "loss": 0.0, "step": 62362 }, { "epoch": 1.21, "learning_rate": 2.9833601821840085e-05, "loss": 0.0, "step": 62364 }, { "epoch": 1.21, "learning_rate": 2.9832954861582857e-05, "loss": 0.0, "step": 62366 }, { "epoch": 1.21, "learning_rate": 2.9832307901325623e-05, "loss": 0.0049, "step": 62368 }, { "epoch": 1.21, "learning_rate": 2.983166094106839e-05, "loss": 0.0, "step": 62370 }, { "epoch": 1.21, "learning_rate": 2.983101398081116e-05, "loss": 0.0, "step": 62372 }, { "epoch": 1.21, "learning_rate": 2.983036702055393e-05, "loss": 0.0, "step": 62374 }, { "epoch": 1.21, "learning_rate": 2.9829720060296696e-05, "loss": 0.0, "step": 62376 }, { "epoch": 1.21, "learning_rate": 2.9829073100039468e-05, "loss": 0.0, "step": 62378 }, { "epoch": 1.21, "learning_rate": 2.9828426139782234e-05, "loss": 0.0009, "step": 62380 }, { "epoch": 1.21, "learning_rate": 2.9827779179525e-05, "loss": 0.0, "step": 62382 }, { "epoch": 1.21, "learning_rate": 2.9827132219267772e-05, "loss": 0.0002, "step": 62384 }, { "epoch": 1.21, "learning_rate": 2.9826485259010538e-05, "loss": 0.0001, "step": 62386 }, { "epoch": 1.21, "learning_rate": 2.982583829875331e-05, "loss": 0.0001, "step": 62388 }, { "epoch": 1.21, "learning_rate": 2.982519133849608e-05, "loss": 0.0, "step": 62390 }, { "epoch": 1.21, "learning_rate": 2.9824544378238845e-05, "loss": 0.0, "step": 62392 }, { "epoch": 1.21, "learning_rate": 2.9823897417981617e-05, "loss": 0.0021, "step": 62394 }, { "epoch": 1.21, "learning_rate": 2.9823250457724383e-05, "loss": 0.0, "step": 62396 }, { "epoch": 1.21, "learning_rate": 2.982260349746715e-05, "loss": 0.0, "step": 62398 }, { "epoch": 1.21, "learning_rate": 2.982195653720992e-05, "loss": 0.0, "step": 62400 }, { "epoch": 1.21, "learning_rate": 2.9821309576952687e-05, "loss": 0.0, "step": 62402 }, { "epoch": 1.21, "learning_rate": 2.982066261669546e-05, "loss": 0.0007, "step": 62404 }, { "epoch": 1.21, "learning_rate": 2.982001565643823e-05, "loss": 0.0, "step": 62406 }, { "epoch": 1.21, "learning_rate": 2.9819368696180994e-05, "loss": 0.0001, "step": 62408 }, { "epoch": 1.21, "learning_rate": 2.9818721735923767e-05, "loss": 0.0, "step": 62410 }, { "epoch": 1.21, "learning_rate": 2.9818074775666532e-05, "loss": 0.0003, "step": 62412 }, { "epoch": 1.21, "learning_rate": 2.9817427815409298e-05, "loss": 0.0, "step": 62414 }, { "epoch": 1.21, "learning_rate": 2.981678085515207e-05, "loss": 0.0, "step": 62416 }, { "epoch": 1.21, "learning_rate": 2.9816133894894836e-05, "loss": 0.0175, "step": 62418 }, { "epoch": 1.21, "learning_rate": 2.9815486934637605e-05, "loss": 0.0011, "step": 62420 }, { "epoch": 1.21, "learning_rate": 2.9814839974380378e-05, "loss": 0.0, "step": 62422 }, { "epoch": 1.21, "learning_rate": 2.9814193014123144e-05, "loss": 0.0, "step": 62424 }, { "epoch": 1.21, "learning_rate": 2.9813546053865916e-05, "loss": 0.0001, "step": 62426 }, { "epoch": 1.21, "learning_rate": 2.9812899093608682e-05, "loss": 0.0, "step": 62428 }, { "epoch": 1.21, "learning_rate": 2.9812252133351447e-05, "loss": 0.0, "step": 62430 }, { "epoch": 1.21, "learning_rate": 2.981160517309422e-05, "loss": 0.0, "step": 62432 }, { "epoch": 1.21, "learning_rate": 2.9810958212836986e-05, "loss": 0.0, "step": 62434 }, { "epoch": 1.21, "learning_rate": 2.9810311252579755e-05, "loss": 0.0, "step": 62436 }, { "epoch": 1.21, "learning_rate": 2.9809664292322527e-05, "loss": 0.0, "step": 62438 }, { "epoch": 1.21, "learning_rate": 2.9809017332065293e-05, "loss": 0.0, "step": 62440 }, { "epoch": 1.21, "learning_rate": 2.980837037180806e-05, "loss": 0.0001, "step": 62442 }, { "epoch": 1.21, "learning_rate": 2.980772341155083e-05, "loss": 0.0, "step": 62444 }, { "epoch": 1.21, "learning_rate": 2.9807076451293597e-05, "loss": 0.0001, "step": 62446 }, { "epoch": 1.21, "learning_rate": 2.980642949103637e-05, "loss": 0.0, "step": 62448 }, { "epoch": 1.21, "learning_rate": 2.9805782530779135e-05, "loss": 0.0, "step": 62450 }, { "epoch": 1.21, "learning_rate": 2.9805135570521904e-05, "loss": 0.0, "step": 62452 }, { "epoch": 1.21, "learning_rate": 2.9804488610264676e-05, "loss": 0.0, "step": 62454 }, { "epoch": 1.21, "learning_rate": 2.9803841650007442e-05, "loss": 0.0, "step": 62456 }, { "epoch": 1.21, "learning_rate": 2.9803194689750208e-05, "loss": 0.0001, "step": 62458 }, { "epoch": 1.21, "learning_rate": 2.980254772949298e-05, "loss": 0.0, "step": 62460 }, { "epoch": 1.21, "learning_rate": 2.9801900769235746e-05, "loss": 0.0029, "step": 62462 }, { "epoch": 1.21, "learning_rate": 2.980125380897851e-05, "loss": 0.0, "step": 62464 }, { "epoch": 1.21, "learning_rate": 2.9800606848721284e-05, "loss": 0.0104, "step": 62466 }, { "epoch": 1.21, "learning_rate": 2.9799959888464053e-05, "loss": 0.0, "step": 62468 }, { "epoch": 1.21, "learning_rate": 2.9799312928206822e-05, "loss": 0.0, "step": 62470 }, { "epoch": 1.21, "learning_rate": 2.979866596794959e-05, "loss": 0.0, "step": 62472 }, { "epoch": 1.21, "learning_rate": 2.9798019007692357e-05, "loss": 0.0001, "step": 62474 }, { "epoch": 1.21, "learning_rate": 2.979737204743513e-05, "loss": 0.0, "step": 62476 }, { "epoch": 1.21, "learning_rate": 2.9796725087177895e-05, "loss": 0.0001, "step": 62478 }, { "epoch": 1.21, "learning_rate": 2.979607812692066e-05, "loss": 0.0, "step": 62480 }, { "epoch": 1.21, "learning_rate": 2.9795431166663433e-05, "loss": 0.0, "step": 62482 }, { "epoch": 1.21, "learning_rate": 2.9794784206406203e-05, "loss": 0.0, "step": 62484 }, { "epoch": 1.21, "learning_rate": 2.979413724614897e-05, "loss": 0.0, "step": 62486 }, { "epoch": 1.21, "learning_rate": 2.979349028589174e-05, "loss": 0.0, "step": 62488 }, { "epoch": 1.21, "learning_rate": 2.9792843325634506e-05, "loss": 0.0, "step": 62490 }, { "epoch": 1.21, "learning_rate": 2.979219636537728e-05, "loss": 0.0, "step": 62492 }, { "epoch": 1.21, "learning_rate": 2.9791549405120045e-05, "loss": 0.0018, "step": 62494 }, { "epoch": 1.21, "learning_rate": 2.979090244486281e-05, "loss": 0.0, "step": 62496 }, { "epoch": 1.21, "learning_rate": 2.9790255484605583e-05, "loss": 0.0, "step": 62498 }, { "epoch": 1.21, "learning_rate": 2.9789608524348352e-05, "loss": 0.0001, "step": 62500 }, { "epoch": 1.21, "learning_rate": 2.9788961564091118e-05, "loss": 0.0001, "step": 62502 }, { "epoch": 1.21, "learning_rate": 2.978831460383389e-05, "loss": 0.0, "step": 62504 }, { "epoch": 1.21, "learning_rate": 2.9787667643576656e-05, "loss": 0.0, "step": 62506 }, { "epoch": 1.21, "learning_rate": 2.9787020683319428e-05, "loss": 0.0004, "step": 62508 }, { "epoch": 1.21, "learning_rate": 2.9786373723062194e-05, "loss": 0.0, "step": 62510 }, { "epoch": 1.21, "learning_rate": 2.978572676280496e-05, "loss": 0.0, "step": 62512 }, { "epoch": 1.21, "learning_rate": 2.9785079802547732e-05, "loss": 0.0, "step": 62514 }, { "epoch": 1.21, "learning_rate": 2.97844328422905e-05, "loss": 0.0, "step": 62516 }, { "epoch": 1.21, "learning_rate": 2.9783785882033267e-05, "loss": 0.0, "step": 62518 }, { "epoch": 1.21, "learning_rate": 2.978313892177604e-05, "loss": 0.0001, "step": 62520 }, { "epoch": 1.21, "learning_rate": 2.9782491961518805e-05, "loss": 0.0, "step": 62522 }, { "epoch": 1.21, "learning_rate": 2.978184500126157e-05, "loss": 0.0032, "step": 62524 }, { "epoch": 1.21, "learning_rate": 2.9781198041004343e-05, "loss": 0.0, "step": 62526 }, { "epoch": 1.21, "learning_rate": 2.978055108074711e-05, "loss": 0.0002, "step": 62528 }, { "epoch": 1.21, "learning_rate": 2.977990412048988e-05, "loss": 0.0, "step": 62530 }, { "epoch": 1.21, "learning_rate": 2.9779257160232647e-05, "loss": 0.0, "step": 62532 }, { "epoch": 1.21, "learning_rate": 2.9778610199975416e-05, "loss": 0.0, "step": 62534 }, { "epoch": 1.21, "learning_rate": 2.977796323971819e-05, "loss": 0.0, "step": 62536 }, { "epoch": 1.21, "learning_rate": 2.9777316279460954e-05, "loss": 0.0026, "step": 62538 }, { "epoch": 1.21, "learning_rate": 2.977666931920372e-05, "loss": 0.0, "step": 62540 }, { "epoch": 1.21, "learning_rate": 2.9776022358946492e-05, "loss": 0.0, "step": 62542 }, { "epoch": 1.21, "learning_rate": 2.9775375398689258e-05, "loss": 0.0, "step": 62544 }, { "epoch": 1.21, "learning_rate": 2.977472843843203e-05, "loss": 0.0, "step": 62546 }, { "epoch": 1.21, "learning_rate": 2.9774081478174796e-05, "loss": 0.0, "step": 62548 }, { "epoch": 1.21, "learning_rate": 2.9773434517917565e-05, "loss": 0.0049, "step": 62550 }, { "epoch": 1.21, "learning_rate": 2.9772787557660338e-05, "loss": 0.0, "step": 62552 }, { "epoch": 1.21, "learning_rate": 2.9772140597403104e-05, "loss": 0.0001, "step": 62554 }, { "epoch": 1.21, "learning_rate": 2.977149363714587e-05, "loss": 0.0, "step": 62556 }, { "epoch": 1.21, "learning_rate": 2.9770846676888642e-05, "loss": 0.0068, "step": 62558 }, { "epoch": 1.21, "learning_rate": 2.9770199716631407e-05, "loss": 0.0, "step": 62560 }, { "epoch": 1.21, "learning_rate": 2.9769552756374176e-05, "loss": 0.0, "step": 62562 }, { "epoch": 1.21, "learning_rate": 2.9768905796116946e-05, "loss": 0.0001, "step": 62564 }, { "epoch": 1.21, "learning_rate": 2.9768258835859715e-05, "loss": 0.0, "step": 62566 }, { "epoch": 1.21, "learning_rate": 2.9767611875602487e-05, "loss": 0.0, "step": 62568 }, { "epoch": 1.21, "learning_rate": 2.9766964915345253e-05, "loss": 0.0051, "step": 62570 }, { "epoch": 1.21, "learning_rate": 2.976631795508802e-05, "loss": 0.0, "step": 62572 }, { "epoch": 1.21, "learning_rate": 2.976567099483079e-05, "loss": 0.0, "step": 62574 }, { "epoch": 1.21, "learning_rate": 2.9765024034573557e-05, "loss": 0.0, "step": 62576 }, { "epoch": 1.21, "learning_rate": 2.9764377074316326e-05, "loss": 0.0, "step": 62578 }, { "epoch": 1.21, "learning_rate": 2.9763730114059095e-05, "loss": 0.0062, "step": 62580 }, { "epoch": 1.21, "learning_rate": 2.9763083153801864e-05, "loss": 0.0, "step": 62582 }, { "epoch": 1.21, "learning_rate": 2.976243619354463e-05, "loss": 0.0018, "step": 62584 }, { "epoch": 1.21, "learning_rate": 2.9761789233287402e-05, "loss": 0.0, "step": 62586 }, { "epoch": 1.21, "learning_rate": 2.9761142273030168e-05, "loss": 0.0, "step": 62588 }, { "epoch": 1.21, "learning_rate": 2.976049531277294e-05, "loss": 0.0, "step": 62590 }, { "epoch": 1.21, "learning_rate": 2.9759848352515706e-05, "loss": 0.0, "step": 62592 }, { "epoch": 1.21, "learning_rate": 2.975920139225847e-05, "loss": 0.0, "step": 62594 }, { "epoch": 1.21, "learning_rate": 2.9758554432001244e-05, "loss": 0.0007, "step": 62596 }, { "epoch": 1.21, "learning_rate": 2.9757907471744013e-05, "loss": 0.0005, "step": 62598 }, { "epoch": 1.21, "learning_rate": 2.975726051148678e-05, "loss": 0.0002, "step": 62600 }, { "epoch": 1.22, "learning_rate": 2.975661355122955e-05, "loss": 0.0, "step": 62602 }, { "epoch": 1.22, "learning_rate": 2.9755966590972317e-05, "loss": 0.0001, "step": 62604 }, { "epoch": 1.22, "learning_rate": 2.9755319630715083e-05, "loss": 0.0, "step": 62606 }, { "epoch": 1.22, "learning_rate": 2.9754672670457855e-05, "loss": 0.0, "step": 62608 }, { "epoch": 1.22, "learning_rate": 2.975402571020062e-05, "loss": 0.0009, "step": 62610 }, { "epoch": 1.22, "learning_rate": 2.9753378749943393e-05, "loss": 0.0, "step": 62612 }, { "epoch": 1.22, "learning_rate": 2.9752731789686163e-05, "loss": 0.0001, "step": 62614 }, { "epoch": 1.22, "learning_rate": 2.9752084829428928e-05, "loss": 0.0, "step": 62616 }, { "epoch": 1.22, "learning_rate": 2.97514378691717e-05, "loss": 0.0, "step": 62618 }, { "epoch": 1.22, "learning_rate": 2.9750790908914466e-05, "loss": 0.0, "step": 62620 }, { "epoch": 1.22, "learning_rate": 2.9750143948657232e-05, "loss": 0.0, "step": 62622 }, { "epoch": 1.22, "learning_rate": 2.9749496988400005e-05, "loss": 0.0078, "step": 62624 }, { "epoch": 1.22, "learning_rate": 2.974885002814277e-05, "loss": 0.0, "step": 62626 }, { "epoch": 1.22, "learning_rate": 2.9748203067885543e-05, "loss": 0.0, "step": 62628 }, { "epoch": 1.22, "learning_rate": 2.9747556107628312e-05, "loss": 0.0, "step": 62630 }, { "epoch": 1.22, "learning_rate": 2.9746909147371078e-05, "loss": 0.0, "step": 62632 }, { "epoch": 1.22, "learning_rate": 2.974626218711385e-05, "loss": 0.0, "step": 62634 }, { "epoch": 1.22, "learning_rate": 2.9745615226856616e-05, "loss": 0.0, "step": 62636 }, { "epoch": 1.22, "learning_rate": 2.974496826659938e-05, "loss": 0.0, "step": 62638 }, { "epoch": 1.22, "learning_rate": 2.9744321306342154e-05, "loss": 0.0, "step": 62640 }, { "epoch": 1.22, "learning_rate": 2.974367434608492e-05, "loss": 0.0, "step": 62642 }, { "epoch": 1.22, "learning_rate": 2.974302738582769e-05, "loss": 0.0, "step": 62644 }, { "epoch": 1.22, "learning_rate": 2.974238042557046e-05, "loss": 0.0, "step": 62646 }, { "epoch": 1.22, "learning_rate": 2.9741733465313227e-05, "loss": 0.0001, "step": 62648 }, { "epoch": 1.22, "learning_rate": 2.9741086505056e-05, "loss": 0.0, "step": 62650 }, { "epoch": 1.22, "learning_rate": 2.9740439544798765e-05, "loss": 0.0, "step": 62652 }, { "epoch": 1.22, "learning_rate": 2.973979258454153e-05, "loss": 0.0001, "step": 62654 }, { "epoch": 1.22, "learning_rate": 2.9739145624284303e-05, "loss": 0.0, "step": 62656 }, { "epoch": 1.22, "learning_rate": 2.973849866402707e-05, "loss": 0.0001, "step": 62658 }, { "epoch": 1.22, "learning_rate": 2.9737851703769838e-05, "loss": 0.0019, "step": 62660 }, { "epoch": 1.22, "learning_rate": 2.973720474351261e-05, "loss": 0.0004, "step": 62662 }, { "epoch": 1.22, "learning_rate": 2.9736557783255376e-05, "loss": 0.0, "step": 62664 }, { "epoch": 1.22, "learning_rate": 2.9735910822998142e-05, "loss": 0.0, "step": 62666 }, { "epoch": 1.22, "learning_rate": 2.9735263862740914e-05, "loss": 0.0, "step": 62668 }, { "epoch": 1.22, "learning_rate": 2.973461690248368e-05, "loss": 0.0523, "step": 62670 }, { "epoch": 1.22, "learning_rate": 2.9733969942226452e-05, "loss": 0.0, "step": 62672 }, { "epoch": 1.22, "learning_rate": 2.9733322981969218e-05, "loss": 0.0001, "step": 62674 }, { "epoch": 1.22, "learning_rate": 2.9732676021711987e-05, "loss": 0.0, "step": 62676 }, { "epoch": 1.22, "learning_rate": 2.973202906145476e-05, "loss": 0.0004, "step": 62678 }, { "epoch": 1.22, "learning_rate": 2.9731382101197525e-05, "loss": 0.0, "step": 62680 }, { "epoch": 1.22, "learning_rate": 2.973073514094029e-05, "loss": 0.0, "step": 62682 }, { "epoch": 1.22, "learning_rate": 2.9730088180683064e-05, "loss": 0.0, "step": 62684 }, { "epoch": 1.22, "learning_rate": 2.972944122042583e-05, "loss": 0.0, "step": 62686 }, { "epoch": 1.22, "learning_rate": 2.9728794260168602e-05, "loss": 0.0, "step": 62688 }, { "epoch": 1.22, "learning_rate": 2.9728147299911367e-05, "loss": 0.0001, "step": 62690 }, { "epoch": 1.22, "learning_rate": 2.9727500339654136e-05, "loss": 0.0, "step": 62692 }, { "epoch": 1.22, "learning_rate": 2.972685337939691e-05, "loss": 0.0, "step": 62694 }, { "epoch": 1.22, "learning_rate": 2.9726206419139675e-05, "loss": 0.0074, "step": 62696 }, { "epoch": 1.22, "learning_rate": 2.972555945888244e-05, "loss": 0.0, "step": 62698 }, { "epoch": 1.22, "learning_rate": 2.9724912498625213e-05, "loss": 0.0, "step": 62700 }, { "epoch": 1.22, "learning_rate": 2.972426553836798e-05, "loss": 0.0081, "step": 62702 }, { "epoch": 1.22, "learning_rate": 2.9723618578110744e-05, "loss": 0.0003, "step": 62704 }, { "epoch": 1.22, "learning_rate": 2.9722971617853517e-05, "loss": 0.0, "step": 62706 }, { "epoch": 1.22, "learning_rate": 2.9722324657596286e-05, "loss": 0.0, "step": 62708 }, { "epoch": 1.22, "learning_rate": 2.9721677697339055e-05, "loss": 0.0, "step": 62710 }, { "epoch": 1.22, "learning_rate": 2.9721030737081824e-05, "loss": 0.0199, "step": 62712 }, { "epoch": 1.22, "learning_rate": 2.972038377682459e-05, "loss": 0.0, "step": 62714 }, { "epoch": 1.22, "learning_rate": 2.9719736816567362e-05, "loss": 0.0, "step": 62716 }, { "epoch": 1.22, "learning_rate": 2.9719089856310128e-05, "loss": 0.0, "step": 62718 }, { "epoch": 1.22, "learning_rate": 2.9718442896052894e-05, "loss": 0.0, "step": 62720 }, { "epoch": 1.22, "learning_rate": 2.9717795935795666e-05, "loss": 0.0, "step": 62722 }, { "epoch": 1.22, "learning_rate": 2.9717148975538435e-05, "loss": 0.0, "step": 62724 }, { "epoch": 1.22, "learning_rate": 2.97165020152812e-05, "loss": 0.0, "step": 62726 }, { "epoch": 1.22, "learning_rate": 2.9715855055023973e-05, "loss": 0.0, "step": 62728 }, { "epoch": 1.22, "learning_rate": 2.971520809476674e-05, "loss": 0.0, "step": 62730 }, { "epoch": 1.22, "learning_rate": 2.971456113450951e-05, "loss": 0.0, "step": 62732 }, { "epoch": 1.22, "learning_rate": 2.9713914174252277e-05, "loss": 0.0018, "step": 62734 }, { "epoch": 1.22, "learning_rate": 2.9713267213995043e-05, "loss": 0.0, "step": 62736 }, { "epoch": 1.22, "learning_rate": 2.9712620253737815e-05, "loss": 0.0, "step": 62738 }, { "epoch": 1.22, "learning_rate": 2.9711973293480584e-05, "loss": 0.0003, "step": 62740 }, { "epoch": 1.22, "learning_rate": 2.971132633322335e-05, "loss": 0.0, "step": 62742 }, { "epoch": 1.22, "learning_rate": 2.9710679372966123e-05, "loss": 0.0, "step": 62744 }, { "epoch": 1.22, "learning_rate": 2.9710032412708888e-05, "loss": 0.0, "step": 62746 }, { "epoch": 1.22, "learning_rate": 2.9709385452451654e-05, "loss": 0.0, "step": 62748 }, { "epoch": 1.22, "learning_rate": 2.9708738492194426e-05, "loss": 0.0, "step": 62750 }, { "epoch": 1.22, "learning_rate": 2.9708091531937192e-05, "loss": 0.0001, "step": 62752 }, { "epoch": 1.22, "learning_rate": 2.9707444571679965e-05, "loss": 0.0039, "step": 62754 }, { "epoch": 1.22, "learning_rate": 2.9706797611422734e-05, "loss": 0.0, "step": 62756 }, { "epoch": 1.22, "learning_rate": 2.97061506511655e-05, "loss": 0.0, "step": 62758 }, { "epoch": 1.22, "learning_rate": 2.9705503690908272e-05, "loss": 0.0002, "step": 62760 }, { "epoch": 1.22, "learning_rate": 2.9704856730651037e-05, "loss": 0.0, "step": 62762 }, { "epoch": 1.22, "learning_rate": 2.9704209770393803e-05, "loss": 0.0, "step": 62764 }, { "epoch": 1.22, "learning_rate": 2.9703562810136576e-05, "loss": 0.0, "step": 62766 }, { "epoch": 1.22, "learning_rate": 2.970291584987934e-05, "loss": 0.0, "step": 62768 }, { "epoch": 1.22, "learning_rate": 2.9702268889622114e-05, "loss": 0.0, "step": 62770 }, { "epoch": 1.22, "learning_rate": 2.970162192936488e-05, "loss": 0.0, "step": 62772 }, { "epoch": 1.22, "learning_rate": 2.970097496910765e-05, "loss": 0.0031, "step": 62774 }, { "epoch": 1.22, "learning_rate": 2.970032800885042e-05, "loss": 0.0, "step": 62776 }, { "epoch": 1.22, "learning_rate": 2.9699681048593187e-05, "loss": 0.0, "step": 62778 }, { "epoch": 1.22, "learning_rate": 2.9699034088335952e-05, "loss": 0.0, "step": 62780 }, { "epoch": 1.22, "learning_rate": 2.9698387128078725e-05, "loss": 0.0, "step": 62782 }, { "epoch": 1.22, "learning_rate": 2.969774016782149e-05, "loss": 0.0, "step": 62784 }, { "epoch": 1.22, "learning_rate": 2.969709320756426e-05, "loss": 0.0, "step": 62786 }, { "epoch": 1.22, "learning_rate": 2.969644624730703e-05, "loss": 0.0, "step": 62788 }, { "epoch": 1.22, "learning_rate": 2.9695799287049798e-05, "loss": 0.0, "step": 62790 }, { "epoch": 1.22, "learning_rate": 2.969515232679257e-05, "loss": 0.0082, "step": 62792 }, { "epoch": 1.22, "learning_rate": 2.9694505366535336e-05, "loss": 0.0001, "step": 62794 }, { "epoch": 1.22, "learning_rate": 2.9693858406278102e-05, "loss": 0.0, "step": 62796 }, { "epoch": 1.22, "learning_rate": 2.9693211446020874e-05, "loss": 0.0, "step": 62798 }, { "epoch": 1.22, "learning_rate": 2.969256448576364e-05, "loss": 0.0, "step": 62800 }, { "epoch": 1.22, "learning_rate": 2.969191752550641e-05, "loss": 0.0, "step": 62802 }, { "epoch": 1.22, "learning_rate": 2.9691270565249178e-05, "loss": 0.0, "step": 62804 }, { "epoch": 1.22, "learning_rate": 2.9690623604991947e-05, "loss": 0.0, "step": 62806 }, { "epoch": 1.22, "learning_rate": 2.9689976644734713e-05, "loss": 0.0108, "step": 62808 }, { "epoch": 1.22, "learning_rate": 2.9689329684477485e-05, "loss": 0.0106, "step": 62810 }, { "epoch": 1.22, "learning_rate": 2.968868272422025e-05, "loss": 0.0, "step": 62812 }, { "epoch": 1.22, "learning_rate": 2.9688035763963024e-05, "loss": 0.0006, "step": 62814 }, { "epoch": 1.22, "learning_rate": 2.968738880370579e-05, "loss": 0.0, "step": 62816 }, { "epoch": 1.22, "learning_rate": 2.968674184344856e-05, "loss": 0.0, "step": 62818 }, { "epoch": 1.22, "learning_rate": 2.9686094883191327e-05, "loss": 0.0, "step": 62820 }, { "epoch": 1.22, "learning_rate": 2.9685447922934096e-05, "loss": 0.0, "step": 62822 }, { "epoch": 1.22, "learning_rate": 2.9684800962676862e-05, "loss": 0.004, "step": 62824 }, { "epoch": 1.22, "learning_rate": 2.9684154002419635e-05, "loss": 0.0, "step": 62826 }, { "epoch": 1.22, "learning_rate": 2.96835070421624e-05, "loss": 0.0, "step": 62828 }, { "epoch": 1.22, "learning_rate": 2.9682860081905173e-05, "loss": 0.0, "step": 62830 }, { "epoch": 1.22, "learning_rate": 2.968221312164794e-05, "loss": 0.0, "step": 62832 }, { "epoch": 1.22, "learning_rate": 2.9681566161390704e-05, "loss": 0.0, "step": 62834 }, { "epoch": 1.22, "learning_rate": 2.9680919201133477e-05, "loss": 0.0001, "step": 62836 }, { "epoch": 1.22, "learning_rate": 2.9680272240876246e-05, "loss": 0.0087, "step": 62838 }, { "epoch": 1.22, "learning_rate": 2.967962528061901e-05, "loss": 0.0058, "step": 62840 }, { "epoch": 1.22, "learning_rate": 2.9678978320361784e-05, "loss": 0.0, "step": 62842 }, { "epoch": 1.22, "learning_rate": 2.967833136010455e-05, "loss": 0.0, "step": 62844 }, { "epoch": 1.22, "learning_rate": 2.9677684399847315e-05, "loss": 0.0, "step": 62846 }, { "epoch": 1.22, "learning_rate": 2.9677037439590088e-05, "loss": 0.0, "step": 62848 }, { "epoch": 1.22, "learning_rate": 2.9676390479332853e-05, "loss": 0.0008, "step": 62850 }, { "epoch": 1.22, "learning_rate": 2.9675743519075626e-05, "loss": 0.0008, "step": 62852 }, { "epoch": 1.22, "learning_rate": 2.9675096558818395e-05, "loss": 0.0, "step": 62854 }, { "epoch": 1.22, "learning_rate": 2.967444959856116e-05, "loss": 0.0003, "step": 62856 }, { "epoch": 1.22, "learning_rate": 2.9673802638303933e-05, "loss": 0.0003, "step": 62858 }, { "epoch": 1.22, "learning_rate": 2.96731556780467e-05, "loss": 0.0, "step": 62860 }, { "epoch": 1.22, "learning_rate": 2.9672508717789465e-05, "loss": 0.0, "step": 62862 }, { "epoch": 1.22, "learning_rate": 2.9671861757532237e-05, "loss": 0.0, "step": 62864 }, { "epoch": 1.22, "learning_rate": 2.9671214797275003e-05, "loss": 0.0, "step": 62866 }, { "epoch": 1.22, "learning_rate": 2.9670567837017772e-05, "loss": 0.0015, "step": 62868 }, { "epoch": 1.22, "learning_rate": 2.9669920876760544e-05, "loss": 0.0019, "step": 62870 }, { "epoch": 1.22, "learning_rate": 2.966927391650331e-05, "loss": 0.0, "step": 62872 }, { "epoch": 1.22, "learning_rate": 2.9668626956246083e-05, "loss": 0.0, "step": 62874 }, { "epoch": 1.22, "learning_rate": 2.9667979995988848e-05, "loss": 0.0002, "step": 62876 }, { "epoch": 1.22, "learning_rate": 2.9667333035731614e-05, "loss": 0.0, "step": 62878 }, { "epoch": 1.22, "learning_rate": 2.9666686075474386e-05, "loss": 0.0, "step": 62880 }, { "epoch": 1.22, "learning_rate": 2.9666039115217152e-05, "loss": 0.0, "step": 62882 }, { "epoch": 1.22, "learning_rate": 2.966539215495992e-05, "loss": 0.0015, "step": 62884 }, { "epoch": 1.22, "learning_rate": 2.9664745194702694e-05, "loss": 0.0002, "step": 62886 }, { "epoch": 1.22, "learning_rate": 2.966409823444546e-05, "loss": 0.0, "step": 62888 }, { "epoch": 1.22, "learning_rate": 2.9663451274188225e-05, "loss": 0.0004, "step": 62890 }, { "epoch": 1.22, "learning_rate": 2.9662804313930997e-05, "loss": 0.0001, "step": 62892 }, { "epoch": 1.22, "learning_rate": 2.9662157353673763e-05, "loss": 0.0001, "step": 62894 }, { "epoch": 1.22, "learning_rate": 2.9661510393416536e-05, "loss": 0.0007, "step": 62896 }, { "epoch": 1.22, "learning_rate": 2.96608634331593e-05, "loss": 0.0, "step": 62898 }, { "epoch": 1.22, "learning_rate": 2.966021647290207e-05, "loss": 0.0, "step": 62900 }, { "epoch": 1.22, "learning_rate": 2.9659569512644843e-05, "loss": 0.0001, "step": 62902 }, { "epoch": 1.22, "learning_rate": 2.965892255238761e-05, "loss": 0.0, "step": 62904 }, { "epoch": 1.22, "learning_rate": 2.9658275592130374e-05, "loss": 0.0002, "step": 62906 }, { "epoch": 1.22, "learning_rate": 2.9657628631873147e-05, "loss": 0.0001, "step": 62908 }, { "epoch": 1.22, "learning_rate": 2.9656981671615912e-05, "loss": 0.0001, "step": 62910 }, { "epoch": 1.22, "learning_rate": 2.9656334711358685e-05, "loss": 0.0, "step": 62912 }, { "epoch": 1.22, "learning_rate": 2.965568775110145e-05, "loss": 0.0, "step": 62914 }, { "epoch": 1.22, "learning_rate": 2.965504079084422e-05, "loss": 0.0112, "step": 62916 }, { "epoch": 1.22, "learning_rate": 2.9654393830586992e-05, "loss": 0.0, "step": 62918 }, { "epoch": 1.22, "learning_rate": 2.9653746870329758e-05, "loss": 0.0, "step": 62920 }, { "epoch": 1.22, "learning_rate": 2.9653099910072524e-05, "loss": 0.0, "step": 62922 }, { "epoch": 1.22, "learning_rate": 2.9652452949815296e-05, "loss": 0.0, "step": 62924 }, { "epoch": 1.22, "learning_rate": 2.9651805989558062e-05, "loss": 0.0, "step": 62926 }, { "epoch": 1.22, "learning_rate": 2.9651159029300827e-05, "loss": 0.0003, "step": 62928 }, { "epoch": 1.22, "learning_rate": 2.96505120690436e-05, "loss": 0.0, "step": 62930 }, { "epoch": 1.22, "learning_rate": 2.964986510878637e-05, "loss": 0.0, "step": 62932 }, { "epoch": 1.22, "learning_rate": 2.964921814852914e-05, "loss": 0.0039, "step": 62934 }, { "epoch": 1.22, "learning_rate": 2.9648571188271907e-05, "loss": 0.0, "step": 62936 }, { "epoch": 1.22, "learning_rate": 2.9647924228014673e-05, "loss": 0.0, "step": 62938 }, { "epoch": 1.22, "learning_rate": 2.9647277267757445e-05, "loss": 0.0001, "step": 62940 }, { "epoch": 1.22, "learning_rate": 2.964663030750021e-05, "loss": 0.0, "step": 62942 }, { "epoch": 1.22, "learning_rate": 2.9645983347242977e-05, "loss": 0.0, "step": 62944 }, { "epoch": 1.22, "learning_rate": 2.964533638698575e-05, "loss": 0.0074, "step": 62946 }, { "epoch": 1.22, "learning_rate": 2.9644689426728518e-05, "loss": 0.0001, "step": 62948 }, { "epoch": 1.22, "learning_rate": 2.9644042466471284e-05, "loss": 0.0, "step": 62950 }, { "epoch": 1.22, "learning_rate": 2.9643395506214056e-05, "loss": 0.0003, "step": 62952 }, { "epoch": 1.22, "learning_rate": 2.9642748545956822e-05, "loss": 0.0, "step": 62954 }, { "epoch": 1.22, "learning_rate": 2.9642101585699595e-05, "loss": 0.0232, "step": 62956 }, { "epoch": 1.22, "learning_rate": 2.964145462544236e-05, "loss": 0.0, "step": 62958 }, { "epoch": 1.22, "learning_rate": 2.9640807665185126e-05, "loss": 0.0, "step": 62960 }, { "epoch": 1.22, "learning_rate": 2.96401607049279e-05, "loss": 0.0, "step": 62962 }, { "epoch": 1.22, "learning_rate": 2.9639513744670668e-05, "loss": 0.0001, "step": 62964 }, { "epoch": 1.22, "learning_rate": 2.9638866784413433e-05, "loss": 0.0029, "step": 62966 }, { "epoch": 1.22, "learning_rate": 2.9638219824156206e-05, "loss": 0.0067, "step": 62968 }, { "epoch": 1.22, "learning_rate": 2.963757286389897e-05, "loss": 0.0, "step": 62970 }, { "epoch": 1.22, "learning_rate": 2.9636925903641737e-05, "loss": 0.0193, "step": 62972 }, { "epoch": 1.22, "learning_rate": 2.963627894338451e-05, "loss": 0.0, "step": 62974 }, { "epoch": 1.22, "learning_rate": 2.9635631983127275e-05, "loss": 0.0, "step": 62976 }, { "epoch": 1.22, "learning_rate": 2.9634985022870048e-05, "loss": 0.0, "step": 62978 }, { "epoch": 1.22, "learning_rate": 2.9634338062612817e-05, "loss": 0.0002, "step": 62980 }, { "epoch": 1.22, "learning_rate": 2.9633691102355583e-05, "loss": 0.0011, "step": 62982 }, { "epoch": 1.22, "learning_rate": 2.9633044142098355e-05, "loss": 0.0001, "step": 62984 }, { "epoch": 1.22, "learning_rate": 2.963239718184112e-05, "loss": 0.0, "step": 62986 }, { "epoch": 1.22, "learning_rate": 2.9631750221583886e-05, "loss": 0.0, "step": 62988 }, { "epoch": 1.22, "learning_rate": 2.963110326132666e-05, "loss": 0.0, "step": 62990 }, { "epoch": 1.22, "learning_rate": 2.9630456301069425e-05, "loss": 0.0, "step": 62992 }, { "epoch": 1.22, "learning_rate": 2.9629809340812197e-05, "loss": 0.0, "step": 62994 }, { "epoch": 1.22, "learning_rate": 2.9629162380554966e-05, "loss": 0.0, "step": 62996 }, { "epoch": 1.22, "learning_rate": 2.9628515420297732e-05, "loss": 0.0, "step": 62998 }, { "epoch": 1.22, "learning_rate": 2.9627868460040504e-05, "loss": 0.0, "step": 63000 }, { "epoch": 1.22, "learning_rate": 2.962722149978327e-05, "loss": 0.0004, "step": 63002 }, { "epoch": 1.22, "learning_rate": 2.9626574539526036e-05, "loss": 0.0, "step": 63004 }, { "epoch": 1.22, "learning_rate": 2.9625927579268808e-05, "loss": 0.0, "step": 63006 }, { "epoch": 1.22, "learning_rate": 2.9625280619011574e-05, "loss": 0.0, "step": 63008 }, { "epoch": 1.22, "learning_rate": 2.9624633658754343e-05, "loss": 0.0, "step": 63010 }, { "epoch": 1.22, "learning_rate": 2.9623986698497112e-05, "loss": 0.0, "step": 63012 }, { "epoch": 1.22, "learning_rate": 2.962333973823988e-05, "loss": 0.0, "step": 63014 }, { "epoch": 1.22, "learning_rate": 2.9622692777982654e-05, "loss": 0.0, "step": 63016 }, { "epoch": 1.22, "learning_rate": 2.962204581772542e-05, "loss": 0.0, "step": 63018 }, { "epoch": 1.22, "learning_rate": 2.9621398857468185e-05, "loss": 0.0001, "step": 63020 }, { "epoch": 1.22, "learning_rate": 2.9620751897210957e-05, "loss": 0.0, "step": 63022 }, { "epoch": 1.22, "learning_rate": 2.9620104936953723e-05, "loss": 0.0, "step": 63024 }, { "epoch": 1.22, "learning_rate": 2.9619457976696492e-05, "loss": 0.0002, "step": 63026 }, { "epoch": 1.22, "learning_rate": 2.961881101643926e-05, "loss": 0.0002, "step": 63028 }, { "epoch": 1.22, "learning_rate": 2.961816405618203e-05, "loss": 0.0, "step": 63030 }, { "epoch": 1.22, "learning_rate": 2.9617517095924796e-05, "loss": 0.0003, "step": 63032 }, { "epoch": 1.22, "learning_rate": 2.961687013566757e-05, "loss": 0.0, "step": 63034 }, { "epoch": 1.22, "learning_rate": 2.9616223175410334e-05, "loss": 0.0001, "step": 63036 }, { "epoch": 1.22, "learning_rate": 2.9615576215153107e-05, "loss": 0.0, "step": 63038 }, { "epoch": 1.22, "learning_rate": 2.9614929254895872e-05, "loss": 0.0, "step": 63040 }, { "epoch": 1.22, "learning_rate": 2.961428229463864e-05, "loss": 0.0, "step": 63042 }, { "epoch": 1.22, "learning_rate": 2.961363533438141e-05, "loss": 0.0, "step": 63044 }, { "epoch": 1.22, "learning_rate": 2.961298837412418e-05, "loss": 0.0, "step": 63046 }, { "epoch": 1.22, "learning_rate": 2.9612341413866945e-05, "loss": 0.0, "step": 63048 }, { "epoch": 1.22, "learning_rate": 2.9611694453609718e-05, "loss": 0.0, "step": 63050 }, { "epoch": 1.22, "learning_rate": 2.9611047493352484e-05, "loss": 0.0347, "step": 63052 }, { "epoch": 1.22, "learning_rate": 2.9610400533095256e-05, "loss": 0.0, "step": 63054 }, { "epoch": 1.22, "learning_rate": 2.9609753572838022e-05, "loss": 0.0, "step": 63056 }, { "epoch": 1.22, "learning_rate": 2.960910661258079e-05, "loss": 0.0005, "step": 63058 }, { "epoch": 1.22, "learning_rate": 2.960845965232356e-05, "loss": 0.0, "step": 63060 }, { "epoch": 1.22, "learning_rate": 2.960781269206633e-05, "loss": 0.0, "step": 63062 }, { "epoch": 1.22, "learning_rate": 2.9607165731809095e-05, "loss": 0.0, "step": 63064 }, { "epoch": 1.22, "learning_rate": 2.9606518771551867e-05, "loss": 0.0, "step": 63066 }, { "epoch": 1.22, "learning_rate": 2.9605871811294633e-05, "loss": 0.0, "step": 63068 }, { "epoch": 1.22, "learning_rate": 2.96052248510374e-05, "loss": 0.0, "step": 63070 }, { "epoch": 1.22, "learning_rate": 2.960457789078017e-05, "loss": 0.0, "step": 63072 }, { "epoch": 1.22, "learning_rate": 2.9603930930522937e-05, "loss": 0.0, "step": 63074 }, { "epoch": 1.22, "learning_rate": 2.960328397026571e-05, "loss": 0.0, "step": 63076 }, { "epoch": 1.22, "learning_rate": 2.9602637010008478e-05, "loss": 0.0, "step": 63078 }, { "epoch": 1.22, "learning_rate": 2.9601990049751244e-05, "loss": 0.0, "step": 63080 }, { "epoch": 1.22, "learning_rate": 2.9601343089494016e-05, "loss": 0.0, "step": 63082 }, { "epoch": 1.22, "learning_rate": 2.9600696129236782e-05, "loss": 0.0001, "step": 63084 }, { "epoch": 1.22, "learning_rate": 2.9600049168979548e-05, "loss": 0.0, "step": 63086 }, { "epoch": 1.22, "learning_rate": 2.959940220872232e-05, "loss": 0.0001, "step": 63088 }, { "epoch": 1.22, "learning_rate": 2.9598755248465086e-05, "loss": 0.0001, "step": 63090 }, { "epoch": 1.22, "learning_rate": 2.9598108288207855e-05, "loss": 0.0008, "step": 63092 }, { "epoch": 1.22, "learning_rate": 2.9597461327950628e-05, "loss": 0.0, "step": 63094 }, { "epoch": 1.22, "learning_rate": 2.9596814367693393e-05, "loss": 0.0, "step": 63096 }, { "epoch": 1.22, "learning_rate": 2.9596167407436166e-05, "loss": 0.001, "step": 63098 }, { "epoch": 1.22, "learning_rate": 2.959552044717893e-05, "loss": 0.0, "step": 63100 }, { "epoch": 1.22, "learning_rate": 2.9594873486921697e-05, "loss": 0.0, "step": 63102 }, { "epoch": 1.22, "learning_rate": 2.959422652666447e-05, "loss": 0.0001, "step": 63104 }, { "epoch": 1.22, "learning_rate": 2.9593579566407235e-05, "loss": 0.0016, "step": 63106 }, { "epoch": 1.22, "learning_rate": 2.9592932606150004e-05, "loss": 0.0003, "step": 63108 }, { "epoch": 1.22, "learning_rate": 2.9592285645892777e-05, "loss": 0.0441, "step": 63110 }, { "epoch": 1.22, "learning_rate": 2.9591638685635543e-05, "loss": 0.0, "step": 63112 }, { "epoch": 1.22, "learning_rate": 2.9590991725378308e-05, "loss": 0.0007, "step": 63114 }, { "epoch": 1.23, "learning_rate": 2.959034476512108e-05, "loss": 0.0055, "step": 63116 }, { "epoch": 1.23, "learning_rate": 2.9589697804863846e-05, "loss": 0.0029, "step": 63118 }, { "epoch": 1.23, "learning_rate": 2.958905084460662e-05, "loss": 0.0, "step": 63120 }, { "epoch": 1.23, "learning_rate": 2.9588403884349385e-05, "loss": 0.0002, "step": 63122 }, { "epoch": 1.23, "learning_rate": 2.9587756924092154e-05, "loss": 0.0, "step": 63124 }, { "epoch": 1.23, "learning_rate": 2.9587109963834926e-05, "loss": 0.0001, "step": 63126 }, { "epoch": 1.23, "learning_rate": 2.9586463003577692e-05, "loss": 0.0, "step": 63128 }, { "epoch": 1.23, "learning_rate": 2.9585816043320458e-05, "loss": 0.0, "step": 63130 }, { "epoch": 1.23, "learning_rate": 2.958516908306323e-05, "loss": 0.0, "step": 63132 }, { "epoch": 1.23, "learning_rate": 2.9584522122805996e-05, "loss": 0.0, "step": 63134 }, { "epoch": 1.23, "learning_rate": 2.9583875162548768e-05, "loss": 0.0, "step": 63136 }, { "epoch": 1.23, "learning_rate": 2.9583228202291534e-05, "loss": 0.0, "step": 63138 }, { "epoch": 1.23, "learning_rate": 2.9582581242034303e-05, "loss": 0.0, "step": 63140 }, { "epoch": 1.23, "learning_rate": 2.9581934281777075e-05, "loss": 0.0, "step": 63142 }, { "epoch": 1.23, "learning_rate": 2.958128732151984e-05, "loss": 0.0, "step": 63144 }, { "epoch": 1.23, "learning_rate": 2.9580640361262607e-05, "loss": 0.0002, "step": 63146 }, { "epoch": 1.23, "learning_rate": 2.957999340100538e-05, "loss": 0.0002, "step": 63148 }, { "epoch": 1.23, "learning_rate": 2.9579346440748145e-05, "loss": 0.0, "step": 63150 }, { "epoch": 1.23, "learning_rate": 2.957869948049091e-05, "loss": 0.0001, "step": 63152 }, { "epoch": 1.23, "learning_rate": 2.9578052520233683e-05, "loss": 0.0004, "step": 63154 }, { "epoch": 1.23, "learning_rate": 2.9577405559976452e-05, "loss": 0.0, "step": 63156 }, { "epoch": 1.23, "learning_rate": 2.9576758599719225e-05, "loss": 0.0025, "step": 63158 }, { "epoch": 1.23, "learning_rate": 2.957611163946199e-05, "loss": 0.0, "step": 63160 }, { "epoch": 1.23, "learning_rate": 2.9575464679204756e-05, "loss": 0.0002, "step": 63162 }, { "epoch": 1.23, "learning_rate": 2.957481771894753e-05, "loss": 0.0134, "step": 63164 }, { "epoch": 1.23, "learning_rate": 2.9574170758690294e-05, "loss": 0.0, "step": 63166 }, { "epoch": 1.23, "learning_rate": 2.957352379843306e-05, "loss": 0.0021, "step": 63168 }, { "epoch": 1.23, "learning_rate": 2.9572876838175832e-05, "loss": 0.0, "step": 63170 }, { "epoch": 1.23, "learning_rate": 2.95722298779186e-05, "loss": 0.0001, "step": 63172 }, { "epoch": 1.23, "learning_rate": 2.9571582917661367e-05, "loss": 0.0, "step": 63174 }, { "epoch": 1.23, "learning_rate": 2.957093595740414e-05, "loss": 0.0013, "step": 63176 }, { "epoch": 1.23, "learning_rate": 2.9570288997146905e-05, "loss": 0.0002, "step": 63178 }, { "epoch": 1.23, "learning_rate": 2.9569642036889678e-05, "loss": 0.0001, "step": 63180 }, { "epoch": 1.23, "learning_rate": 2.9568995076632444e-05, "loss": 0.0115, "step": 63182 }, { "epoch": 1.23, "learning_rate": 2.956834811637521e-05, "loss": 0.0005, "step": 63184 }, { "epoch": 1.23, "learning_rate": 2.9567701156117982e-05, "loss": 0.0, "step": 63186 }, { "epoch": 1.23, "learning_rate": 2.956705419586075e-05, "loss": 0.0, "step": 63188 }, { "epoch": 1.23, "learning_rate": 2.9566407235603517e-05, "loss": 0.0031, "step": 63190 }, { "epoch": 1.23, "learning_rate": 2.956576027534629e-05, "loss": 0.0, "step": 63192 }, { "epoch": 1.23, "learning_rate": 2.9565113315089055e-05, "loss": 0.0, "step": 63194 }, { "epoch": 1.23, "learning_rate": 2.9564466354831827e-05, "loss": 0.0, "step": 63196 }, { "epoch": 1.23, "learning_rate": 2.9563819394574593e-05, "loss": 0.0007, "step": 63198 }, { "epoch": 1.23, "learning_rate": 2.956317243431736e-05, "loss": 0.0, "step": 63200 }, { "epoch": 1.23, "learning_rate": 2.956252547406013e-05, "loss": 0.0, "step": 63202 }, { "epoch": 1.23, "learning_rate": 2.95618785138029e-05, "loss": 0.0001, "step": 63204 }, { "epoch": 1.23, "learning_rate": 2.9561231553545666e-05, "loss": 0.0, "step": 63206 }, { "epoch": 1.23, "learning_rate": 2.9560584593288438e-05, "loss": 0.0, "step": 63208 }, { "epoch": 1.23, "learning_rate": 2.9559937633031204e-05, "loss": 0.0062, "step": 63210 }, { "epoch": 1.23, "learning_rate": 2.955929067277397e-05, "loss": 0.0, "step": 63212 }, { "epoch": 1.23, "learning_rate": 2.9558643712516742e-05, "loss": 0.0, "step": 63214 }, { "epoch": 1.23, "learning_rate": 2.9557996752259508e-05, "loss": 0.0, "step": 63216 }, { "epoch": 1.23, "learning_rate": 2.955734979200228e-05, "loss": 0.0, "step": 63218 }, { "epoch": 1.23, "learning_rate": 2.955670283174505e-05, "loss": 0.0, "step": 63220 }, { "epoch": 1.23, "learning_rate": 2.9556055871487815e-05, "loss": 0.0005, "step": 63222 }, { "epoch": 1.23, "learning_rate": 2.9555408911230588e-05, "loss": 0.0, "step": 63224 }, { "epoch": 1.23, "learning_rate": 2.9554761950973353e-05, "loss": 0.0, "step": 63226 }, { "epoch": 1.23, "learning_rate": 2.955411499071612e-05, "loss": 0.0, "step": 63228 }, { "epoch": 1.23, "learning_rate": 2.955346803045889e-05, "loss": 0.0, "step": 63230 }, { "epoch": 1.23, "learning_rate": 2.9552821070201657e-05, "loss": 0.0, "step": 63232 }, { "epoch": 1.23, "learning_rate": 2.9552174109944426e-05, "loss": 0.0004, "step": 63234 }, { "epoch": 1.23, "learning_rate": 2.95515271496872e-05, "loss": 0.0, "step": 63236 }, { "epoch": 1.23, "learning_rate": 2.9550880189429964e-05, "loss": 0.0, "step": 63238 }, { "epoch": 1.23, "learning_rate": 2.9550233229172737e-05, "loss": 0.0, "step": 63240 }, { "epoch": 1.23, "learning_rate": 2.9549586268915503e-05, "loss": 0.0, "step": 63242 }, { "epoch": 1.23, "learning_rate": 2.9548939308658268e-05, "loss": 0.0, "step": 63244 }, { "epoch": 1.23, "learning_rate": 2.954829234840104e-05, "loss": 0.0004, "step": 63246 }, { "epoch": 1.23, "learning_rate": 2.9547645388143806e-05, "loss": 0.013, "step": 63248 }, { "epoch": 1.23, "learning_rate": 2.9546998427886575e-05, "loss": 0.0, "step": 63250 }, { "epoch": 1.23, "learning_rate": 2.9546351467629345e-05, "loss": 0.0, "step": 63252 }, { "epoch": 1.23, "learning_rate": 2.9545704507372114e-05, "loss": 0.0, "step": 63254 }, { "epoch": 1.23, "learning_rate": 2.954505754711488e-05, "loss": 0.0, "step": 63256 }, { "epoch": 1.23, "learning_rate": 2.9544410586857652e-05, "loss": 0.0015, "step": 63258 }, { "epoch": 1.23, "learning_rate": 2.9543763626600418e-05, "loss": 0.0, "step": 63260 }, { "epoch": 1.23, "learning_rate": 2.954311666634319e-05, "loss": 0.0006, "step": 63262 }, { "epoch": 1.23, "learning_rate": 2.9542469706085956e-05, "loss": 0.0, "step": 63264 }, { "epoch": 1.23, "learning_rate": 2.9541822745828725e-05, "loss": 0.0, "step": 63266 }, { "epoch": 1.23, "learning_rate": 2.9541175785571494e-05, "loss": 0.0, "step": 63268 }, { "epoch": 1.23, "learning_rate": 2.9540528825314263e-05, "loss": 0.0, "step": 63270 }, { "epoch": 1.23, "learning_rate": 2.953988186505703e-05, "loss": 0.0, "step": 63272 }, { "epoch": 1.23, "learning_rate": 2.95392349047998e-05, "loss": 0.0, "step": 63274 }, { "epoch": 1.23, "learning_rate": 2.9538587944542567e-05, "loss": 0.0, "step": 63276 }, { "epoch": 1.23, "learning_rate": 2.953794098428534e-05, "loss": 0.0, "step": 63278 }, { "epoch": 1.23, "learning_rate": 2.9537294024028105e-05, "loss": 0.0, "step": 63280 }, { "epoch": 1.23, "learning_rate": 2.9536647063770874e-05, "loss": 0.0, "step": 63282 }, { "epoch": 1.23, "learning_rate": 2.9536000103513643e-05, "loss": 0.0001, "step": 63284 }, { "epoch": 1.23, "learning_rate": 2.9535353143256412e-05, "loss": 0.0, "step": 63286 }, { "epoch": 1.23, "learning_rate": 2.9534706182999178e-05, "loss": 0.0056, "step": 63288 }, { "epoch": 1.23, "learning_rate": 2.953405922274195e-05, "loss": 0.0, "step": 63290 }, { "epoch": 1.23, "learning_rate": 2.9533412262484716e-05, "loss": 0.0062, "step": 63292 }, { "epoch": 1.23, "learning_rate": 2.9532765302227482e-05, "loss": 0.0034, "step": 63294 }, { "epoch": 1.23, "learning_rate": 2.9532118341970254e-05, "loss": 0.0, "step": 63296 }, { "epoch": 1.23, "learning_rate": 2.9531471381713023e-05, "loss": 0.0, "step": 63298 }, { "epoch": 1.23, "learning_rate": 2.9530824421455792e-05, "loss": 0.0, "step": 63300 }, { "epoch": 1.23, "learning_rate": 2.953017746119856e-05, "loss": 0.0004, "step": 63302 }, { "epoch": 1.23, "learning_rate": 2.9529530500941327e-05, "loss": 0.0, "step": 63304 }, { "epoch": 1.23, "learning_rate": 2.95288835406841e-05, "loss": 0.0, "step": 63306 }, { "epoch": 1.23, "learning_rate": 2.9528236580426865e-05, "loss": 0.0, "step": 63308 }, { "epoch": 1.23, "learning_rate": 2.952758962016963e-05, "loss": 0.0, "step": 63310 }, { "epoch": 1.23, "learning_rate": 2.9526942659912404e-05, "loss": 0.0, "step": 63312 }, { "epoch": 1.23, "learning_rate": 2.952629569965517e-05, "loss": 0.0, "step": 63314 }, { "epoch": 1.23, "learning_rate": 2.952564873939794e-05, "loss": 0.0, "step": 63316 }, { "epoch": 1.23, "learning_rate": 2.952500177914071e-05, "loss": 0.0, "step": 63318 }, { "epoch": 1.23, "learning_rate": 2.9524354818883477e-05, "loss": 0.0, "step": 63320 }, { "epoch": 1.23, "learning_rate": 2.952370785862625e-05, "loss": 0.0, "step": 63322 }, { "epoch": 1.23, "learning_rate": 2.9523060898369015e-05, "loss": 0.0017, "step": 63324 }, { "epoch": 1.23, "learning_rate": 2.952241393811178e-05, "loss": 0.0, "step": 63326 }, { "epoch": 1.23, "learning_rate": 2.9521766977854553e-05, "loss": 0.0001, "step": 63328 }, { "epoch": 1.23, "learning_rate": 2.952112001759732e-05, "loss": 0.0, "step": 63330 }, { "epoch": 1.23, "learning_rate": 2.9520473057340088e-05, "loss": 0.0037, "step": 63332 }, { "epoch": 1.23, "learning_rate": 2.951982609708286e-05, "loss": 0.0, "step": 63334 }, { "epoch": 1.23, "learning_rate": 2.9519179136825626e-05, "loss": 0.0, "step": 63336 }, { "epoch": 1.23, "learning_rate": 2.9518532176568398e-05, "loss": 0.0001, "step": 63338 }, { "epoch": 1.23, "learning_rate": 2.9517885216311164e-05, "loss": 0.0, "step": 63340 }, { "epoch": 1.23, "learning_rate": 2.951723825605393e-05, "loss": 0.0, "step": 63342 }, { "epoch": 1.23, "learning_rate": 2.9516591295796702e-05, "loss": 0.0001, "step": 63344 }, { "epoch": 1.23, "learning_rate": 2.9515944335539468e-05, "loss": 0.0051, "step": 63346 }, { "epoch": 1.23, "learning_rate": 2.9515297375282237e-05, "loss": 0.0, "step": 63348 }, { "epoch": 1.23, "learning_rate": 2.951465041502501e-05, "loss": 0.0006, "step": 63350 }, { "epoch": 1.23, "learning_rate": 2.9514003454767775e-05, "loss": 0.0, "step": 63352 }, { "epoch": 1.23, "learning_rate": 2.951335649451054e-05, "loss": 0.0, "step": 63354 }, { "epoch": 1.23, "learning_rate": 2.9512709534253313e-05, "loss": 0.0001, "step": 63356 }, { "epoch": 1.23, "learning_rate": 2.951206257399608e-05, "loss": 0.0, "step": 63358 }, { "epoch": 1.23, "learning_rate": 2.951141561373885e-05, "loss": 0.0, "step": 63360 }, { "epoch": 1.23, "learning_rate": 2.9510768653481617e-05, "loss": 0.0, "step": 63362 }, { "epoch": 1.23, "learning_rate": 2.9510121693224386e-05, "loss": 0.0, "step": 63364 }, { "epoch": 1.23, "learning_rate": 2.950947473296716e-05, "loss": 0.0002, "step": 63366 }, { "epoch": 1.23, "learning_rate": 2.9508827772709924e-05, "loss": 0.0, "step": 63368 }, { "epoch": 1.23, "learning_rate": 2.950818081245269e-05, "loss": 0.0, "step": 63370 }, { "epoch": 1.23, "learning_rate": 2.9507533852195463e-05, "loss": 0.0, "step": 63372 }, { "epoch": 1.23, "learning_rate": 2.9506886891938228e-05, "loss": 0.0, "step": 63374 }, { "epoch": 1.23, "learning_rate": 2.9506239931680997e-05, "loss": 0.0024, "step": 63376 }, { "epoch": 1.23, "learning_rate": 2.9505592971423766e-05, "loss": 0.0001, "step": 63378 }, { "epoch": 1.23, "learning_rate": 2.9504946011166535e-05, "loss": 0.0001, "step": 63380 }, { "epoch": 1.23, "learning_rate": 2.9504299050909308e-05, "loss": 0.0, "step": 63382 }, { "epoch": 1.23, "learning_rate": 2.9503652090652074e-05, "loss": 0.0, "step": 63384 }, { "epoch": 1.23, "learning_rate": 2.950300513039484e-05, "loss": 0.0, "step": 63386 }, { "epoch": 1.23, "learning_rate": 2.9502358170137612e-05, "loss": 0.0, "step": 63388 }, { "epoch": 1.23, "learning_rate": 2.9501711209880378e-05, "loss": 0.0, "step": 63390 }, { "epoch": 1.23, "learning_rate": 2.9501064249623143e-05, "loss": 0.0001, "step": 63392 }, { "epoch": 1.23, "learning_rate": 2.9500417289365916e-05, "loss": 0.0, "step": 63394 }, { "epoch": 1.23, "learning_rate": 2.9499770329108685e-05, "loss": 0.0, "step": 63396 }, { "epoch": 1.23, "learning_rate": 2.949912336885145e-05, "loss": 0.0, "step": 63398 }, { "epoch": 1.23, "learning_rate": 2.9498476408594223e-05, "loss": 0.0, "step": 63400 }, { "epoch": 1.23, "learning_rate": 2.949782944833699e-05, "loss": 0.0, "step": 63402 }, { "epoch": 1.23, "learning_rate": 2.949718248807976e-05, "loss": 0.0001, "step": 63404 }, { "epoch": 1.23, "learning_rate": 2.9496535527822527e-05, "loss": 0.0, "step": 63406 }, { "epoch": 1.23, "learning_rate": 2.9495888567565293e-05, "loss": 0.0, "step": 63408 }, { "epoch": 1.23, "learning_rate": 2.9495241607308065e-05, "loss": 0.0, "step": 63410 }, { "epoch": 1.23, "learning_rate": 2.9494594647050834e-05, "loss": 0.0, "step": 63412 }, { "epoch": 1.23, "learning_rate": 2.94939476867936e-05, "loss": 0.0018, "step": 63414 }, { "epoch": 1.23, "learning_rate": 2.9493300726536372e-05, "loss": 0.0, "step": 63416 }, { "epoch": 1.23, "learning_rate": 2.9492653766279138e-05, "loss": 0.0, "step": 63418 }, { "epoch": 1.23, "learning_rate": 2.949200680602191e-05, "loss": 0.0, "step": 63420 }, { "epoch": 1.23, "learning_rate": 2.9491359845764676e-05, "loss": 0.0, "step": 63422 }, { "epoch": 1.23, "learning_rate": 2.9490712885507442e-05, "loss": 0.0001, "step": 63424 }, { "epoch": 1.23, "learning_rate": 2.9490065925250214e-05, "loss": 0.0, "step": 63426 }, { "epoch": 1.23, "learning_rate": 2.9489418964992983e-05, "loss": 0.0, "step": 63428 }, { "epoch": 1.23, "learning_rate": 2.948877200473575e-05, "loss": 0.0126, "step": 63430 }, { "epoch": 1.23, "learning_rate": 2.948812504447852e-05, "loss": 0.0, "step": 63432 }, { "epoch": 1.23, "learning_rate": 2.9487478084221287e-05, "loss": 0.0, "step": 63434 }, { "epoch": 1.23, "learning_rate": 2.9486831123964053e-05, "loss": 0.0, "step": 63436 }, { "epoch": 1.23, "learning_rate": 2.9486184163706825e-05, "loss": 0.0, "step": 63438 }, { "epoch": 1.23, "learning_rate": 2.948553720344959e-05, "loss": 0.0, "step": 63440 }, { "epoch": 1.23, "learning_rate": 2.9484890243192364e-05, "loss": 0.0001, "step": 63442 }, { "epoch": 1.23, "learning_rate": 2.9484243282935133e-05, "loss": 0.0, "step": 63444 }, { "epoch": 1.23, "learning_rate": 2.94835963226779e-05, "loss": 0.0, "step": 63446 }, { "epoch": 1.23, "learning_rate": 2.948294936242067e-05, "loss": 0.0, "step": 63448 }, { "epoch": 1.23, "learning_rate": 2.9482302402163437e-05, "loss": 0.0, "step": 63450 }, { "epoch": 1.23, "learning_rate": 2.9481655441906202e-05, "loss": 0.0, "step": 63452 }, { "epoch": 1.23, "learning_rate": 2.9481008481648975e-05, "loss": 0.0001, "step": 63454 }, { "epoch": 1.23, "learning_rate": 2.948036152139174e-05, "loss": 0.0, "step": 63456 }, { "epoch": 1.23, "learning_rate": 2.947971456113451e-05, "loss": 0.0, "step": 63458 }, { "epoch": 1.23, "learning_rate": 2.9479067600877282e-05, "loss": 0.0277, "step": 63460 }, { "epoch": 1.23, "learning_rate": 2.9478420640620048e-05, "loss": 0.0, "step": 63462 }, { "epoch": 1.23, "learning_rate": 2.947777368036282e-05, "loss": 0.0, "step": 63464 }, { "epoch": 1.23, "learning_rate": 2.9477126720105586e-05, "loss": 0.0021, "step": 63466 }, { "epoch": 1.23, "learning_rate": 2.947647975984835e-05, "loss": 0.0, "step": 63468 }, { "epoch": 1.23, "learning_rate": 2.9475832799591124e-05, "loss": 0.0, "step": 63470 }, { "epoch": 1.23, "learning_rate": 2.947518583933389e-05, "loss": 0.0028, "step": 63472 }, { "epoch": 1.23, "learning_rate": 2.947453887907666e-05, "loss": 0.0, "step": 63474 }, { "epoch": 1.23, "learning_rate": 2.947389191881943e-05, "loss": 0.0, "step": 63476 }, { "epoch": 1.23, "learning_rate": 2.9473244958562197e-05, "loss": 0.0, "step": 63478 }, { "epoch": 1.23, "learning_rate": 2.947259799830497e-05, "loss": 0.0, "step": 63480 }, { "epoch": 1.23, "learning_rate": 2.9471951038047735e-05, "loss": 0.0068, "step": 63482 }, { "epoch": 1.23, "learning_rate": 2.94713040777905e-05, "loss": 0.0083, "step": 63484 }, { "epoch": 1.23, "learning_rate": 2.9470657117533273e-05, "loss": 0.0, "step": 63486 }, { "epoch": 1.23, "learning_rate": 2.947001015727604e-05, "loss": 0.0, "step": 63488 }, { "epoch": 1.23, "learning_rate": 2.9469363197018808e-05, "loss": 0.0001, "step": 63490 }, { "epoch": 1.23, "learning_rate": 2.9468716236761577e-05, "loss": 0.0001, "step": 63492 }, { "epoch": 1.23, "learning_rate": 2.9468069276504346e-05, "loss": 0.0002, "step": 63494 }, { "epoch": 1.23, "learning_rate": 2.9467422316247112e-05, "loss": 0.0, "step": 63496 }, { "epoch": 1.23, "learning_rate": 2.9466775355989884e-05, "loss": 0.0003, "step": 63498 }, { "epoch": 1.23, "learning_rate": 2.946612839573265e-05, "loss": 0.0, "step": 63500 }, { "epoch": 1.23, "learning_rate": 2.9465481435475423e-05, "loss": 0.0, "step": 63502 }, { "epoch": 1.23, "learning_rate": 2.9464834475218188e-05, "loss": 0.0, "step": 63504 }, { "epoch": 1.23, "learning_rate": 2.9464187514960957e-05, "loss": 0.0, "step": 63506 }, { "epoch": 1.23, "learning_rate": 2.9463540554703726e-05, "loss": 0.0, "step": 63508 }, { "epoch": 1.23, "learning_rate": 2.9462893594446495e-05, "loss": 0.0041, "step": 63510 }, { "epoch": 1.23, "learning_rate": 2.946224663418926e-05, "loss": 0.011, "step": 63512 }, { "epoch": 1.23, "learning_rate": 2.9461599673932034e-05, "loss": 0.0009, "step": 63514 }, { "epoch": 1.23, "learning_rate": 2.94609527136748e-05, "loss": 0.0, "step": 63516 }, { "epoch": 1.23, "learning_rate": 2.9460305753417565e-05, "loss": 0.0, "step": 63518 }, { "epoch": 1.23, "learning_rate": 2.9459658793160338e-05, "loss": 0.0, "step": 63520 }, { "epoch": 1.23, "learning_rate": 2.9459011832903107e-05, "loss": 0.0, "step": 63522 }, { "epoch": 1.23, "learning_rate": 2.9458364872645876e-05, "loss": 0.0247, "step": 63524 }, { "epoch": 1.23, "learning_rate": 2.9457717912388645e-05, "loss": 0.0, "step": 63526 }, { "epoch": 1.23, "learning_rate": 2.945707095213141e-05, "loss": 0.0, "step": 63528 }, { "epoch": 1.23, "learning_rate": 2.9456423991874183e-05, "loss": 0.0001, "step": 63530 }, { "epoch": 1.23, "learning_rate": 2.945577703161695e-05, "loss": 0.0002, "step": 63532 }, { "epoch": 1.23, "learning_rate": 2.9455130071359714e-05, "loss": 0.0, "step": 63534 }, { "epoch": 1.23, "learning_rate": 2.9454483111102487e-05, "loss": 0.0001, "step": 63536 }, { "epoch": 1.23, "learning_rate": 2.9453836150845256e-05, "loss": 0.0159, "step": 63538 }, { "epoch": 1.23, "learning_rate": 2.945318919058802e-05, "loss": 0.0, "step": 63540 }, { "epoch": 1.23, "learning_rate": 2.9452542230330794e-05, "loss": 0.0, "step": 63542 }, { "epoch": 1.23, "learning_rate": 2.945189527007356e-05, "loss": 0.0, "step": 63544 }, { "epoch": 1.23, "learning_rate": 2.9451248309816332e-05, "loss": 0.0, "step": 63546 }, { "epoch": 1.23, "learning_rate": 2.9450601349559098e-05, "loss": 0.0, "step": 63548 }, { "epoch": 1.23, "learning_rate": 2.9449954389301864e-05, "loss": 0.0, "step": 63550 }, { "epoch": 1.23, "learning_rate": 2.9449307429044636e-05, "loss": 0.0, "step": 63552 }, { "epoch": 1.23, "learning_rate": 2.9448660468787402e-05, "loss": 0.0, "step": 63554 }, { "epoch": 1.23, "learning_rate": 2.944801350853017e-05, "loss": 0.0004, "step": 63556 }, { "epoch": 1.23, "learning_rate": 2.9447366548272943e-05, "loss": 0.0, "step": 63558 }, { "epoch": 1.23, "learning_rate": 2.944671958801571e-05, "loss": 0.0, "step": 63560 }, { "epoch": 1.23, "learning_rate": 2.944607262775848e-05, "loss": 0.0, "step": 63562 }, { "epoch": 1.23, "learning_rate": 2.9445425667501247e-05, "loss": 0.0001, "step": 63564 }, { "epoch": 1.23, "learning_rate": 2.9444778707244013e-05, "loss": 0.0013, "step": 63566 }, { "epoch": 1.23, "learning_rate": 2.9444131746986785e-05, "loss": 0.0, "step": 63568 }, { "epoch": 1.23, "learning_rate": 2.944348478672955e-05, "loss": 0.0012, "step": 63570 }, { "epoch": 1.23, "learning_rate": 2.944283782647232e-05, "loss": 0.0, "step": 63572 }, { "epoch": 1.23, "learning_rate": 2.9442190866215093e-05, "loss": 0.0, "step": 63574 }, { "epoch": 1.23, "learning_rate": 2.944154390595786e-05, "loss": 0.0039, "step": 63576 }, { "epoch": 1.23, "learning_rate": 2.9440896945700624e-05, "loss": 0.0072, "step": 63578 }, { "epoch": 1.23, "learning_rate": 2.9440249985443396e-05, "loss": 0.0, "step": 63580 }, { "epoch": 1.23, "learning_rate": 2.9439603025186162e-05, "loss": 0.0, "step": 63582 }, { "epoch": 1.23, "learning_rate": 2.9438956064928935e-05, "loss": 0.0, "step": 63584 }, { "epoch": 1.23, "learning_rate": 2.94383091046717e-05, "loss": 0.0001, "step": 63586 }, { "epoch": 1.23, "learning_rate": 2.943766214441447e-05, "loss": 0.0, "step": 63588 }, { "epoch": 1.23, "learning_rate": 2.9437015184157242e-05, "loss": 0.0001, "step": 63590 }, { "epoch": 1.23, "learning_rate": 2.9436368223900008e-05, "loss": 0.0, "step": 63592 }, { "epoch": 1.23, "learning_rate": 2.9435721263642773e-05, "loss": 0.0, "step": 63594 }, { "epoch": 1.23, "learning_rate": 2.9435074303385546e-05, "loss": 0.0131, "step": 63596 }, { "epoch": 1.23, "learning_rate": 2.943442734312831e-05, "loss": 0.0, "step": 63598 }, { "epoch": 1.23, "learning_rate": 2.943378038287108e-05, "loss": 0.0033, "step": 63600 }, { "epoch": 1.23, "learning_rate": 2.943313342261385e-05, "loss": 0.0, "step": 63602 }, { "epoch": 1.23, "learning_rate": 2.943248646235662e-05, "loss": 0.0, "step": 63604 }, { "epoch": 1.23, "learning_rate": 2.943183950209939e-05, "loss": 0.0, "step": 63606 }, { "epoch": 1.23, "learning_rate": 2.9431192541842157e-05, "loss": 0.0, "step": 63608 }, { "epoch": 1.23, "learning_rate": 2.9430545581584923e-05, "loss": 0.0006, "step": 63610 }, { "epoch": 1.23, "learning_rate": 2.9429898621327695e-05, "loss": 0.0024, "step": 63612 }, { "epoch": 1.23, "learning_rate": 2.942925166107046e-05, "loss": 0.0, "step": 63614 }, { "epoch": 1.23, "learning_rate": 2.942860470081323e-05, "loss": 0.0, "step": 63616 }, { "epoch": 1.23, "learning_rate": 2.9427957740556e-05, "loss": 0.0, "step": 63618 }, { "epoch": 1.23, "learning_rate": 2.9427310780298768e-05, "loss": 0.0, "step": 63620 }, { "epoch": 1.23, "learning_rate": 2.942666382004154e-05, "loss": 0.0, "step": 63622 }, { "epoch": 1.23, "learning_rate": 2.9426016859784306e-05, "loss": 0.0, "step": 63624 }, { "epoch": 1.23, "learning_rate": 2.9425369899527072e-05, "loss": 0.0002, "step": 63626 }, { "epoch": 1.23, "learning_rate": 2.9424722939269844e-05, "loss": 0.0, "step": 63628 }, { "epoch": 1.23, "learning_rate": 2.942407597901261e-05, "loss": 0.0, "step": 63630 }, { "epoch": 1.24, "learning_rate": 2.9423429018755376e-05, "loss": 0.0, "step": 63632 }, { "epoch": 1.24, "learning_rate": 2.9422782058498148e-05, "loss": 0.0001, "step": 63634 }, { "epoch": 1.24, "learning_rate": 2.9422135098240917e-05, "loss": 0.0, "step": 63636 }, { "epoch": 1.24, "learning_rate": 2.9421488137983683e-05, "loss": 0.0, "step": 63638 }, { "epoch": 1.24, "learning_rate": 2.9420841177726455e-05, "loss": 0.0015, "step": 63640 }, { "epoch": 1.24, "learning_rate": 2.942019421746922e-05, "loss": 0.0, "step": 63642 }, { "epoch": 1.24, "learning_rate": 2.9419547257211994e-05, "loss": 0.0, "step": 63644 }, { "epoch": 1.24, "learning_rate": 2.941890029695476e-05, "loss": 0.0, "step": 63646 }, { "epoch": 1.24, "learning_rate": 2.9418253336697525e-05, "loss": 0.0, "step": 63648 }, { "epoch": 1.24, "learning_rate": 2.9417606376440298e-05, "loss": 0.0, "step": 63650 }, { "epoch": 1.24, "learning_rate": 2.9416959416183067e-05, "loss": 0.0001, "step": 63652 }, { "epoch": 1.24, "learning_rate": 2.9416312455925832e-05, "loss": 0.0, "step": 63654 }, { "epoch": 1.24, "learning_rate": 2.9415665495668605e-05, "loss": 0.0, "step": 63656 }, { "epoch": 1.24, "learning_rate": 2.941501853541137e-05, "loss": 0.0001, "step": 63658 }, { "epoch": 1.24, "learning_rate": 2.9414371575154136e-05, "loss": 0.0, "step": 63660 }, { "epoch": 1.24, "learning_rate": 2.941372461489691e-05, "loss": 0.0, "step": 63662 }, { "epoch": 1.24, "learning_rate": 2.9413077654639674e-05, "loss": 0.0, "step": 63664 }, { "epoch": 1.24, "learning_rate": 2.9412430694382447e-05, "loss": 0.0019, "step": 63666 }, { "epoch": 1.24, "learning_rate": 2.9411783734125216e-05, "loss": 0.0, "step": 63668 }, { "epoch": 1.24, "learning_rate": 2.941113677386798e-05, "loss": 0.0, "step": 63670 }, { "epoch": 1.24, "learning_rate": 2.9410489813610754e-05, "loss": 0.0072, "step": 63672 }, { "epoch": 1.24, "learning_rate": 2.940984285335352e-05, "loss": 0.0, "step": 63674 }, { "epoch": 1.24, "learning_rate": 2.9409195893096285e-05, "loss": 0.0, "step": 63676 }, { "epoch": 1.24, "learning_rate": 2.9408548932839058e-05, "loss": 0.0, "step": 63678 }, { "epoch": 1.24, "learning_rate": 2.9407901972581824e-05, "loss": 0.0, "step": 63680 }, { "epoch": 1.24, "learning_rate": 2.9407255012324593e-05, "loss": 0.0, "step": 63682 }, { "epoch": 1.24, "learning_rate": 2.9406608052067365e-05, "loss": 0.0, "step": 63684 }, { "epoch": 1.24, "learning_rate": 2.940596109181013e-05, "loss": 0.0087, "step": 63686 }, { "epoch": 1.24, "learning_rate": 2.9405314131552903e-05, "loss": 0.0, "step": 63688 }, { "epoch": 1.24, "learning_rate": 2.940466717129567e-05, "loss": 0.0, "step": 63690 }, { "epoch": 1.24, "learning_rate": 2.9404020211038435e-05, "loss": 0.0, "step": 63692 }, { "epoch": 1.24, "learning_rate": 2.9403373250781207e-05, "loss": 0.0, "step": 63694 }, { "epoch": 1.24, "learning_rate": 2.9402726290523973e-05, "loss": 0.0149, "step": 63696 }, { "epoch": 1.24, "learning_rate": 2.9402079330266742e-05, "loss": 0.0, "step": 63698 }, { "epoch": 1.24, "learning_rate": 2.9401432370009514e-05, "loss": 0.0, "step": 63700 }, { "epoch": 1.24, "learning_rate": 2.940078540975228e-05, "loss": 0.0099, "step": 63702 }, { "epoch": 1.24, "learning_rate": 2.9400138449495053e-05, "loss": 0.0002, "step": 63704 }, { "epoch": 1.24, "learning_rate": 2.939949148923782e-05, "loss": 0.0001, "step": 63706 }, { "epoch": 1.24, "learning_rate": 2.9398844528980584e-05, "loss": 0.0, "step": 63708 }, { "epoch": 1.24, "learning_rate": 2.9398197568723356e-05, "loss": 0.0147, "step": 63710 }, { "epoch": 1.24, "learning_rate": 2.9397550608466122e-05, "loss": 0.006, "step": 63712 }, { "epoch": 1.24, "learning_rate": 2.939690364820889e-05, "loss": 0.0, "step": 63714 }, { "epoch": 1.24, "learning_rate": 2.9396256687951664e-05, "loss": 0.0, "step": 63716 }, { "epoch": 1.24, "learning_rate": 2.939560972769443e-05, "loss": 0.0, "step": 63718 }, { "epoch": 1.24, "learning_rate": 2.9394962767437195e-05, "loss": 0.0, "step": 63720 }, { "epoch": 1.24, "learning_rate": 2.9394315807179968e-05, "loss": 0.0, "step": 63722 }, { "epoch": 1.24, "learning_rate": 2.9393668846922733e-05, "loss": 0.0, "step": 63724 }, { "epoch": 1.24, "learning_rate": 2.9393021886665506e-05, "loss": 0.0, "step": 63726 }, { "epoch": 1.24, "learning_rate": 2.939237492640827e-05, "loss": 0.0, "step": 63728 }, { "epoch": 1.24, "learning_rate": 2.939172796615104e-05, "loss": 0.0, "step": 63730 }, { "epoch": 1.24, "learning_rate": 2.939108100589381e-05, "loss": 0.0001, "step": 63732 }, { "epoch": 1.24, "learning_rate": 2.939043404563658e-05, "loss": 0.0, "step": 63734 }, { "epoch": 1.24, "learning_rate": 2.9389787085379344e-05, "loss": 0.0, "step": 63736 }, { "epoch": 1.24, "learning_rate": 2.9389140125122117e-05, "loss": 0.0173, "step": 63738 }, { "epoch": 1.24, "learning_rate": 2.9388493164864883e-05, "loss": 0.0, "step": 63740 }, { "epoch": 1.24, "learning_rate": 2.9387846204607648e-05, "loss": 0.0, "step": 63742 }, { "epoch": 1.24, "learning_rate": 2.938719924435042e-05, "loss": 0.0, "step": 63744 }, { "epoch": 1.24, "learning_rate": 2.938655228409319e-05, "loss": 0.0001, "step": 63746 }, { "epoch": 1.24, "learning_rate": 2.938590532383596e-05, "loss": 0.0, "step": 63748 }, { "epoch": 1.24, "learning_rate": 2.9385258363578728e-05, "loss": 0.0001, "step": 63750 }, { "epoch": 1.24, "learning_rate": 2.9384611403321494e-05, "loss": 0.0, "step": 63752 }, { "epoch": 1.24, "learning_rate": 2.9383964443064266e-05, "loss": 0.0043, "step": 63754 }, { "epoch": 1.24, "learning_rate": 2.9383317482807032e-05, "loss": 0.0, "step": 63756 }, { "epoch": 1.24, "learning_rate": 2.9382670522549798e-05, "loss": 0.0063, "step": 63758 }, { "epoch": 1.24, "learning_rate": 2.938202356229257e-05, "loss": 0.0, "step": 63760 }, { "epoch": 1.24, "learning_rate": 2.938137660203534e-05, "loss": 0.0, "step": 63762 }, { "epoch": 1.24, "learning_rate": 2.9380729641778108e-05, "loss": 0.0, "step": 63764 }, { "epoch": 1.24, "learning_rate": 2.9380082681520877e-05, "loss": 0.0, "step": 63766 }, { "epoch": 1.24, "learning_rate": 2.9379435721263643e-05, "loss": 0.0001, "step": 63768 }, { "epoch": 1.24, "learning_rate": 2.9378788761006415e-05, "loss": 0.0099, "step": 63770 }, { "epoch": 1.24, "learning_rate": 2.937814180074918e-05, "loss": 0.0, "step": 63772 }, { "epoch": 1.24, "learning_rate": 2.9377494840491947e-05, "loss": 0.0, "step": 63774 }, { "epoch": 1.24, "learning_rate": 2.937684788023472e-05, "loss": 0.0, "step": 63776 }, { "epoch": 1.24, "learning_rate": 2.937620091997749e-05, "loss": 0.0, "step": 63778 }, { "epoch": 1.24, "learning_rate": 2.9375553959720254e-05, "loss": 0.0, "step": 63780 }, { "epoch": 1.24, "learning_rate": 2.9374906999463027e-05, "loss": 0.0, "step": 63782 }, { "epoch": 1.24, "learning_rate": 2.9374260039205792e-05, "loss": 0.0, "step": 63784 }, { "epoch": 1.24, "learning_rate": 2.9373613078948565e-05, "loss": 0.0088, "step": 63786 }, { "epoch": 1.24, "learning_rate": 2.937296611869133e-05, "loss": 0.0, "step": 63788 }, { "epoch": 1.24, "learning_rate": 2.9372319158434096e-05, "loss": 0.0, "step": 63790 }, { "epoch": 1.24, "learning_rate": 2.937167219817687e-05, "loss": 0.0, "step": 63792 }, { "epoch": 1.24, "learning_rate": 2.9371025237919638e-05, "loss": 0.0124, "step": 63794 }, { "epoch": 1.24, "learning_rate": 2.9370378277662403e-05, "loss": 0.0, "step": 63796 }, { "epoch": 1.24, "learning_rate": 2.9369731317405176e-05, "loss": 0.0, "step": 63798 }, { "epoch": 1.24, "learning_rate": 2.936908435714794e-05, "loss": 0.0013, "step": 63800 }, { "epoch": 1.24, "learning_rate": 2.9368437396890707e-05, "loss": 0.001, "step": 63802 }, { "epoch": 1.24, "learning_rate": 2.936779043663348e-05, "loss": 0.0, "step": 63804 }, { "epoch": 1.24, "learning_rate": 2.9367143476376245e-05, "loss": 0.0001, "step": 63806 }, { "epoch": 1.24, "learning_rate": 2.9366496516119018e-05, "loss": 0.0, "step": 63808 }, { "epoch": 1.24, "learning_rate": 2.9365849555861784e-05, "loss": 0.0, "step": 63810 }, { "epoch": 1.24, "learning_rate": 2.9365202595604553e-05, "loss": 0.0, "step": 63812 }, { "epoch": 1.24, "learning_rate": 2.9364555635347325e-05, "loss": 0.0, "step": 63814 }, { "epoch": 1.24, "learning_rate": 2.936390867509009e-05, "loss": 0.0003, "step": 63816 }, { "epoch": 1.24, "learning_rate": 2.9363261714832857e-05, "loss": 0.0, "step": 63818 }, { "epoch": 1.24, "learning_rate": 2.936261475457563e-05, "loss": 0.0, "step": 63820 }, { "epoch": 1.24, "learning_rate": 2.9361967794318395e-05, "loss": 0.0, "step": 63822 }, { "epoch": 1.24, "learning_rate": 2.9361320834061164e-05, "loss": 0.0001, "step": 63824 }, { "epoch": 1.24, "learning_rate": 2.9360673873803933e-05, "loss": 0.0, "step": 63826 }, { "epoch": 1.24, "learning_rate": 2.9360026913546702e-05, "loss": 0.0001, "step": 63828 }, { "epoch": 1.24, "learning_rate": 2.9359379953289474e-05, "loss": 0.0, "step": 63830 }, { "epoch": 1.24, "learning_rate": 2.935873299303224e-05, "loss": 0.0, "step": 63832 }, { "epoch": 1.24, "learning_rate": 2.9358086032775006e-05, "loss": 0.0, "step": 63834 }, { "epoch": 1.24, "learning_rate": 2.935743907251778e-05, "loss": 0.0, "step": 63836 }, { "epoch": 1.24, "learning_rate": 2.9356792112260544e-05, "loss": 0.0, "step": 63838 }, { "epoch": 1.24, "learning_rate": 2.9356145152003313e-05, "loss": 0.0001, "step": 63840 }, { "epoch": 1.24, "learning_rate": 2.9355498191746082e-05, "loss": 0.0003, "step": 63842 }, { "epoch": 1.24, "learning_rate": 2.935485123148885e-05, "loss": 0.0, "step": 63844 }, { "epoch": 1.24, "learning_rate": 2.9354204271231624e-05, "loss": 0.0018, "step": 63846 }, { "epoch": 1.24, "learning_rate": 2.935355731097439e-05, "loss": 0.0001, "step": 63848 }, { "epoch": 1.24, "learning_rate": 2.9352910350717155e-05, "loss": 0.0, "step": 63850 }, { "epoch": 1.24, "learning_rate": 2.9352263390459928e-05, "loss": 0.0, "step": 63852 }, { "epoch": 1.24, "learning_rate": 2.9351616430202693e-05, "loss": 0.0, "step": 63854 }, { "epoch": 1.24, "learning_rate": 2.9350969469945462e-05, "loss": 0.0, "step": 63856 }, { "epoch": 1.24, "learning_rate": 2.935032250968823e-05, "loss": 0.0, "step": 63858 }, { "epoch": 1.24, "learning_rate": 2.9349675549431e-05, "loss": 0.0, "step": 63860 }, { "epoch": 1.24, "learning_rate": 2.9349028589173766e-05, "loss": 0.0, "step": 63862 }, { "epoch": 1.24, "learning_rate": 2.934838162891654e-05, "loss": 0.0, "step": 63864 }, { "epoch": 1.24, "learning_rate": 2.9347734668659304e-05, "loss": 0.0, "step": 63866 }, { "epoch": 1.24, "learning_rate": 2.9347087708402077e-05, "loss": 0.0, "step": 63868 }, { "epoch": 1.24, "learning_rate": 2.9346440748144843e-05, "loss": 0.0, "step": 63870 }, { "epoch": 1.24, "learning_rate": 2.9345793787887608e-05, "loss": 0.0, "step": 63872 }, { "epoch": 1.24, "learning_rate": 2.934514682763038e-05, "loss": 0.0054, "step": 63874 }, { "epoch": 1.24, "learning_rate": 2.934449986737315e-05, "loss": 0.0, "step": 63876 }, { "epoch": 1.24, "learning_rate": 2.9343852907115916e-05, "loss": 0.0, "step": 63878 }, { "epoch": 1.24, "learning_rate": 2.9343205946858688e-05, "loss": 0.0, "step": 63880 }, { "epoch": 1.24, "learning_rate": 2.9342558986601454e-05, "loss": 0.0001, "step": 63882 }, { "epoch": 1.24, "learning_rate": 2.934191202634422e-05, "loss": 0.0, "step": 63884 }, { "epoch": 1.24, "learning_rate": 2.9341265066086992e-05, "loss": 0.0, "step": 63886 }, { "epoch": 1.24, "learning_rate": 2.9340618105829758e-05, "loss": 0.0, "step": 63888 }, { "epoch": 1.24, "learning_rate": 2.933997114557253e-05, "loss": 0.0, "step": 63890 }, { "epoch": 1.24, "learning_rate": 2.93393241853153e-05, "loss": 0.0, "step": 63892 }, { "epoch": 1.24, "learning_rate": 2.9338677225058065e-05, "loss": 0.0001, "step": 63894 }, { "epoch": 1.24, "learning_rate": 2.9338030264800837e-05, "loss": 0.0, "step": 63896 }, { "epoch": 1.24, "learning_rate": 2.9337383304543603e-05, "loss": 0.0, "step": 63898 }, { "epoch": 1.24, "learning_rate": 2.933673634428637e-05, "loss": 0.0, "step": 63900 }, { "epoch": 1.24, "learning_rate": 2.933608938402914e-05, "loss": 0.0001, "step": 63902 }, { "epoch": 1.24, "learning_rate": 2.9335442423771907e-05, "loss": 0.0, "step": 63904 }, { "epoch": 1.24, "learning_rate": 2.933479546351468e-05, "loss": 0.0002, "step": 63906 }, { "epoch": 1.24, "learning_rate": 2.933414850325745e-05, "loss": 0.0001, "step": 63908 }, { "epoch": 1.24, "learning_rate": 2.9333501543000214e-05, "loss": 0.0001, "step": 63910 }, { "epoch": 1.24, "learning_rate": 2.9332854582742987e-05, "loss": 0.0, "step": 63912 }, { "epoch": 1.24, "learning_rate": 2.9332207622485752e-05, "loss": 0.0001, "step": 63914 }, { "epoch": 1.24, "learning_rate": 2.9331560662228518e-05, "loss": 0.0, "step": 63916 }, { "epoch": 1.24, "learning_rate": 2.933091370197129e-05, "loss": 0.0, "step": 63918 }, { "epoch": 1.24, "learning_rate": 2.9330266741714056e-05, "loss": 0.0, "step": 63920 }, { "epoch": 1.24, "learning_rate": 2.9329619781456825e-05, "loss": 0.0014, "step": 63922 }, { "epoch": 1.24, "learning_rate": 2.9328972821199598e-05, "loss": 0.0014, "step": 63924 }, { "epoch": 1.24, "learning_rate": 2.9328325860942363e-05, "loss": 0.0, "step": 63926 }, { "epoch": 1.24, "learning_rate": 2.9327678900685136e-05, "loss": 0.0, "step": 63928 }, { "epoch": 1.24, "learning_rate": 2.93270319404279e-05, "loss": 0.0001, "step": 63930 }, { "epoch": 1.24, "learning_rate": 2.9326384980170667e-05, "loss": 0.0001, "step": 63932 }, { "epoch": 1.24, "learning_rate": 2.932573801991344e-05, "loss": 0.0, "step": 63934 }, { "epoch": 1.24, "learning_rate": 2.9325091059656205e-05, "loss": 0.0, "step": 63936 }, { "epoch": 1.24, "learning_rate": 2.9324444099398975e-05, "loss": 0.0, "step": 63938 }, { "epoch": 1.24, "learning_rate": 2.9323797139141747e-05, "loss": 0.0, "step": 63940 }, { "epoch": 1.24, "learning_rate": 2.9323150178884513e-05, "loss": 0.0, "step": 63942 }, { "epoch": 1.24, "learning_rate": 2.932250321862728e-05, "loss": 0.0001, "step": 63944 }, { "epoch": 1.24, "learning_rate": 2.932185625837005e-05, "loss": 0.0, "step": 63946 }, { "epoch": 1.24, "learning_rate": 2.9321209298112817e-05, "loss": 0.0, "step": 63948 }, { "epoch": 1.24, "learning_rate": 2.932056233785559e-05, "loss": 0.0, "step": 63950 }, { "epoch": 1.24, "learning_rate": 2.9319915377598355e-05, "loss": 0.0, "step": 63952 }, { "epoch": 1.24, "learning_rate": 2.9319268417341124e-05, "loss": 0.0, "step": 63954 }, { "epoch": 1.24, "learning_rate": 2.9318621457083896e-05, "loss": 0.0081, "step": 63956 }, { "epoch": 1.24, "learning_rate": 2.9317974496826662e-05, "loss": 0.0, "step": 63958 }, { "epoch": 1.24, "learning_rate": 2.9317327536569428e-05, "loss": 0.0, "step": 63960 }, { "epoch": 1.24, "learning_rate": 2.93166805763122e-05, "loss": 0.0, "step": 63962 }, { "epoch": 1.24, "learning_rate": 2.9316033616054966e-05, "loss": 0.0001, "step": 63964 }, { "epoch": 1.24, "learning_rate": 2.931538665579773e-05, "loss": 0.0, "step": 63966 }, { "epoch": 1.24, "learning_rate": 2.9314739695540504e-05, "loss": 0.0, "step": 63968 }, { "epoch": 1.24, "learning_rate": 2.9314092735283273e-05, "loss": 0.0, "step": 63970 }, { "epoch": 1.24, "learning_rate": 2.9313445775026042e-05, "loss": 0.0002, "step": 63972 }, { "epoch": 1.24, "learning_rate": 2.931279881476881e-05, "loss": 0.0, "step": 63974 }, { "epoch": 1.24, "learning_rate": 2.9312151854511577e-05, "loss": 0.0, "step": 63976 }, { "epoch": 1.24, "learning_rate": 2.931150489425435e-05, "loss": 0.0, "step": 63978 }, { "epoch": 1.24, "learning_rate": 2.9310857933997115e-05, "loss": 0.0, "step": 63980 }, { "epoch": 1.24, "learning_rate": 2.931021097373988e-05, "loss": 0.0001, "step": 63982 }, { "epoch": 1.24, "learning_rate": 2.9309564013482653e-05, "loss": 0.0002, "step": 63984 }, { "epoch": 1.24, "learning_rate": 2.9308917053225422e-05, "loss": 0.0, "step": 63986 }, { "epoch": 1.24, "learning_rate": 2.930827009296819e-05, "loss": 0.0, "step": 63988 }, { "epoch": 1.24, "learning_rate": 2.930762313271096e-05, "loss": 0.0, "step": 63990 }, { "epoch": 1.24, "learning_rate": 2.9306976172453726e-05, "loss": 0.0, "step": 63992 }, { "epoch": 1.24, "learning_rate": 2.93063292121965e-05, "loss": 0.0001, "step": 63994 }, { "epoch": 1.24, "learning_rate": 2.9305682251939264e-05, "loss": 0.0, "step": 63996 }, { "epoch": 1.24, "learning_rate": 2.930503529168203e-05, "loss": 0.0, "step": 63998 }, { "epoch": 1.24, "learning_rate": 2.9304388331424803e-05, "loss": 0.0001, "step": 64000 }, { "epoch": 1.24, "learning_rate": 2.930374137116757e-05, "loss": 0.0, "step": 64002 }, { "epoch": 1.24, "learning_rate": 2.9303094410910337e-05, "loss": 0.0, "step": 64004 }, { "epoch": 1.24, "learning_rate": 2.930244745065311e-05, "loss": 0.0004, "step": 64006 }, { "epoch": 1.24, "learning_rate": 2.9301800490395876e-05, "loss": 0.006, "step": 64008 }, { "epoch": 1.24, "learning_rate": 2.9301153530138648e-05, "loss": 0.0, "step": 64010 }, { "epoch": 1.24, "learning_rate": 2.9300506569881414e-05, "loss": 0.0, "step": 64012 }, { "epoch": 1.24, "learning_rate": 2.929985960962418e-05, "loss": 0.0, "step": 64014 }, { "epoch": 1.24, "learning_rate": 2.9299212649366952e-05, "loss": 0.0, "step": 64016 }, { "epoch": 1.24, "learning_rate": 2.929856568910972e-05, "loss": 0.0, "step": 64018 }, { "epoch": 1.24, "learning_rate": 2.9297918728852487e-05, "loss": 0.0, "step": 64020 }, { "epoch": 1.24, "learning_rate": 2.929727176859526e-05, "loss": 0.0001, "step": 64022 }, { "epoch": 1.24, "learning_rate": 2.9296624808338025e-05, "loss": 0.0, "step": 64024 }, { "epoch": 1.24, "learning_rate": 2.929597784808079e-05, "loss": 0.0027, "step": 64026 }, { "epoch": 1.24, "learning_rate": 2.9295330887823563e-05, "loss": 0.0029, "step": 64028 }, { "epoch": 1.24, "learning_rate": 2.929468392756633e-05, "loss": 0.0, "step": 64030 }, { "epoch": 1.24, "learning_rate": 2.92940369673091e-05, "loss": 0.0, "step": 64032 }, { "epoch": 1.24, "learning_rate": 2.929339000705187e-05, "loss": 0.0, "step": 64034 }, { "epoch": 1.24, "learning_rate": 2.9292743046794636e-05, "loss": 0.0, "step": 64036 }, { "epoch": 1.24, "learning_rate": 2.929209608653741e-05, "loss": 0.0, "step": 64038 }, { "epoch": 1.24, "learning_rate": 2.9291449126280174e-05, "loss": 0.0056, "step": 64040 }, { "epoch": 1.24, "learning_rate": 2.929080216602294e-05, "loss": 0.0, "step": 64042 }, { "epoch": 1.24, "learning_rate": 2.9290155205765712e-05, "loss": 0.0, "step": 64044 }, { "epoch": 1.24, "learning_rate": 2.9289508245508478e-05, "loss": 0.0, "step": 64046 }, { "epoch": 1.24, "learning_rate": 2.928886128525125e-05, "loss": 0.0, "step": 64048 }, { "epoch": 1.24, "learning_rate": 2.9288214324994016e-05, "loss": 0.0032, "step": 64050 }, { "epoch": 1.24, "learning_rate": 2.9287567364736785e-05, "loss": 0.0, "step": 64052 }, { "epoch": 1.24, "learning_rate": 2.9286920404479558e-05, "loss": 0.0083, "step": 64054 }, { "epoch": 1.24, "learning_rate": 2.9286273444222323e-05, "loss": 0.0, "step": 64056 }, { "epoch": 1.24, "learning_rate": 2.928562648396509e-05, "loss": 0.0, "step": 64058 }, { "epoch": 1.24, "learning_rate": 2.928497952370786e-05, "loss": 0.0, "step": 64060 }, { "epoch": 1.24, "learning_rate": 2.9284332563450627e-05, "loss": 0.0, "step": 64062 }, { "epoch": 1.24, "learning_rate": 2.9283685603193396e-05, "loss": 0.0001, "step": 64064 }, { "epoch": 1.24, "learning_rate": 2.9283038642936165e-05, "loss": 0.0, "step": 64066 }, { "epoch": 1.24, "learning_rate": 2.9282391682678934e-05, "loss": 0.0149, "step": 64068 }, { "epoch": 1.24, "learning_rate": 2.9281744722421707e-05, "loss": 0.0003, "step": 64070 }, { "epoch": 1.24, "learning_rate": 2.9281097762164473e-05, "loss": 0.0, "step": 64072 }, { "epoch": 1.24, "learning_rate": 2.928045080190724e-05, "loss": 0.0, "step": 64074 }, { "epoch": 1.24, "learning_rate": 2.927980384165001e-05, "loss": 0.0, "step": 64076 }, { "epoch": 1.24, "learning_rate": 2.9279156881392777e-05, "loss": 0.0039, "step": 64078 }, { "epoch": 1.24, "learning_rate": 2.9278509921135546e-05, "loss": 0.0001, "step": 64080 }, { "epoch": 1.24, "learning_rate": 2.9277862960878315e-05, "loss": 0.0, "step": 64082 }, { "epoch": 1.24, "learning_rate": 2.9277216000621084e-05, "loss": 0.0, "step": 64084 }, { "epoch": 1.24, "learning_rate": 2.927656904036385e-05, "loss": 0.0, "step": 64086 }, { "epoch": 1.24, "learning_rate": 2.9275922080106622e-05, "loss": 0.0, "step": 64088 }, { "epoch": 1.24, "learning_rate": 2.9275275119849388e-05, "loss": 0.0, "step": 64090 }, { "epoch": 1.24, "learning_rate": 2.927462815959216e-05, "loss": 0.0011, "step": 64092 }, { "epoch": 1.24, "learning_rate": 2.9273981199334926e-05, "loss": 0.0, "step": 64094 }, { "epoch": 1.24, "learning_rate": 2.9273334239077695e-05, "loss": 0.0062, "step": 64096 }, { "epoch": 1.24, "learning_rate": 2.9272687278820464e-05, "loss": 0.0, "step": 64098 }, { "epoch": 1.24, "learning_rate": 2.9272040318563233e-05, "loss": 0.0, "step": 64100 }, { "epoch": 1.24, "learning_rate": 2.9271393358306e-05, "loss": 0.0, "step": 64102 }, { "epoch": 1.24, "learning_rate": 2.927074639804877e-05, "loss": 0.0, "step": 64104 }, { "epoch": 1.24, "learning_rate": 2.9270099437791537e-05, "loss": 0.0, "step": 64106 }, { "epoch": 1.24, "learning_rate": 2.9269452477534303e-05, "loss": 0.0, "step": 64108 }, { "epoch": 1.24, "learning_rate": 2.9268805517277075e-05, "loss": 0.0055, "step": 64110 }, { "epoch": 1.24, "learning_rate": 2.926815855701984e-05, "loss": 0.0, "step": 64112 }, { "epoch": 1.24, "learning_rate": 2.9267511596762613e-05, "loss": 0.0, "step": 64114 }, { "epoch": 1.24, "learning_rate": 2.9266864636505382e-05, "loss": 0.0, "step": 64116 }, { "epoch": 1.24, "learning_rate": 2.9266217676248148e-05, "loss": 0.0001, "step": 64118 }, { "epoch": 1.24, "learning_rate": 2.926557071599092e-05, "loss": 0.0, "step": 64120 }, { "epoch": 1.24, "learning_rate": 2.9264923755733686e-05, "loss": 0.0001, "step": 64122 }, { "epoch": 1.24, "learning_rate": 2.9264276795476452e-05, "loss": 0.0, "step": 64124 }, { "epoch": 1.24, "learning_rate": 2.9263629835219224e-05, "loss": 0.0001, "step": 64126 }, { "epoch": 1.24, "learning_rate": 2.926298287496199e-05, "loss": 0.0, "step": 64128 }, { "epoch": 1.24, "learning_rate": 2.9262335914704763e-05, "loss": 0.0, "step": 64130 }, { "epoch": 1.24, "learning_rate": 2.926168895444753e-05, "loss": 0.0, "step": 64132 }, { "epoch": 1.24, "learning_rate": 2.9261041994190297e-05, "loss": 0.0095, "step": 64134 }, { "epoch": 1.24, "learning_rate": 2.926039503393307e-05, "loss": 0.0002, "step": 64136 }, { "epoch": 1.24, "learning_rate": 2.9259748073675836e-05, "loss": 0.0, "step": 64138 }, { "epoch": 1.24, "learning_rate": 2.92591011134186e-05, "loss": 0.0, "step": 64140 }, { "epoch": 1.24, "learning_rate": 2.9258454153161374e-05, "loss": 0.0, "step": 64142 }, { "epoch": 1.24, "learning_rate": 2.925780719290414e-05, "loss": 0.0001, "step": 64144 }, { "epoch": 1.24, "learning_rate": 2.925716023264691e-05, "loss": 0.0, "step": 64146 }, { "epoch": 1.25, "learning_rate": 2.925651327238968e-05, "loss": 0.0002, "step": 64148 }, { "epoch": 1.25, "learning_rate": 2.9255866312132447e-05, "loss": 0.0, "step": 64150 }, { "epoch": 1.25, "learning_rate": 2.925521935187522e-05, "loss": 0.0009, "step": 64152 }, { "epoch": 1.25, "learning_rate": 2.9254572391617985e-05, "loss": 0.0006, "step": 64154 }, { "epoch": 1.25, "learning_rate": 2.925392543136075e-05, "loss": 0.0, "step": 64156 }, { "epoch": 1.25, "learning_rate": 2.9253278471103523e-05, "loss": 0.0, "step": 64158 }, { "epoch": 1.25, "learning_rate": 2.925263151084629e-05, "loss": 0.0, "step": 64160 }, { "epoch": 1.25, "learning_rate": 2.9251984550589058e-05, "loss": 0.0, "step": 64162 }, { "epoch": 1.25, "learning_rate": 2.925133759033183e-05, "loss": 0.0, "step": 64164 }, { "epoch": 1.25, "learning_rate": 2.9250690630074596e-05, "loss": 0.0, "step": 64166 }, { "epoch": 1.25, "learning_rate": 2.925004366981736e-05, "loss": 0.0, "step": 64168 }, { "epoch": 1.25, "learning_rate": 2.9249396709560134e-05, "loss": 0.0, "step": 64170 }, { "epoch": 1.25, "learning_rate": 2.92487497493029e-05, "loss": 0.0, "step": 64172 }, { "epoch": 1.25, "learning_rate": 2.9248102789045672e-05, "loss": 0.0006, "step": 64174 }, { "epoch": 1.25, "learning_rate": 2.9247455828788438e-05, "loss": 0.0, "step": 64176 }, { "epoch": 1.25, "learning_rate": 2.9246808868531207e-05, "loss": 0.0, "step": 64178 }, { "epoch": 1.25, "learning_rate": 2.924616190827398e-05, "loss": 0.0, "step": 64180 }, { "epoch": 1.25, "learning_rate": 2.9245514948016745e-05, "loss": 0.0001, "step": 64182 }, { "epoch": 1.25, "learning_rate": 2.924486798775951e-05, "loss": 0.0, "step": 64184 }, { "epoch": 1.25, "learning_rate": 2.9244221027502283e-05, "loss": 0.0, "step": 64186 }, { "epoch": 1.25, "learning_rate": 2.924357406724505e-05, "loss": 0.0, "step": 64188 }, { "epoch": 1.25, "learning_rate": 2.924292710698782e-05, "loss": 0.0, "step": 64190 }, { "epoch": 1.25, "learning_rate": 2.9242280146730587e-05, "loss": 0.0, "step": 64192 }, { "epoch": 1.25, "learning_rate": 2.9241633186473356e-05, "loss": 0.0002, "step": 64194 }, { "epoch": 1.25, "learning_rate": 2.924098622621613e-05, "loss": 0.0, "step": 64196 }, { "epoch": 1.25, "learning_rate": 2.9240339265958894e-05, "loss": 0.0, "step": 64198 }, { "epoch": 1.25, "learning_rate": 2.923969230570166e-05, "loss": 0.0, "step": 64200 }, { "epoch": 1.25, "learning_rate": 2.9239045345444433e-05, "loss": 0.0006, "step": 64202 }, { "epoch": 1.25, "learning_rate": 2.92383983851872e-05, "loss": 0.0001, "step": 64204 }, { "epoch": 1.25, "learning_rate": 2.9237751424929964e-05, "loss": 0.0001, "step": 64206 }, { "epoch": 1.25, "learning_rate": 2.9237104464672737e-05, "loss": 0.0, "step": 64208 }, { "epoch": 1.25, "learning_rate": 2.9236457504415506e-05, "loss": 0.0001, "step": 64210 }, { "epoch": 1.25, "learning_rate": 2.9235810544158275e-05, "loss": 0.0, "step": 64212 }, { "epoch": 1.25, "learning_rate": 2.9235163583901044e-05, "loss": 0.0, "step": 64214 }, { "epoch": 1.25, "learning_rate": 2.923451662364381e-05, "loss": 0.0001, "step": 64216 }, { "epoch": 1.25, "learning_rate": 2.9233869663386582e-05, "loss": 0.005, "step": 64218 }, { "epoch": 1.25, "learning_rate": 2.9233222703129348e-05, "loss": 0.0, "step": 64220 }, { "epoch": 1.25, "learning_rate": 2.9232575742872113e-05, "loss": 0.0, "step": 64222 }, { "epoch": 1.25, "learning_rate": 2.9231928782614886e-05, "loss": 0.0033, "step": 64224 }, { "epoch": 1.25, "learning_rate": 2.9231281822357655e-05, "loss": 0.0, "step": 64226 }, { "epoch": 1.25, "learning_rate": 2.923063486210042e-05, "loss": 0.0115, "step": 64228 }, { "epoch": 1.25, "learning_rate": 2.9229987901843193e-05, "loss": 0.0002, "step": 64230 }, { "epoch": 1.25, "learning_rate": 2.922934094158596e-05, "loss": 0.0006, "step": 64232 }, { "epoch": 1.25, "learning_rate": 2.922869398132873e-05, "loss": 0.0, "step": 64234 }, { "epoch": 1.25, "learning_rate": 2.9228047021071497e-05, "loss": 0.0, "step": 64236 }, { "epoch": 1.25, "learning_rate": 2.9227400060814263e-05, "loss": 0.0001, "step": 64238 }, { "epoch": 1.25, "learning_rate": 2.9226753100557035e-05, "loss": 0.0, "step": 64240 }, { "epoch": 1.25, "learning_rate": 2.9226106140299804e-05, "loss": 0.0, "step": 64242 }, { "epoch": 1.25, "learning_rate": 2.922545918004257e-05, "loss": 0.0001, "step": 64244 }, { "epoch": 1.25, "learning_rate": 2.9224812219785342e-05, "loss": 0.0006, "step": 64246 }, { "epoch": 1.25, "learning_rate": 2.9224165259528108e-05, "loss": 0.004, "step": 64248 }, { "epoch": 1.25, "learning_rate": 2.9223518299270874e-05, "loss": 0.0002, "step": 64250 }, { "epoch": 1.25, "learning_rate": 2.9222871339013646e-05, "loss": 0.0, "step": 64252 }, { "epoch": 1.25, "learning_rate": 2.9222224378756412e-05, "loss": 0.0, "step": 64254 }, { "epoch": 1.25, "learning_rate": 2.9221577418499184e-05, "loss": 0.0, "step": 64256 }, { "epoch": 1.25, "learning_rate": 2.9220930458241953e-05, "loss": 0.0, "step": 64258 }, { "epoch": 1.25, "learning_rate": 2.922028349798472e-05, "loss": 0.0, "step": 64260 }, { "epoch": 1.25, "learning_rate": 2.921963653772749e-05, "loss": 0.0, "step": 64262 }, { "epoch": 1.25, "learning_rate": 2.9218989577470257e-05, "loss": 0.0, "step": 64264 }, { "epoch": 1.25, "learning_rate": 2.9218342617213023e-05, "loss": 0.0001, "step": 64266 }, { "epoch": 1.25, "learning_rate": 2.9217695656955795e-05, "loss": 0.0015, "step": 64268 }, { "epoch": 1.25, "learning_rate": 2.921704869669856e-05, "loss": 0.0, "step": 64270 }, { "epoch": 1.25, "learning_rate": 2.9216401736441334e-05, "loss": 0.0, "step": 64272 }, { "epoch": 1.25, "learning_rate": 2.9215754776184103e-05, "loss": 0.0001, "step": 64274 }, { "epoch": 1.25, "learning_rate": 2.921510781592687e-05, "loss": 0.0082, "step": 64276 }, { "epoch": 1.25, "learning_rate": 2.921446085566964e-05, "loss": 0.0, "step": 64278 }, { "epoch": 1.25, "learning_rate": 2.9213813895412407e-05, "loss": 0.0001, "step": 64280 }, { "epoch": 1.25, "learning_rate": 2.9213166935155172e-05, "loss": 0.0001, "step": 64282 }, { "epoch": 1.25, "learning_rate": 2.9212519974897945e-05, "loss": 0.0006, "step": 64284 }, { "epoch": 1.25, "learning_rate": 2.921187301464071e-05, "loss": 0.0, "step": 64286 }, { "epoch": 1.25, "learning_rate": 2.921122605438348e-05, "loss": 0.0005, "step": 64288 }, { "epoch": 1.25, "learning_rate": 2.921057909412625e-05, "loss": 0.0, "step": 64290 }, { "epoch": 1.25, "learning_rate": 2.9209932133869018e-05, "loss": 0.0, "step": 64292 }, { "epoch": 1.25, "learning_rate": 2.920928517361179e-05, "loss": 0.0, "step": 64294 }, { "epoch": 1.25, "learning_rate": 2.9208638213354556e-05, "loss": 0.0001, "step": 64296 }, { "epoch": 1.25, "learning_rate": 2.920799125309732e-05, "loss": 0.0002, "step": 64298 }, { "epoch": 1.25, "learning_rate": 2.9207344292840094e-05, "loss": 0.0, "step": 64300 }, { "epoch": 1.25, "learning_rate": 2.920669733258286e-05, "loss": 0.0, "step": 64302 }, { "epoch": 1.25, "learning_rate": 2.920605037232563e-05, "loss": 0.0, "step": 64304 }, { "epoch": 1.25, "learning_rate": 2.9205403412068398e-05, "loss": 0.0, "step": 64306 }, { "epoch": 1.25, "learning_rate": 2.9204756451811167e-05, "loss": 0.0, "step": 64308 }, { "epoch": 1.25, "learning_rate": 2.9204109491553933e-05, "loss": 0.0, "step": 64310 }, { "epoch": 1.25, "learning_rate": 2.9203462531296705e-05, "loss": 0.0, "step": 64312 }, { "epoch": 1.25, "learning_rate": 2.920281557103947e-05, "loss": 0.0015, "step": 64314 }, { "epoch": 1.25, "learning_rate": 2.9202168610782243e-05, "loss": 0.0, "step": 64316 }, { "epoch": 1.25, "learning_rate": 2.920152165052501e-05, "loss": 0.0, "step": 64318 }, { "epoch": 1.25, "learning_rate": 2.9200874690267778e-05, "loss": 0.0002, "step": 64320 }, { "epoch": 1.25, "learning_rate": 2.9200227730010547e-05, "loss": 0.0, "step": 64322 }, { "epoch": 1.25, "learning_rate": 2.9199580769753316e-05, "loss": 0.0, "step": 64324 }, { "epoch": 1.25, "learning_rate": 2.9198933809496082e-05, "loss": 0.0, "step": 64326 }, { "epoch": 1.25, "learning_rate": 2.9198286849238854e-05, "loss": 0.0, "step": 64328 }, { "epoch": 1.25, "learning_rate": 2.919763988898162e-05, "loss": 0.0002, "step": 64330 }, { "epoch": 1.25, "learning_rate": 2.9196992928724393e-05, "loss": 0.0001, "step": 64332 }, { "epoch": 1.25, "learning_rate": 2.919634596846716e-05, "loss": 0.0038, "step": 64334 }, { "epoch": 1.25, "learning_rate": 2.9195699008209927e-05, "loss": 0.0, "step": 64336 }, { "epoch": 1.25, "learning_rate": 2.9195052047952697e-05, "loss": 0.0, "step": 64338 }, { "epoch": 1.25, "learning_rate": 2.9194405087695466e-05, "loss": 0.0, "step": 64340 }, { "epoch": 1.25, "learning_rate": 2.919375812743823e-05, "loss": 0.0368, "step": 64342 }, { "epoch": 1.25, "learning_rate": 2.9193111167181004e-05, "loss": 0.0087, "step": 64344 }, { "epoch": 1.25, "learning_rate": 2.919246420692377e-05, "loss": 0.0, "step": 64346 }, { "epoch": 1.25, "learning_rate": 2.9191817246666535e-05, "loss": 0.0, "step": 64348 }, { "epoch": 1.25, "learning_rate": 2.9191170286409308e-05, "loss": 0.0, "step": 64350 }, { "epoch": 1.25, "learning_rate": 2.9190523326152073e-05, "loss": 0.0, "step": 64352 }, { "epoch": 1.25, "learning_rate": 2.9189876365894846e-05, "loss": 0.0, "step": 64354 }, { "epoch": 1.25, "learning_rate": 2.9189229405637615e-05, "loss": 0.0, "step": 64356 }, { "epoch": 1.25, "learning_rate": 2.918858244538038e-05, "loss": 0.0, "step": 64358 }, { "epoch": 1.25, "learning_rate": 2.9187935485123153e-05, "loss": 0.0, "step": 64360 }, { "epoch": 1.25, "learning_rate": 2.918728852486592e-05, "loss": 0.0, "step": 64362 }, { "epoch": 1.25, "learning_rate": 2.9186641564608684e-05, "loss": 0.0, "step": 64364 }, { "epoch": 1.25, "learning_rate": 2.9185994604351457e-05, "loss": 0.0003, "step": 64366 }, { "epoch": 1.25, "learning_rate": 2.9185347644094223e-05, "loss": 0.0002, "step": 64368 }, { "epoch": 1.25, "learning_rate": 2.9184700683836992e-05, "loss": 0.0001, "step": 64370 }, { "epoch": 1.25, "learning_rate": 2.9184053723579764e-05, "loss": 0.0001, "step": 64372 }, { "epoch": 1.25, "learning_rate": 2.918340676332253e-05, "loss": 0.0063, "step": 64374 }, { "epoch": 1.25, "learning_rate": 2.9182759803065302e-05, "loss": 0.0, "step": 64376 }, { "epoch": 1.25, "learning_rate": 2.9182112842808068e-05, "loss": 0.0, "step": 64378 }, { "epoch": 1.25, "learning_rate": 2.9181465882550834e-05, "loss": 0.0, "step": 64380 }, { "epoch": 1.25, "learning_rate": 2.9180818922293606e-05, "loss": 0.0001, "step": 64382 }, { "epoch": 1.25, "learning_rate": 2.9180171962036372e-05, "loss": 0.0, "step": 64384 }, { "epoch": 1.25, "learning_rate": 2.917952500177914e-05, "loss": 0.0134, "step": 64386 }, { "epoch": 1.25, "learning_rate": 2.9178878041521913e-05, "loss": 0.0, "step": 64388 }, { "epoch": 1.25, "learning_rate": 2.917823108126468e-05, "loss": 0.0, "step": 64390 }, { "epoch": 1.25, "learning_rate": 2.9177584121007445e-05, "loss": 0.0, "step": 64392 }, { "epoch": 1.25, "learning_rate": 2.9176937160750217e-05, "loss": 0.0, "step": 64394 }, { "epoch": 1.25, "learning_rate": 2.9176290200492983e-05, "loss": 0.0, "step": 64396 }, { "epoch": 1.25, "learning_rate": 2.9175643240235755e-05, "loss": 0.0, "step": 64398 }, { "epoch": 1.25, "learning_rate": 2.917499627997852e-05, "loss": 0.0, "step": 64400 }, { "epoch": 1.25, "learning_rate": 2.917434931972129e-05, "loss": 0.0034, "step": 64402 }, { "epoch": 1.25, "learning_rate": 2.9173702359464063e-05, "loss": 0.0, "step": 64404 }, { "epoch": 1.25, "learning_rate": 2.917305539920683e-05, "loss": 0.0, "step": 64406 }, { "epoch": 1.25, "learning_rate": 2.9172408438949594e-05, "loss": 0.0, "step": 64408 }, { "epoch": 1.25, "learning_rate": 2.9171761478692367e-05, "loss": 0.0, "step": 64410 }, { "epoch": 1.25, "learning_rate": 2.9171114518435132e-05, "loss": 0.0, "step": 64412 }, { "epoch": 1.25, "learning_rate": 2.9170467558177905e-05, "loss": 0.001, "step": 64414 }, { "epoch": 1.25, "learning_rate": 2.916982059792067e-05, "loss": 0.0005, "step": 64416 }, { "epoch": 1.25, "learning_rate": 2.916917363766344e-05, "loss": 0.0, "step": 64418 }, { "epoch": 1.25, "learning_rate": 2.9168526677406212e-05, "loss": 0.0, "step": 64420 }, { "epoch": 1.25, "learning_rate": 2.9167879717148978e-05, "loss": 0.0, "step": 64422 }, { "epoch": 1.25, "learning_rate": 2.9167232756891743e-05, "loss": 0.0, "step": 64424 }, { "epoch": 1.25, "learning_rate": 2.9166585796634516e-05, "loss": 0.0, "step": 64426 }, { "epoch": 1.25, "learning_rate": 2.916593883637728e-05, "loss": 0.0003, "step": 64428 }, { "epoch": 1.25, "learning_rate": 2.9165291876120047e-05, "loss": 0.0, "step": 64430 }, { "epoch": 1.25, "learning_rate": 2.916464491586282e-05, "loss": 0.0, "step": 64432 }, { "epoch": 1.25, "learning_rate": 2.916399795560559e-05, "loss": 0.0, "step": 64434 }, { "epoch": 1.25, "learning_rate": 2.916335099534836e-05, "loss": 0.0, "step": 64436 }, { "epoch": 1.25, "learning_rate": 2.9162704035091127e-05, "loss": 0.0, "step": 64438 }, { "epoch": 1.25, "learning_rate": 2.9162057074833893e-05, "loss": 0.0007, "step": 64440 }, { "epoch": 1.25, "learning_rate": 2.9161410114576665e-05, "loss": 0.0, "step": 64442 }, { "epoch": 1.25, "learning_rate": 2.916076315431943e-05, "loss": 0.0001, "step": 64444 }, { "epoch": 1.25, "learning_rate": 2.9160116194062197e-05, "loss": 0.0001, "step": 64446 }, { "epoch": 1.25, "learning_rate": 2.915946923380497e-05, "loss": 0.0, "step": 64448 }, { "epoch": 1.25, "learning_rate": 2.9158822273547738e-05, "loss": 0.0, "step": 64450 }, { "epoch": 1.25, "learning_rate": 2.9158175313290504e-05, "loss": 0.0, "step": 64452 }, { "epoch": 1.25, "learning_rate": 2.9157528353033276e-05, "loss": 0.0006, "step": 64454 }, { "epoch": 1.25, "learning_rate": 2.9156881392776042e-05, "loss": 0.0, "step": 64456 }, { "epoch": 1.25, "learning_rate": 2.9156234432518814e-05, "loss": 0.0001, "step": 64458 }, { "epoch": 1.25, "learning_rate": 2.915558747226158e-05, "loss": 0.0001, "step": 64460 }, { "epoch": 1.25, "learning_rate": 2.9154940512004346e-05, "loss": 0.0, "step": 64462 }, { "epoch": 1.25, "learning_rate": 2.915429355174712e-05, "loss": 0.0, "step": 64464 }, { "epoch": 1.25, "learning_rate": 2.9153646591489887e-05, "loss": 0.0052, "step": 64466 }, { "epoch": 1.25, "learning_rate": 2.9152999631232653e-05, "loss": 0.0, "step": 64468 }, { "epoch": 1.25, "learning_rate": 2.9152352670975426e-05, "loss": 0.0, "step": 64470 }, { "epoch": 1.25, "learning_rate": 2.915170571071819e-05, "loss": 0.0, "step": 64472 }, { "epoch": 1.25, "learning_rate": 2.9151058750460964e-05, "loss": 0.0078, "step": 64474 }, { "epoch": 1.25, "learning_rate": 2.915041179020373e-05, "loss": 0.0, "step": 64476 }, { "epoch": 1.25, "learning_rate": 2.9149764829946495e-05, "loss": 0.0144, "step": 64478 }, { "epoch": 1.25, "learning_rate": 2.9149117869689268e-05, "loss": 0.0, "step": 64480 }, { "epoch": 1.25, "learning_rate": 2.9148470909432037e-05, "loss": 0.0, "step": 64482 }, { "epoch": 1.25, "learning_rate": 2.9147823949174802e-05, "loss": 0.0, "step": 64484 }, { "epoch": 1.25, "learning_rate": 2.9147176988917575e-05, "loss": 0.0085, "step": 64486 }, { "epoch": 1.25, "learning_rate": 2.914653002866034e-05, "loss": 0.0, "step": 64488 }, { "epoch": 1.25, "learning_rate": 2.9145883068403106e-05, "loss": 0.0, "step": 64490 }, { "epoch": 1.25, "learning_rate": 2.914523610814588e-05, "loss": 0.0, "step": 64492 }, { "epoch": 1.25, "learning_rate": 2.9144589147888644e-05, "loss": 0.0, "step": 64494 }, { "epoch": 1.25, "learning_rate": 2.9143942187631417e-05, "loss": 0.0, "step": 64496 }, { "epoch": 1.25, "learning_rate": 2.9143295227374186e-05, "loss": 0.0036, "step": 64498 }, { "epoch": 1.25, "learning_rate": 2.914264826711695e-05, "loss": 0.0, "step": 64500 }, { "epoch": 1.25, "learning_rate": 2.9142001306859724e-05, "loss": 0.0, "step": 64502 }, { "epoch": 1.25, "learning_rate": 2.914135434660249e-05, "loss": 0.0105, "step": 64504 }, { "epoch": 1.25, "learning_rate": 2.9140707386345256e-05, "loss": 0.0, "step": 64506 }, { "epoch": 1.25, "learning_rate": 2.9140060426088028e-05, "loss": 0.0, "step": 64508 }, { "epoch": 1.25, "learning_rate": 2.9139413465830794e-05, "loss": 0.0, "step": 64510 }, { "epoch": 1.25, "learning_rate": 2.9138766505573563e-05, "loss": 0.0, "step": 64512 }, { "epoch": 1.25, "learning_rate": 2.9138119545316335e-05, "loss": 0.0, "step": 64514 }, { "epoch": 1.25, "learning_rate": 2.91374725850591e-05, "loss": 0.004, "step": 64516 }, { "epoch": 1.25, "learning_rate": 2.9136825624801873e-05, "loss": 0.0, "step": 64518 }, { "epoch": 1.25, "learning_rate": 2.913617866454464e-05, "loss": 0.0015, "step": 64520 }, { "epoch": 1.25, "learning_rate": 2.9135531704287405e-05, "loss": 0.0029, "step": 64522 }, { "epoch": 1.25, "learning_rate": 2.9134884744030177e-05, "loss": 0.0, "step": 64524 }, { "epoch": 1.25, "learning_rate": 2.9134237783772943e-05, "loss": 0.0, "step": 64526 }, { "epoch": 1.25, "learning_rate": 2.9133590823515712e-05, "loss": 0.0, "step": 64528 }, { "epoch": 1.25, "learning_rate": 2.913294386325848e-05, "loss": 0.0, "step": 64530 }, { "epoch": 1.25, "learning_rate": 2.913229690300125e-05, "loss": 0.0, "step": 64532 }, { "epoch": 1.25, "learning_rate": 2.9131649942744016e-05, "loss": 0.0, "step": 64534 }, { "epoch": 1.25, "learning_rate": 2.913100298248679e-05, "loss": 0.0, "step": 64536 }, { "epoch": 1.25, "learning_rate": 2.9130356022229554e-05, "loss": 0.0, "step": 64538 }, { "epoch": 1.25, "learning_rate": 2.9129709061972327e-05, "loss": 0.0002, "step": 64540 }, { "epoch": 1.25, "learning_rate": 2.9129062101715092e-05, "loss": 0.0, "step": 64542 }, { "epoch": 1.25, "learning_rate": 2.912841514145786e-05, "loss": 0.0, "step": 64544 }, { "epoch": 1.25, "learning_rate": 2.912776818120063e-05, "loss": 0.0, "step": 64546 }, { "epoch": 1.25, "learning_rate": 2.91271212209434e-05, "loss": 0.0, "step": 64548 }, { "epoch": 1.25, "learning_rate": 2.9126474260686165e-05, "loss": 0.0, "step": 64550 }, { "epoch": 1.25, "learning_rate": 2.9125827300428938e-05, "loss": 0.0002, "step": 64552 }, { "epoch": 1.25, "learning_rate": 2.9125180340171703e-05, "loss": 0.0, "step": 64554 }, { "epoch": 1.25, "learning_rate": 2.9124533379914476e-05, "loss": 0.0, "step": 64556 }, { "epoch": 1.25, "learning_rate": 2.912388641965724e-05, "loss": 0.0, "step": 64558 }, { "epoch": 1.25, "learning_rate": 2.912323945940001e-05, "loss": 0.007, "step": 64560 }, { "epoch": 1.25, "learning_rate": 2.912259249914278e-05, "loss": 0.0006, "step": 64562 }, { "epoch": 1.25, "learning_rate": 2.912194553888555e-05, "loss": 0.0, "step": 64564 }, { "epoch": 1.25, "learning_rate": 2.9121298578628315e-05, "loss": 0.0, "step": 64566 }, { "epoch": 1.25, "learning_rate": 2.9120651618371087e-05, "loss": 0.0, "step": 64568 }, { "epoch": 1.25, "learning_rate": 2.9120004658113853e-05, "loss": 0.0, "step": 64570 }, { "epoch": 1.25, "learning_rate": 2.911935769785662e-05, "loss": 0.0, "step": 64572 }, { "epoch": 1.25, "learning_rate": 2.911871073759939e-05, "loss": 0.0038, "step": 64574 }, { "epoch": 1.25, "learning_rate": 2.911806377734216e-05, "loss": 0.0, "step": 64576 }, { "epoch": 1.25, "learning_rate": 2.911741681708493e-05, "loss": 0.0, "step": 64578 }, { "epoch": 1.25, "learning_rate": 2.9116769856827698e-05, "loss": 0.0, "step": 64580 }, { "epoch": 1.25, "learning_rate": 2.9116122896570464e-05, "loss": 0.0, "step": 64582 }, { "epoch": 1.25, "learning_rate": 2.9115475936313236e-05, "loss": 0.0, "step": 64584 }, { "epoch": 1.25, "learning_rate": 2.9114828976056002e-05, "loss": 0.0, "step": 64586 }, { "epoch": 1.25, "learning_rate": 2.9114182015798768e-05, "loss": 0.0, "step": 64588 }, { "epoch": 1.25, "learning_rate": 2.911353505554154e-05, "loss": 0.004, "step": 64590 }, { "epoch": 1.25, "learning_rate": 2.9112888095284306e-05, "loss": 0.0, "step": 64592 }, { "epoch": 1.25, "learning_rate": 2.9112241135027075e-05, "loss": 0.0, "step": 64594 }, { "epoch": 1.25, "learning_rate": 2.9111594174769847e-05, "loss": 0.0079, "step": 64596 }, { "epoch": 1.25, "learning_rate": 2.9110947214512613e-05, "loss": 0.0, "step": 64598 }, { "epoch": 1.25, "learning_rate": 2.9110300254255386e-05, "loss": 0.0, "step": 64600 }, { "epoch": 1.25, "learning_rate": 2.910965329399815e-05, "loss": 0.0, "step": 64602 }, { "epoch": 1.25, "learning_rate": 2.9109006333740917e-05, "loss": 0.0, "step": 64604 }, { "epoch": 1.25, "learning_rate": 2.910835937348369e-05, "loss": 0.0042, "step": 64606 }, { "epoch": 1.25, "learning_rate": 2.9107712413226455e-05, "loss": 0.0, "step": 64608 }, { "epoch": 1.25, "learning_rate": 2.9107065452969224e-05, "loss": 0.0, "step": 64610 }, { "epoch": 1.25, "learning_rate": 2.9106418492711997e-05, "loss": 0.0, "step": 64612 }, { "epoch": 1.25, "learning_rate": 2.9105771532454762e-05, "loss": 0.0, "step": 64614 }, { "epoch": 1.25, "learning_rate": 2.9105124572197528e-05, "loss": 0.0, "step": 64616 }, { "epoch": 1.25, "learning_rate": 2.91044776119403e-05, "loss": 0.0, "step": 64618 }, { "epoch": 1.25, "learning_rate": 2.9103830651683066e-05, "loss": 0.0, "step": 64620 }, { "epoch": 1.25, "learning_rate": 2.910318369142584e-05, "loss": 0.0005, "step": 64622 }, { "epoch": 1.25, "learning_rate": 2.9102536731168604e-05, "loss": 0.0, "step": 64624 }, { "epoch": 1.25, "learning_rate": 2.9101889770911374e-05, "loss": 0.0, "step": 64626 }, { "epoch": 1.25, "learning_rate": 2.9101242810654146e-05, "loss": 0.0, "step": 64628 }, { "epoch": 1.25, "learning_rate": 2.910059585039691e-05, "loss": 0.0, "step": 64630 }, { "epoch": 1.25, "learning_rate": 2.9099948890139677e-05, "loss": 0.0076, "step": 64632 }, { "epoch": 1.25, "learning_rate": 2.909930192988245e-05, "loss": 0.0, "step": 64634 }, { "epoch": 1.25, "learning_rate": 2.9098654969625216e-05, "loss": 0.0, "step": 64636 }, { "epoch": 1.25, "learning_rate": 2.9098008009367988e-05, "loss": 0.0225, "step": 64638 }, { "epoch": 1.25, "learning_rate": 2.9097361049110754e-05, "loss": 0.0, "step": 64640 }, { "epoch": 1.25, "learning_rate": 2.9096714088853523e-05, "loss": 0.0, "step": 64642 }, { "epoch": 1.25, "learning_rate": 2.9096067128596295e-05, "loss": 0.0, "step": 64644 }, { "epoch": 1.25, "learning_rate": 2.909542016833906e-05, "loss": 0.0, "step": 64646 }, { "epoch": 1.25, "learning_rate": 2.9094773208081827e-05, "loss": 0.0, "step": 64648 }, { "epoch": 1.25, "learning_rate": 2.90941262478246e-05, "loss": 0.0036, "step": 64650 }, { "epoch": 1.25, "learning_rate": 2.9093479287567365e-05, "loss": 0.0127, "step": 64652 }, { "epoch": 1.25, "learning_rate": 2.909283232731013e-05, "loss": 0.0, "step": 64654 }, { "epoch": 1.25, "learning_rate": 2.9092185367052903e-05, "loss": 0.0018, "step": 64656 }, { "epoch": 1.25, "learning_rate": 2.9091538406795672e-05, "loss": 0.0, "step": 64658 }, { "epoch": 1.25, "learning_rate": 2.9090891446538445e-05, "loss": 0.0, "step": 64660 }, { "epoch": 1.26, "learning_rate": 2.909024448628121e-05, "loss": 0.0, "step": 64662 }, { "epoch": 1.26, "learning_rate": 2.9089597526023976e-05, "loss": 0.0, "step": 64664 }, { "epoch": 1.26, "learning_rate": 2.908895056576675e-05, "loss": 0.0, "step": 64666 }, { "epoch": 1.26, "learning_rate": 2.9088303605509514e-05, "loss": 0.0, "step": 64668 }, { "epoch": 1.26, "learning_rate": 2.908765664525228e-05, "loss": 0.0001, "step": 64670 }, { "epoch": 1.26, "learning_rate": 2.9087009684995052e-05, "loss": 0.0, "step": 64672 }, { "epoch": 1.26, "learning_rate": 2.908636272473782e-05, "loss": 0.0, "step": 64674 }, { "epoch": 1.26, "learning_rate": 2.9085715764480587e-05, "loss": 0.0088, "step": 64676 }, { "epoch": 1.26, "learning_rate": 2.908506880422336e-05, "loss": 0.0, "step": 64678 }, { "epoch": 1.26, "learning_rate": 2.9084421843966125e-05, "loss": 0.0, "step": 64680 }, { "epoch": 1.26, "learning_rate": 2.9083774883708898e-05, "loss": 0.0, "step": 64682 }, { "epoch": 1.26, "learning_rate": 2.9083127923451663e-05, "loss": 0.0011, "step": 64684 }, { "epoch": 1.26, "learning_rate": 2.908248096319443e-05, "loss": 0.0, "step": 64686 }, { "epoch": 1.26, "learning_rate": 2.90818340029372e-05, "loss": 0.0005, "step": 64688 }, { "epoch": 1.26, "learning_rate": 2.908118704267997e-05, "loss": 0.0, "step": 64690 }, { "epoch": 1.26, "learning_rate": 2.9080540082422736e-05, "loss": 0.0, "step": 64692 }, { "epoch": 1.26, "learning_rate": 2.907989312216551e-05, "loss": 0.0, "step": 64694 }, { "epoch": 1.26, "learning_rate": 2.9079246161908275e-05, "loss": 0.0, "step": 64696 }, { "epoch": 1.26, "learning_rate": 2.9078599201651047e-05, "loss": 0.0, "step": 64698 }, { "epoch": 1.26, "learning_rate": 2.9077952241393813e-05, "loss": 0.0, "step": 64700 }, { "epoch": 1.26, "learning_rate": 2.907730528113658e-05, "loss": 0.0058, "step": 64702 }, { "epoch": 1.26, "learning_rate": 2.907665832087935e-05, "loss": 0.0, "step": 64704 }, { "epoch": 1.26, "learning_rate": 2.907601136062212e-05, "loss": 0.0, "step": 64706 }, { "epoch": 1.26, "learning_rate": 2.9075364400364886e-05, "loss": 0.0388, "step": 64708 }, { "epoch": 1.26, "learning_rate": 2.9074717440107658e-05, "loss": 0.003, "step": 64710 }, { "epoch": 1.26, "learning_rate": 2.9074070479850424e-05, "loss": 0.0, "step": 64712 }, { "epoch": 1.26, "learning_rate": 2.907342351959319e-05, "loss": 0.0, "step": 64714 }, { "epoch": 1.26, "learning_rate": 2.9072776559335962e-05, "loss": 0.0, "step": 64716 }, { "epoch": 1.26, "learning_rate": 2.9072129599078728e-05, "loss": 0.0, "step": 64718 }, { "epoch": 1.26, "learning_rate": 2.90714826388215e-05, "loss": 0.0, "step": 64720 }, { "epoch": 1.26, "learning_rate": 2.907083567856427e-05, "loss": 0.0124, "step": 64722 }, { "epoch": 1.26, "learning_rate": 2.9070188718307035e-05, "loss": 0.003, "step": 64724 }, { "epoch": 1.26, "learning_rate": 2.9069541758049807e-05, "loss": 0.0, "step": 64726 }, { "epoch": 1.26, "learning_rate": 2.9068894797792573e-05, "loss": 0.0001, "step": 64728 }, { "epoch": 1.26, "learning_rate": 2.906824783753534e-05, "loss": 0.0, "step": 64730 }, { "epoch": 1.26, "learning_rate": 2.906760087727811e-05, "loss": 0.0, "step": 64732 }, { "epoch": 1.26, "learning_rate": 2.9066953917020877e-05, "loss": 0.0005, "step": 64734 }, { "epoch": 1.26, "learning_rate": 2.9066306956763646e-05, "loss": 0.0073, "step": 64736 }, { "epoch": 1.26, "learning_rate": 2.906565999650642e-05, "loss": 0.0, "step": 64738 }, { "epoch": 1.26, "learning_rate": 2.9065013036249184e-05, "loss": 0.0, "step": 64740 }, { "epoch": 1.26, "learning_rate": 2.9064366075991957e-05, "loss": 0.0033, "step": 64742 }, { "epoch": 1.26, "learning_rate": 2.9063719115734722e-05, "loss": 0.0, "step": 64744 }, { "epoch": 1.26, "learning_rate": 2.9063072155477488e-05, "loss": 0.0101, "step": 64746 }, { "epoch": 1.26, "learning_rate": 2.906242519522026e-05, "loss": 0.0, "step": 64748 }, { "epoch": 1.26, "learning_rate": 2.9061778234963026e-05, "loss": 0.0028, "step": 64750 }, { "epoch": 1.26, "learning_rate": 2.9061131274705795e-05, "loss": 0.002, "step": 64752 }, { "epoch": 1.26, "learning_rate": 2.9060484314448568e-05, "loss": 0.0, "step": 64754 }, { "epoch": 1.26, "learning_rate": 2.9059837354191333e-05, "loss": 0.0, "step": 64756 }, { "epoch": 1.26, "learning_rate": 2.90591903939341e-05, "loss": 0.0001, "step": 64758 }, { "epoch": 1.26, "learning_rate": 2.905854343367687e-05, "loss": 0.0, "step": 64760 }, { "epoch": 1.26, "learning_rate": 2.9057896473419637e-05, "loss": 0.0003, "step": 64762 }, { "epoch": 1.26, "learning_rate": 2.905724951316241e-05, "loss": 0.0, "step": 64764 }, { "epoch": 1.26, "learning_rate": 2.9056602552905176e-05, "loss": 0.0, "step": 64766 }, { "epoch": 1.26, "learning_rate": 2.9055955592647945e-05, "loss": 0.0, "step": 64768 }, { "epoch": 1.26, "learning_rate": 2.9055308632390714e-05, "loss": 0.0, "step": 64770 }, { "epoch": 1.26, "learning_rate": 2.9054661672133483e-05, "loss": 0.0001, "step": 64772 }, { "epoch": 1.26, "learning_rate": 2.905401471187625e-05, "loss": 0.0, "step": 64774 }, { "epoch": 1.26, "learning_rate": 2.905336775161902e-05, "loss": 0.0, "step": 64776 }, { "epoch": 1.26, "learning_rate": 2.9052720791361787e-05, "loss": 0.0, "step": 64778 }, { "epoch": 1.26, "learning_rate": 2.905207383110456e-05, "loss": 0.0, "step": 64780 }, { "epoch": 1.26, "learning_rate": 2.9051426870847325e-05, "loss": 0.0005, "step": 64782 }, { "epoch": 1.26, "learning_rate": 2.9050779910590094e-05, "loss": 0.0, "step": 64784 }, { "epoch": 1.26, "learning_rate": 2.9050132950332863e-05, "loss": 0.0, "step": 64786 }, { "epoch": 1.26, "learning_rate": 2.9049485990075632e-05, "loss": 0.0, "step": 64788 }, { "epoch": 1.26, "learning_rate": 2.9048839029818398e-05, "loss": 0.0, "step": 64790 }, { "epoch": 1.26, "learning_rate": 2.904819206956117e-05, "loss": 0.0201, "step": 64792 }, { "epoch": 1.26, "learning_rate": 2.9047545109303936e-05, "loss": 0.0, "step": 64794 }, { "epoch": 1.26, "learning_rate": 2.90468981490467e-05, "loss": 0.0, "step": 64796 }, { "epoch": 1.26, "learning_rate": 2.9046251188789474e-05, "loss": 0.0, "step": 64798 }, { "epoch": 1.26, "learning_rate": 2.9045604228532243e-05, "loss": 0.0027, "step": 64800 }, { "epoch": 1.26, "learning_rate": 2.9044957268275012e-05, "loss": 0.0, "step": 64802 }, { "epoch": 1.26, "learning_rate": 2.904431030801778e-05, "loss": 0.0003, "step": 64804 }, { "epoch": 1.26, "learning_rate": 2.9043663347760547e-05, "loss": 0.0, "step": 64806 }, { "epoch": 1.26, "learning_rate": 2.904301638750332e-05, "loss": 0.0001, "step": 64808 }, { "epoch": 1.26, "learning_rate": 2.9042369427246085e-05, "loss": 0.0075, "step": 64810 }, { "epoch": 1.26, "learning_rate": 2.904172246698885e-05, "loss": 0.0002, "step": 64812 }, { "epoch": 1.26, "learning_rate": 2.9041075506731623e-05, "loss": 0.0004, "step": 64814 }, { "epoch": 1.26, "learning_rate": 2.9040428546474392e-05, "loss": 0.0004, "step": 64816 }, { "epoch": 1.26, "learning_rate": 2.9039781586217158e-05, "loss": 0.0, "step": 64818 }, { "epoch": 1.26, "learning_rate": 2.903913462595993e-05, "loss": 0.0, "step": 64820 }, { "epoch": 1.26, "learning_rate": 2.9038487665702696e-05, "loss": 0.0, "step": 64822 }, { "epoch": 1.26, "learning_rate": 2.903784070544547e-05, "loss": 0.0, "step": 64824 }, { "epoch": 1.26, "learning_rate": 2.9037193745188235e-05, "loss": 0.0, "step": 64826 }, { "epoch": 1.26, "learning_rate": 2.9036546784931e-05, "loss": 0.0, "step": 64828 }, { "epoch": 1.26, "learning_rate": 2.9035899824673773e-05, "loss": 0.0, "step": 64830 }, { "epoch": 1.26, "learning_rate": 2.903525286441654e-05, "loss": 0.0014, "step": 64832 }, { "epoch": 1.26, "learning_rate": 2.9034605904159307e-05, "loss": 0.0, "step": 64834 }, { "epoch": 1.26, "learning_rate": 2.903395894390208e-05, "loss": 0.0, "step": 64836 }, { "epoch": 1.26, "learning_rate": 2.9033311983644846e-05, "loss": 0.0, "step": 64838 }, { "epoch": 1.26, "learning_rate": 2.9032665023387618e-05, "loss": 0.0, "step": 64840 }, { "epoch": 1.26, "learning_rate": 2.9032018063130384e-05, "loss": 0.0, "step": 64842 }, { "epoch": 1.26, "learning_rate": 2.903137110287315e-05, "loss": 0.0, "step": 64844 }, { "epoch": 1.26, "learning_rate": 2.9030724142615922e-05, "loss": 0.0, "step": 64846 }, { "epoch": 1.26, "learning_rate": 2.9030077182358688e-05, "loss": 0.0, "step": 64848 }, { "epoch": 1.26, "learning_rate": 2.9029430222101457e-05, "loss": 0.001, "step": 64850 }, { "epoch": 1.26, "learning_rate": 2.902878326184423e-05, "loss": 0.0, "step": 64852 }, { "epoch": 1.26, "learning_rate": 2.9028136301586995e-05, "loss": 0.0, "step": 64854 }, { "epoch": 1.26, "learning_rate": 2.902748934132976e-05, "loss": 0.0, "step": 64856 }, { "epoch": 1.26, "learning_rate": 2.9026842381072533e-05, "loss": 0.0, "step": 64858 }, { "epoch": 1.26, "learning_rate": 2.90261954208153e-05, "loss": 0.0005, "step": 64860 }, { "epoch": 1.26, "learning_rate": 2.902554846055807e-05, "loss": 0.0, "step": 64862 }, { "epoch": 1.26, "learning_rate": 2.9024901500300837e-05, "loss": 0.0, "step": 64864 }, { "epoch": 1.26, "learning_rate": 2.9024254540043606e-05, "loss": 0.0, "step": 64866 }, { "epoch": 1.26, "learning_rate": 2.902360757978638e-05, "loss": 0.0, "step": 64868 }, { "epoch": 1.26, "learning_rate": 2.9022960619529144e-05, "loss": 0.0, "step": 64870 }, { "epoch": 1.26, "learning_rate": 2.902231365927191e-05, "loss": 0.0001, "step": 64872 }, { "epoch": 1.26, "learning_rate": 2.9021666699014682e-05, "loss": 0.0, "step": 64874 }, { "epoch": 1.26, "learning_rate": 2.9021019738757448e-05, "loss": 0.0, "step": 64876 }, { "epoch": 1.26, "learning_rate": 2.9020372778500217e-05, "loss": 0.0, "step": 64878 }, { "epoch": 1.26, "learning_rate": 2.9019725818242986e-05, "loss": 0.0017, "step": 64880 }, { "epoch": 1.26, "learning_rate": 2.9019078857985755e-05, "loss": 0.0, "step": 64882 }, { "epoch": 1.26, "learning_rate": 2.9018431897728528e-05, "loss": 0.0, "step": 64884 }, { "epoch": 1.26, "learning_rate": 2.9017784937471293e-05, "loss": 0.0, "step": 64886 }, { "epoch": 1.26, "learning_rate": 2.901713797721406e-05, "loss": 0.0, "step": 64888 }, { "epoch": 1.26, "learning_rate": 2.901649101695683e-05, "loss": 0.0011, "step": 64890 }, { "epoch": 1.26, "learning_rate": 2.9015844056699597e-05, "loss": 0.0, "step": 64892 }, { "epoch": 1.26, "learning_rate": 2.9015197096442363e-05, "loss": 0.0, "step": 64894 }, { "epoch": 1.26, "learning_rate": 2.9014550136185136e-05, "loss": 0.0, "step": 64896 }, { "epoch": 1.26, "learning_rate": 2.9013903175927905e-05, "loss": 0.0138, "step": 64898 }, { "epoch": 1.26, "learning_rate": 2.901325621567067e-05, "loss": 0.0054, "step": 64900 }, { "epoch": 1.26, "learning_rate": 2.9012609255413443e-05, "loss": 0.0, "step": 64902 }, { "epoch": 1.26, "learning_rate": 2.901196229515621e-05, "loss": 0.002, "step": 64904 }, { "epoch": 1.26, "learning_rate": 2.901131533489898e-05, "loss": 0.0003, "step": 64906 }, { "epoch": 1.26, "learning_rate": 2.9010668374641747e-05, "loss": 0.0, "step": 64908 }, { "epoch": 1.26, "learning_rate": 2.9010021414384512e-05, "loss": 0.0, "step": 64910 }, { "epoch": 1.26, "learning_rate": 2.9009374454127285e-05, "loss": 0.0094, "step": 64912 }, { "epoch": 1.26, "learning_rate": 2.9008727493870054e-05, "loss": 0.0001, "step": 64914 }, { "epoch": 1.26, "learning_rate": 2.900808053361282e-05, "loss": 0.0, "step": 64916 }, { "epoch": 1.26, "learning_rate": 2.9007433573355592e-05, "loss": 0.0, "step": 64918 }, { "epoch": 1.26, "learning_rate": 2.9006786613098358e-05, "loss": 0.0, "step": 64920 }, { "epoch": 1.26, "learning_rate": 2.900613965284113e-05, "loss": 0.0006, "step": 64922 }, { "epoch": 1.26, "learning_rate": 2.9005492692583896e-05, "loss": 0.0001, "step": 64924 }, { "epoch": 1.26, "learning_rate": 2.900484573232666e-05, "loss": 0.0, "step": 64926 }, { "epoch": 1.26, "learning_rate": 2.9004198772069434e-05, "loss": 0.0, "step": 64928 }, { "epoch": 1.26, "learning_rate": 2.9003551811812203e-05, "loss": 0.0, "step": 64930 }, { "epoch": 1.26, "learning_rate": 2.900290485155497e-05, "loss": 0.0, "step": 64932 }, { "epoch": 1.26, "learning_rate": 2.900225789129774e-05, "loss": 0.0001, "step": 64934 }, { "epoch": 1.26, "learning_rate": 2.9001610931040507e-05, "loss": 0.0, "step": 64936 }, { "epoch": 1.26, "learning_rate": 2.9000963970783273e-05, "loss": 0.0021, "step": 64938 }, { "epoch": 1.26, "learning_rate": 2.9000317010526045e-05, "loss": 0.0, "step": 64940 }, { "epoch": 1.26, "learning_rate": 2.899967005026881e-05, "loss": 0.0, "step": 64942 }, { "epoch": 1.26, "learning_rate": 2.8999023090011583e-05, "loss": 0.0, "step": 64944 }, { "epoch": 1.26, "learning_rate": 2.8998376129754352e-05, "loss": 0.0, "step": 64946 }, { "epoch": 1.26, "learning_rate": 2.8997729169497118e-05, "loss": 0.0001, "step": 64948 }, { "epoch": 1.26, "learning_rate": 2.899708220923989e-05, "loss": 0.0, "step": 64950 }, { "epoch": 1.26, "learning_rate": 2.8996435248982656e-05, "loss": 0.0, "step": 64952 }, { "epoch": 1.26, "learning_rate": 2.8995788288725422e-05, "loss": 0.0, "step": 64954 }, { "epoch": 1.26, "learning_rate": 2.8995141328468195e-05, "loss": 0.002, "step": 64956 }, { "epoch": 1.26, "learning_rate": 2.899449436821096e-05, "loss": 0.0, "step": 64958 }, { "epoch": 1.26, "learning_rate": 2.899384740795373e-05, "loss": 0.0, "step": 64960 }, { "epoch": 1.26, "learning_rate": 2.8993200447696502e-05, "loss": 0.0, "step": 64962 }, { "epoch": 1.26, "learning_rate": 2.8992553487439267e-05, "loss": 0.0, "step": 64964 }, { "epoch": 1.26, "learning_rate": 2.899190652718204e-05, "loss": 0.0, "step": 64966 }, { "epoch": 1.26, "learning_rate": 2.8991259566924806e-05, "loss": 0.0, "step": 64968 }, { "epoch": 1.26, "learning_rate": 2.899061260666757e-05, "loss": 0.0002, "step": 64970 }, { "epoch": 1.26, "learning_rate": 2.8989965646410344e-05, "loss": 0.0, "step": 64972 }, { "epoch": 1.26, "learning_rate": 2.898931868615311e-05, "loss": 0.0005, "step": 64974 }, { "epoch": 1.26, "learning_rate": 2.898867172589588e-05, "loss": 0.0, "step": 64976 }, { "epoch": 1.26, "learning_rate": 2.898802476563865e-05, "loss": 0.0, "step": 64978 }, { "epoch": 1.26, "learning_rate": 2.8987377805381417e-05, "loss": 0.0, "step": 64980 }, { "epoch": 1.26, "learning_rate": 2.898673084512419e-05, "loss": 0.0001, "step": 64982 }, { "epoch": 1.26, "learning_rate": 2.8986083884866955e-05, "loss": 0.0001, "step": 64984 }, { "epoch": 1.26, "learning_rate": 2.898543692460972e-05, "loss": 0.0037, "step": 64986 }, { "epoch": 1.26, "learning_rate": 2.8984789964352493e-05, "loss": 0.0, "step": 64988 }, { "epoch": 1.26, "learning_rate": 2.898414300409526e-05, "loss": 0.0, "step": 64990 }, { "epoch": 1.26, "learning_rate": 2.8983496043838028e-05, "loss": 0.0, "step": 64992 }, { "epoch": 1.26, "learning_rate": 2.89828490835808e-05, "loss": 0.0002, "step": 64994 }, { "epoch": 1.26, "learning_rate": 2.8982202123323566e-05, "loss": 0.0, "step": 64996 }, { "epoch": 1.26, "learning_rate": 2.8981555163066332e-05, "loss": 0.0, "step": 64998 }, { "epoch": 1.26, "learning_rate": 2.8980908202809104e-05, "loss": 0.0, "step": 65000 }, { "epoch": 1.26, "learning_rate": 2.898026124255187e-05, "loss": 0.0, "step": 65002 }, { "epoch": 1.26, "learning_rate": 2.8979614282294642e-05, "loss": 0.0, "step": 65004 }, { "epoch": 1.26, "learning_rate": 2.8978967322037408e-05, "loss": 0.0, "step": 65006 }, { "epoch": 1.26, "learning_rate": 2.8978320361780177e-05, "loss": 0.0032, "step": 65008 }, { "epoch": 1.26, "learning_rate": 2.8977673401522946e-05, "loss": 0.0002, "step": 65010 }, { "epoch": 1.26, "learning_rate": 2.8977026441265715e-05, "loss": 0.0, "step": 65012 }, { "epoch": 1.26, "learning_rate": 2.897637948100848e-05, "loss": 0.0156, "step": 65014 }, { "epoch": 1.26, "learning_rate": 2.8975732520751253e-05, "loss": 0.0, "step": 65016 }, { "epoch": 1.26, "learning_rate": 2.897508556049402e-05, "loss": 0.0, "step": 65018 }, { "epoch": 1.26, "learning_rate": 2.8974438600236785e-05, "loss": 0.0053, "step": 65020 }, { "epoch": 1.26, "learning_rate": 2.8973791639979557e-05, "loss": 0.0, "step": 65022 }, { "epoch": 1.26, "learning_rate": 2.8973144679722326e-05, "loss": 0.0, "step": 65024 }, { "epoch": 1.26, "learning_rate": 2.8972497719465096e-05, "loss": 0.0, "step": 65026 }, { "epoch": 1.26, "learning_rate": 2.8971850759207865e-05, "loss": 0.0081, "step": 65028 }, { "epoch": 1.26, "learning_rate": 2.897120379895063e-05, "loss": 0.0002, "step": 65030 }, { "epoch": 1.26, "learning_rate": 2.8970556838693403e-05, "loss": 0.0, "step": 65032 }, { "epoch": 1.26, "learning_rate": 2.896990987843617e-05, "loss": 0.0, "step": 65034 }, { "epoch": 1.26, "learning_rate": 2.8969262918178934e-05, "loss": 0.0, "step": 65036 }, { "epoch": 1.26, "learning_rate": 2.8968615957921707e-05, "loss": 0.0, "step": 65038 }, { "epoch": 1.26, "learning_rate": 2.8967968997664476e-05, "loss": 0.0, "step": 65040 }, { "epoch": 1.26, "learning_rate": 2.896732203740724e-05, "loss": 0.0, "step": 65042 }, { "epoch": 1.26, "learning_rate": 2.8966675077150014e-05, "loss": 0.0, "step": 65044 }, { "epoch": 1.26, "learning_rate": 2.896602811689278e-05, "loss": 0.0001, "step": 65046 }, { "epoch": 1.26, "learning_rate": 2.8965381156635552e-05, "loss": 0.0, "step": 65048 }, { "epoch": 1.26, "learning_rate": 2.8964734196378318e-05, "loss": 0.0003, "step": 65050 }, { "epoch": 1.26, "learning_rate": 2.8964087236121083e-05, "loss": 0.0005, "step": 65052 }, { "epoch": 1.26, "learning_rate": 2.8963440275863856e-05, "loss": 0.0006, "step": 65054 }, { "epoch": 1.26, "learning_rate": 2.8962793315606625e-05, "loss": 0.0, "step": 65056 }, { "epoch": 1.26, "learning_rate": 2.896214635534939e-05, "loss": 0.0001, "step": 65058 }, { "epoch": 1.26, "learning_rate": 2.8961499395092163e-05, "loss": 0.0, "step": 65060 }, { "epoch": 1.26, "learning_rate": 2.896085243483493e-05, "loss": 0.0, "step": 65062 }, { "epoch": 1.26, "learning_rate": 2.89602054745777e-05, "loss": 0.0, "step": 65064 }, { "epoch": 1.26, "learning_rate": 2.8959558514320467e-05, "loss": 0.0, "step": 65066 }, { "epoch": 1.26, "learning_rate": 2.8958911554063233e-05, "loss": 0.0, "step": 65068 }, { "epoch": 1.26, "learning_rate": 2.8958264593806005e-05, "loss": 0.0042, "step": 65070 }, { "epoch": 1.26, "learning_rate": 2.895761763354877e-05, "loss": 0.0011, "step": 65072 }, { "epoch": 1.26, "learning_rate": 2.895697067329154e-05, "loss": 0.0001, "step": 65074 }, { "epoch": 1.26, "learning_rate": 2.8956323713034312e-05, "loss": 0.0, "step": 65076 }, { "epoch": 1.26, "learning_rate": 2.8955676752777078e-05, "loss": 0.0002, "step": 65078 }, { "epoch": 1.26, "learning_rate": 2.8955029792519844e-05, "loss": 0.0001, "step": 65080 }, { "epoch": 1.26, "learning_rate": 2.8954382832262616e-05, "loss": 0.0002, "step": 65082 }, { "epoch": 1.26, "learning_rate": 2.8953735872005382e-05, "loss": 0.0, "step": 65084 }, { "epoch": 1.26, "learning_rate": 2.8953088911748154e-05, "loss": 0.0, "step": 65086 }, { "epoch": 1.26, "learning_rate": 2.895244195149092e-05, "loss": 0.0, "step": 65088 }, { "epoch": 1.26, "learning_rate": 2.895179499123369e-05, "loss": 0.0, "step": 65090 }, { "epoch": 1.26, "learning_rate": 2.8951148030976462e-05, "loss": 0.0, "step": 65092 }, { "epoch": 1.26, "learning_rate": 2.8950501070719227e-05, "loss": 0.0, "step": 65094 }, { "epoch": 1.26, "learning_rate": 2.8949854110461993e-05, "loss": 0.0001, "step": 65096 }, { "epoch": 1.26, "learning_rate": 2.8949207150204766e-05, "loss": 0.0001, "step": 65098 }, { "epoch": 1.26, "learning_rate": 2.894856018994753e-05, "loss": 0.0, "step": 65100 }, { "epoch": 1.26, "learning_rate": 2.89479132296903e-05, "loss": 0.0, "step": 65102 }, { "epoch": 1.26, "learning_rate": 2.894726626943307e-05, "loss": 0.0001, "step": 65104 }, { "epoch": 1.26, "learning_rate": 2.894661930917584e-05, "loss": 0.0001, "step": 65106 }, { "epoch": 1.26, "learning_rate": 2.894597234891861e-05, "loss": 0.068, "step": 65108 }, { "epoch": 1.26, "learning_rate": 2.8945325388661377e-05, "loss": 0.0, "step": 65110 }, { "epoch": 1.26, "learning_rate": 2.8944678428404142e-05, "loss": 0.0, "step": 65112 }, { "epoch": 1.26, "learning_rate": 2.8944031468146915e-05, "loss": 0.0, "step": 65114 }, { "epoch": 1.26, "learning_rate": 2.894338450788968e-05, "loss": 0.0, "step": 65116 }, { "epoch": 1.26, "learning_rate": 2.894273754763245e-05, "loss": 0.0, "step": 65118 }, { "epoch": 1.26, "learning_rate": 2.894209058737522e-05, "loss": 0.0, "step": 65120 }, { "epoch": 1.26, "learning_rate": 2.8941443627117988e-05, "loss": 0.0, "step": 65122 }, { "epoch": 1.26, "learning_rate": 2.894079666686076e-05, "loss": 0.0, "step": 65124 }, { "epoch": 1.26, "learning_rate": 2.8940149706603526e-05, "loss": 0.0, "step": 65126 }, { "epoch": 1.26, "learning_rate": 2.8939502746346292e-05, "loss": 0.0, "step": 65128 }, { "epoch": 1.26, "learning_rate": 2.8938855786089064e-05, "loss": 0.0, "step": 65130 }, { "epoch": 1.26, "learning_rate": 2.893820882583183e-05, "loss": 0.0, "step": 65132 }, { "epoch": 1.26, "learning_rate": 2.8937561865574596e-05, "loss": 0.0, "step": 65134 }, { "epoch": 1.26, "learning_rate": 2.8936914905317368e-05, "loss": 0.0, "step": 65136 }, { "epoch": 1.26, "learning_rate": 2.8936267945060137e-05, "loss": 0.0, "step": 65138 }, { "epoch": 1.26, "learning_rate": 2.8935620984802903e-05, "loss": 0.0001, "step": 65140 }, { "epoch": 1.26, "learning_rate": 2.8934974024545675e-05, "loss": 0.0, "step": 65142 }, { "epoch": 1.26, "learning_rate": 2.893432706428844e-05, "loss": 0.0, "step": 65144 }, { "epoch": 1.26, "learning_rate": 2.8933680104031213e-05, "loss": 0.0, "step": 65146 }, { "epoch": 1.26, "learning_rate": 2.893303314377398e-05, "loss": 0.0, "step": 65148 }, { "epoch": 1.26, "learning_rate": 2.8932386183516745e-05, "loss": 0.0, "step": 65150 }, { "epoch": 1.26, "learning_rate": 2.8931739223259517e-05, "loss": 0.0, "step": 65152 }, { "epoch": 1.26, "learning_rate": 2.8931092263002286e-05, "loss": 0.0, "step": 65154 }, { "epoch": 1.26, "learning_rate": 2.8930445302745052e-05, "loss": 0.0, "step": 65156 }, { "epoch": 1.26, "learning_rate": 2.8929798342487825e-05, "loss": 0.0001, "step": 65158 }, { "epoch": 1.26, "learning_rate": 2.892915138223059e-05, "loss": 0.0, "step": 65160 }, { "epoch": 1.26, "learning_rate": 2.8928504421973356e-05, "loss": 0.0, "step": 65162 }, { "epoch": 1.26, "learning_rate": 2.892785746171613e-05, "loss": 0.0, "step": 65164 }, { "epoch": 1.26, "learning_rate": 2.8927210501458894e-05, "loss": 0.0, "step": 65166 }, { "epoch": 1.26, "learning_rate": 2.8926563541201667e-05, "loss": 0.0001, "step": 65168 }, { "epoch": 1.26, "learning_rate": 2.8925916580944436e-05, "loss": 0.0, "step": 65170 }, { "epoch": 1.26, "learning_rate": 2.89252696206872e-05, "loss": 0.0, "step": 65172 }, { "epoch": 1.26, "learning_rate": 2.8924622660429974e-05, "loss": 0.0, "step": 65174 }, { "epoch": 1.26, "learning_rate": 2.892397570017274e-05, "loss": 0.0004, "step": 65176 }, { "epoch": 1.27, "learning_rate": 2.8923328739915505e-05, "loss": 0.0, "step": 65178 }, { "epoch": 1.27, "learning_rate": 2.8922681779658278e-05, "loss": 0.0, "step": 65180 }, { "epoch": 1.27, "learning_rate": 2.8922034819401043e-05, "loss": 0.0, "step": 65182 }, { "epoch": 1.27, "learning_rate": 2.8921387859143813e-05, "loss": 0.0, "step": 65184 }, { "epoch": 1.27, "learning_rate": 2.8920740898886585e-05, "loss": 0.0001, "step": 65186 }, { "epoch": 1.27, "learning_rate": 2.892009393862935e-05, "loss": 0.0, "step": 65188 }, { "epoch": 1.27, "learning_rate": 2.8919446978372123e-05, "loss": 0.0, "step": 65190 }, { "epoch": 1.27, "learning_rate": 2.891880001811489e-05, "loss": 0.0, "step": 65192 }, { "epoch": 1.27, "learning_rate": 2.8918153057857655e-05, "loss": 0.0, "step": 65194 }, { "epoch": 1.27, "learning_rate": 2.8917506097600427e-05, "loss": 0.0002, "step": 65196 }, { "epoch": 1.27, "learning_rate": 2.8916859137343193e-05, "loss": 0.0053, "step": 65198 }, { "epoch": 1.27, "learning_rate": 2.8916212177085962e-05, "loss": 0.0, "step": 65200 }, { "epoch": 1.27, "learning_rate": 2.8915565216828734e-05, "loss": 0.0, "step": 65202 }, { "epoch": 1.27, "learning_rate": 2.89149182565715e-05, "loss": 0.0, "step": 65204 }, { "epoch": 1.27, "learning_rate": 2.8914271296314272e-05, "loss": 0.0, "step": 65206 }, { "epoch": 1.27, "learning_rate": 2.8913624336057038e-05, "loss": 0.0, "step": 65208 }, { "epoch": 1.27, "learning_rate": 2.8912977375799804e-05, "loss": 0.0, "step": 65210 }, { "epoch": 1.27, "learning_rate": 2.8912330415542576e-05, "loss": 0.0003, "step": 65212 }, { "epoch": 1.27, "learning_rate": 2.8911683455285342e-05, "loss": 0.0, "step": 65214 }, { "epoch": 1.27, "learning_rate": 2.891103649502811e-05, "loss": 0.0006, "step": 65216 }, { "epoch": 1.27, "learning_rate": 2.8910389534770884e-05, "loss": 0.0, "step": 65218 }, { "epoch": 1.27, "learning_rate": 2.890974257451365e-05, "loss": 0.0158, "step": 65220 }, { "epoch": 1.27, "learning_rate": 2.8909095614256415e-05, "loss": 0.0, "step": 65222 }, { "epoch": 1.27, "learning_rate": 2.8908448653999187e-05, "loss": 0.0, "step": 65224 }, { "epoch": 1.27, "learning_rate": 2.8907801693741953e-05, "loss": 0.0, "step": 65226 }, { "epoch": 1.27, "learning_rate": 2.8907154733484726e-05, "loss": 0.0, "step": 65228 }, { "epoch": 1.27, "learning_rate": 2.890650777322749e-05, "loss": 0.0, "step": 65230 }, { "epoch": 1.27, "learning_rate": 2.890586081297026e-05, "loss": 0.0, "step": 65232 }, { "epoch": 1.27, "learning_rate": 2.8905213852713033e-05, "loss": 0.0, "step": 65234 }, { "epoch": 1.27, "learning_rate": 2.89045668924558e-05, "loss": 0.0, "step": 65236 }, { "epoch": 1.27, "learning_rate": 2.8903919932198564e-05, "loss": 0.01, "step": 65238 }, { "epoch": 1.27, "learning_rate": 2.8903272971941337e-05, "loss": 0.0, "step": 65240 }, { "epoch": 1.27, "learning_rate": 2.8902626011684102e-05, "loss": 0.0, "step": 65242 }, { "epoch": 1.27, "learning_rate": 2.8901979051426868e-05, "loss": 0.0, "step": 65244 }, { "epoch": 1.27, "learning_rate": 2.890133209116964e-05, "loss": 0.0118, "step": 65246 }, { "epoch": 1.27, "learning_rate": 2.890068513091241e-05, "loss": 0.0, "step": 65248 }, { "epoch": 1.27, "learning_rate": 2.890003817065518e-05, "loss": 0.0, "step": 65250 }, { "epoch": 1.27, "learning_rate": 2.8899391210397948e-05, "loss": 0.0, "step": 65252 }, { "epoch": 1.27, "learning_rate": 2.8898744250140714e-05, "loss": 0.0019, "step": 65254 }, { "epoch": 1.27, "learning_rate": 2.8898097289883486e-05, "loss": 0.0, "step": 65256 }, { "epoch": 1.27, "learning_rate": 2.8897450329626252e-05, "loss": 0.0, "step": 65258 }, { "epoch": 1.27, "learning_rate": 2.8896803369369017e-05, "loss": 0.0, "step": 65260 }, { "epoch": 1.27, "learning_rate": 2.889615640911179e-05, "loss": 0.0, "step": 65262 }, { "epoch": 1.27, "learning_rate": 2.889550944885456e-05, "loss": 0.0, "step": 65264 }, { "epoch": 1.27, "learning_rate": 2.8894862488597328e-05, "loss": 0.0207, "step": 65266 }, { "epoch": 1.27, "learning_rate": 2.8894215528340097e-05, "loss": 0.0, "step": 65268 }, { "epoch": 1.27, "learning_rate": 2.8893568568082863e-05, "loss": 0.0001, "step": 65270 }, { "epoch": 1.27, "learning_rate": 2.8892921607825635e-05, "loss": 0.0001, "step": 65272 }, { "epoch": 1.27, "learning_rate": 2.88922746475684e-05, "loss": 0.0, "step": 65274 }, { "epoch": 1.27, "learning_rate": 2.8891627687311167e-05, "loss": 0.0, "step": 65276 }, { "epoch": 1.27, "learning_rate": 2.889098072705394e-05, "loss": 0.0, "step": 65278 }, { "epoch": 1.27, "learning_rate": 2.8890333766796708e-05, "loss": 0.0, "step": 65280 }, { "epoch": 1.27, "learning_rate": 2.8889686806539474e-05, "loss": 0.0001, "step": 65282 }, { "epoch": 1.27, "learning_rate": 2.8889039846282246e-05, "loss": 0.0043, "step": 65284 }, { "epoch": 1.27, "learning_rate": 2.8888392886025012e-05, "loss": 0.0001, "step": 65286 }, { "epoch": 1.27, "learning_rate": 2.8887745925767785e-05, "loss": 0.0, "step": 65288 }, { "epoch": 1.27, "learning_rate": 2.888709896551055e-05, "loss": 0.0, "step": 65290 }, { "epoch": 1.27, "learning_rate": 2.8886452005253316e-05, "loss": 0.0, "step": 65292 }, { "epoch": 1.27, "learning_rate": 2.888580504499609e-05, "loss": 0.0, "step": 65294 }, { "epoch": 1.27, "learning_rate": 2.8885158084738858e-05, "loss": 0.0, "step": 65296 }, { "epoch": 1.27, "learning_rate": 2.8884511124481623e-05, "loss": 0.0, "step": 65298 }, { "epoch": 1.27, "learning_rate": 2.8883864164224396e-05, "loss": 0.0001, "step": 65300 }, { "epoch": 1.27, "learning_rate": 2.888321720396716e-05, "loss": 0.0, "step": 65302 }, { "epoch": 1.27, "learning_rate": 2.8882570243709927e-05, "loss": 0.0, "step": 65304 }, { "epoch": 1.27, "learning_rate": 2.88819232834527e-05, "loss": 0.0098, "step": 65306 }, { "epoch": 1.27, "learning_rate": 2.8881276323195465e-05, "loss": 0.0007, "step": 65308 }, { "epoch": 1.27, "learning_rate": 2.8880629362938238e-05, "loss": 0.0, "step": 65310 }, { "epoch": 1.27, "learning_rate": 2.8879982402681003e-05, "loss": 0.0, "step": 65312 }, { "epoch": 1.27, "learning_rate": 2.8879335442423773e-05, "loss": 0.0, "step": 65314 }, { "epoch": 1.27, "learning_rate": 2.8878688482166545e-05, "loss": 0.0001, "step": 65316 }, { "epoch": 1.27, "learning_rate": 2.887804152190931e-05, "loss": 0.0, "step": 65318 }, { "epoch": 1.27, "learning_rate": 2.8877394561652076e-05, "loss": 0.0, "step": 65320 }, { "epoch": 1.27, "learning_rate": 2.887674760139485e-05, "loss": 0.0, "step": 65322 }, { "epoch": 1.27, "learning_rate": 2.8876100641137615e-05, "loss": 0.0, "step": 65324 }, { "epoch": 1.27, "learning_rate": 2.8875453680880384e-05, "loss": 0.0001, "step": 65326 }, { "epoch": 1.27, "learning_rate": 2.8874806720623153e-05, "loss": 0.0, "step": 65328 }, { "epoch": 1.27, "learning_rate": 2.8874159760365922e-05, "loss": 0.0, "step": 65330 }, { "epoch": 1.27, "learning_rate": 2.8873512800108694e-05, "loss": 0.0, "step": 65332 }, { "epoch": 1.27, "learning_rate": 2.887286583985146e-05, "loss": 0.0, "step": 65334 }, { "epoch": 1.27, "learning_rate": 2.8872218879594226e-05, "loss": 0.0, "step": 65336 }, { "epoch": 1.27, "learning_rate": 2.8871571919336998e-05, "loss": 0.0, "step": 65338 }, { "epoch": 1.27, "learning_rate": 2.8870924959079764e-05, "loss": 0.0001, "step": 65340 }, { "epoch": 1.27, "learning_rate": 2.8870277998822533e-05, "loss": 0.0018, "step": 65342 }, { "epoch": 1.27, "learning_rate": 2.8869631038565302e-05, "loss": 0.0, "step": 65344 }, { "epoch": 1.27, "learning_rate": 2.886898407830807e-05, "loss": 0.0055, "step": 65346 }, { "epoch": 1.27, "learning_rate": 2.8868337118050844e-05, "loss": 0.0, "step": 65348 }, { "epoch": 1.27, "learning_rate": 2.886769015779361e-05, "loss": 0.0, "step": 65350 }, { "epoch": 1.27, "learning_rate": 2.8867043197536375e-05, "loss": 0.0, "step": 65352 }, { "epoch": 1.27, "learning_rate": 2.8866396237279147e-05, "loss": 0.0, "step": 65354 }, { "epoch": 1.27, "learning_rate": 2.8865749277021913e-05, "loss": 0.0118, "step": 65356 }, { "epoch": 1.27, "learning_rate": 2.8865102316764682e-05, "loss": 0.0004, "step": 65358 }, { "epoch": 1.27, "learning_rate": 2.886445535650745e-05, "loss": 0.0, "step": 65360 }, { "epoch": 1.27, "learning_rate": 2.886380839625022e-05, "loss": 0.0302, "step": 65362 }, { "epoch": 1.27, "learning_rate": 2.8863161435992986e-05, "loss": 0.0, "step": 65364 }, { "epoch": 1.27, "learning_rate": 2.886251447573576e-05, "loss": 0.0, "step": 65366 }, { "epoch": 1.27, "learning_rate": 2.8861867515478524e-05, "loss": 0.0008, "step": 65368 }, { "epoch": 1.27, "learning_rate": 2.8861220555221297e-05, "loss": 0.0, "step": 65370 }, { "epoch": 1.27, "learning_rate": 2.8860573594964062e-05, "loss": 0.003, "step": 65372 }, { "epoch": 1.27, "learning_rate": 2.8859926634706828e-05, "loss": 0.0, "step": 65374 }, { "epoch": 1.27, "learning_rate": 2.88592796744496e-05, "loss": 0.0, "step": 65376 }, { "epoch": 1.27, "learning_rate": 2.885863271419237e-05, "loss": 0.0, "step": 65378 }, { "epoch": 1.27, "learning_rate": 2.8857985753935135e-05, "loss": 0.0001, "step": 65380 }, { "epoch": 1.27, "learning_rate": 2.8857338793677908e-05, "loss": 0.0, "step": 65382 }, { "epoch": 1.27, "learning_rate": 2.8856691833420674e-05, "loss": 0.0, "step": 65384 }, { "epoch": 1.27, "learning_rate": 2.885604487316344e-05, "loss": 0.0066, "step": 65386 }, { "epoch": 1.27, "learning_rate": 2.8855397912906212e-05, "loss": 0.0, "step": 65388 }, { "epoch": 1.27, "learning_rate": 2.8854750952648977e-05, "loss": 0.0002, "step": 65390 }, { "epoch": 1.27, "learning_rate": 2.885410399239175e-05, "loss": 0.0, "step": 65392 }, { "epoch": 1.27, "learning_rate": 2.885345703213452e-05, "loss": 0.0, "step": 65394 }, { "epoch": 1.27, "learning_rate": 2.8852810071877285e-05, "loss": 0.0029, "step": 65396 }, { "epoch": 1.27, "learning_rate": 2.8852163111620057e-05, "loss": 0.0, "step": 65398 }, { "epoch": 1.27, "learning_rate": 2.8851516151362823e-05, "loss": 0.0, "step": 65400 }, { "epoch": 1.27, "learning_rate": 2.885086919110559e-05, "loss": 0.0, "step": 65402 }, { "epoch": 1.27, "learning_rate": 2.885022223084836e-05, "loss": 0.0, "step": 65404 }, { "epoch": 1.27, "learning_rate": 2.8849575270591127e-05, "loss": 0.0, "step": 65406 }, { "epoch": 1.27, "learning_rate": 2.88489283103339e-05, "loss": 0.0009, "step": 65408 }, { "epoch": 1.27, "learning_rate": 2.8848281350076668e-05, "loss": 0.0001, "step": 65410 }, { "epoch": 1.27, "learning_rate": 2.8847634389819434e-05, "loss": 0.0, "step": 65412 }, { "epoch": 1.27, "learning_rate": 2.8846987429562206e-05, "loss": 0.0, "step": 65414 }, { "epoch": 1.27, "learning_rate": 2.8846340469304972e-05, "loss": 0.0, "step": 65416 }, { "epoch": 1.27, "learning_rate": 2.8845693509047738e-05, "loss": 0.0, "step": 65418 }, { "epoch": 1.27, "learning_rate": 2.884504654879051e-05, "loss": 0.0, "step": 65420 }, { "epoch": 1.27, "learning_rate": 2.8844399588533276e-05, "loss": 0.0, "step": 65422 }, { "epoch": 1.27, "learning_rate": 2.8843752628276045e-05, "loss": 0.0065, "step": 65424 }, { "epoch": 1.27, "learning_rate": 2.8843105668018818e-05, "loss": 0.0001, "step": 65426 }, { "epoch": 1.27, "learning_rate": 2.8842458707761583e-05, "loss": 0.0, "step": 65428 }, { "epoch": 1.27, "learning_rate": 2.8841811747504356e-05, "loss": 0.0, "step": 65430 }, { "epoch": 1.27, "learning_rate": 2.884116478724712e-05, "loss": 0.0, "step": 65432 }, { "epoch": 1.27, "learning_rate": 2.8840517826989887e-05, "loss": 0.0, "step": 65434 }, { "epoch": 1.27, "learning_rate": 2.883987086673266e-05, "loss": 0.0, "step": 65436 }, { "epoch": 1.27, "learning_rate": 2.8839223906475425e-05, "loss": 0.0, "step": 65438 }, { "epoch": 1.27, "learning_rate": 2.8838576946218194e-05, "loss": 0.0, "step": 65440 }, { "epoch": 1.27, "learning_rate": 2.8837929985960967e-05, "loss": 0.0, "step": 65442 }, { "epoch": 1.27, "learning_rate": 2.8837283025703733e-05, "loss": 0.0001, "step": 65444 }, { "epoch": 1.27, "learning_rate": 2.8836636065446498e-05, "loss": 0.0, "step": 65446 }, { "epoch": 1.27, "learning_rate": 2.883598910518927e-05, "loss": 0.0, "step": 65448 }, { "epoch": 1.27, "learning_rate": 2.8835342144932036e-05, "loss": 0.0001, "step": 65450 }, { "epoch": 1.27, "learning_rate": 2.883469518467481e-05, "loss": 0.0005, "step": 65452 }, { "epoch": 1.27, "learning_rate": 2.8834048224417575e-05, "loss": 0.0, "step": 65454 }, { "epoch": 1.27, "learning_rate": 2.8833401264160344e-05, "loss": 0.0, "step": 65456 }, { "epoch": 1.27, "learning_rate": 2.8832754303903116e-05, "loss": 0.0009, "step": 65458 }, { "epoch": 1.27, "learning_rate": 2.8832107343645882e-05, "loss": 0.0, "step": 65460 }, { "epoch": 1.27, "learning_rate": 2.8831460383388647e-05, "loss": 0.0, "step": 65462 }, { "epoch": 1.27, "learning_rate": 2.883081342313142e-05, "loss": 0.0002, "step": 65464 }, { "epoch": 1.27, "learning_rate": 2.8830166462874186e-05, "loss": 0.0, "step": 65466 }, { "epoch": 1.27, "learning_rate": 2.882951950261695e-05, "loss": 0.0, "step": 65468 }, { "epoch": 1.27, "learning_rate": 2.8828872542359724e-05, "loss": 0.0, "step": 65470 }, { "epoch": 1.27, "learning_rate": 2.8828225582102493e-05, "loss": 0.0, "step": 65472 }, { "epoch": 1.27, "learning_rate": 2.8827578621845265e-05, "loss": 0.0, "step": 65474 }, { "epoch": 1.27, "learning_rate": 2.882693166158803e-05, "loss": 0.0, "step": 65476 }, { "epoch": 1.27, "learning_rate": 2.8826284701330797e-05, "loss": 0.0, "step": 65478 }, { "epoch": 1.27, "learning_rate": 2.882563774107357e-05, "loss": 0.0, "step": 65480 }, { "epoch": 1.27, "learning_rate": 2.8824990780816335e-05, "loss": 0.0006, "step": 65482 }, { "epoch": 1.27, "learning_rate": 2.88243438205591e-05, "loss": 0.0, "step": 65484 }, { "epoch": 1.27, "learning_rate": 2.8823696860301873e-05, "loss": 0.0, "step": 65486 }, { "epoch": 1.27, "learning_rate": 2.8823049900044642e-05, "loss": 0.0, "step": 65488 }, { "epoch": 1.27, "learning_rate": 2.882240293978741e-05, "loss": 0.0009, "step": 65490 }, { "epoch": 1.27, "learning_rate": 2.882175597953018e-05, "loss": 0.0001, "step": 65492 }, { "epoch": 1.27, "learning_rate": 2.8821109019272946e-05, "loss": 0.0001, "step": 65494 }, { "epoch": 1.27, "learning_rate": 2.882046205901572e-05, "loss": 0.0105, "step": 65496 }, { "epoch": 1.27, "learning_rate": 2.8819815098758484e-05, "loss": 0.0, "step": 65498 }, { "epoch": 1.27, "learning_rate": 2.881916813850125e-05, "loss": 0.0, "step": 65500 }, { "epoch": 1.27, "learning_rate": 2.8818521178244022e-05, "loss": 0.0, "step": 65502 }, { "epoch": 1.27, "learning_rate": 2.881787421798679e-05, "loss": 0.0, "step": 65504 }, { "epoch": 1.27, "learning_rate": 2.8817227257729557e-05, "loss": 0.0, "step": 65506 }, { "epoch": 1.27, "learning_rate": 2.881658029747233e-05, "loss": 0.0102, "step": 65508 }, { "epoch": 1.27, "learning_rate": 2.8815933337215095e-05, "loss": 0.0042, "step": 65510 }, { "epoch": 1.27, "learning_rate": 2.8815286376957868e-05, "loss": 0.0, "step": 65512 }, { "epoch": 1.27, "learning_rate": 2.8814639416700634e-05, "loss": 0.0002, "step": 65514 }, { "epoch": 1.27, "learning_rate": 2.88139924564434e-05, "loss": 0.0, "step": 65516 }, { "epoch": 1.27, "learning_rate": 2.881334549618617e-05, "loss": 0.0, "step": 65518 }, { "epoch": 1.27, "learning_rate": 2.881269853592894e-05, "loss": 0.0003, "step": 65520 }, { "epoch": 1.27, "learning_rate": 2.8812051575671706e-05, "loss": 0.0, "step": 65522 }, { "epoch": 1.27, "learning_rate": 2.881140461541448e-05, "loss": 0.0003, "step": 65524 }, { "epoch": 1.27, "learning_rate": 2.8810757655157245e-05, "loss": 0.0, "step": 65526 }, { "epoch": 1.27, "learning_rate": 2.881011069490001e-05, "loss": 0.0109, "step": 65528 }, { "epoch": 1.27, "learning_rate": 2.8809463734642783e-05, "loss": 0.0007, "step": 65530 }, { "epoch": 1.27, "learning_rate": 2.880881677438555e-05, "loss": 0.0, "step": 65532 }, { "epoch": 1.27, "learning_rate": 2.880816981412832e-05, "loss": 0.0, "step": 65534 }, { "epoch": 1.27, "learning_rate": 2.880752285387109e-05, "loss": 0.0005, "step": 65536 }, { "epoch": 1.27, "learning_rate": 2.8806875893613856e-05, "loss": 0.0, "step": 65538 }, { "epoch": 1.27, "learning_rate": 2.8806228933356628e-05, "loss": 0.0, "step": 65540 }, { "epoch": 1.27, "learning_rate": 2.8805581973099394e-05, "loss": 0.0002, "step": 65542 }, { "epoch": 1.27, "learning_rate": 2.880493501284216e-05, "loss": 0.0215, "step": 65544 }, { "epoch": 1.27, "learning_rate": 2.8804288052584932e-05, "loss": 0.0003, "step": 65546 }, { "epoch": 1.27, "learning_rate": 2.8803641092327698e-05, "loss": 0.0, "step": 65548 }, { "epoch": 1.27, "learning_rate": 2.880299413207047e-05, "loss": 0.0, "step": 65550 }, { "epoch": 1.27, "learning_rate": 2.8802347171813236e-05, "loss": 0.0, "step": 65552 }, { "epoch": 1.27, "learning_rate": 2.8801700211556005e-05, "loss": 0.0001, "step": 65554 }, { "epoch": 1.27, "learning_rate": 2.8801053251298778e-05, "loss": 0.0, "step": 65556 }, { "epoch": 1.27, "learning_rate": 2.8800406291041543e-05, "loss": 0.0001, "step": 65558 }, { "epoch": 1.27, "learning_rate": 2.879975933078431e-05, "loss": 0.002, "step": 65560 }, { "epoch": 1.27, "learning_rate": 2.879911237052708e-05, "loss": 0.0, "step": 65562 }, { "epoch": 1.27, "learning_rate": 2.8798465410269847e-05, "loss": 0.0, "step": 65564 }, { "epoch": 1.27, "learning_rate": 2.8797818450012616e-05, "loss": 0.0, "step": 65566 }, { "epoch": 1.27, "learning_rate": 2.8797171489755385e-05, "loss": 0.0, "step": 65568 }, { "epoch": 1.27, "learning_rate": 2.8796524529498154e-05, "loss": 0.0064, "step": 65570 }, { "epoch": 1.27, "learning_rate": 2.8795877569240927e-05, "loss": 0.0, "step": 65572 }, { "epoch": 1.27, "learning_rate": 2.8795230608983692e-05, "loss": 0.0, "step": 65574 }, { "epoch": 1.27, "learning_rate": 2.8794583648726458e-05, "loss": 0.0004, "step": 65576 }, { "epoch": 1.27, "learning_rate": 2.879393668846923e-05, "loss": 0.0, "step": 65578 }, { "epoch": 1.27, "learning_rate": 2.8793289728211996e-05, "loss": 0.0, "step": 65580 }, { "epoch": 1.27, "learning_rate": 2.8792642767954765e-05, "loss": 0.0, "step": 65582 }, { "epoch": 1.27, "learning_rate": 2.8791995807697535e-05, "loss": 0.0, "step": 65584 }, { "epoch": 1.27, "learning_rate": 2.8791348847440304e-05, "loss": 0.0, "step": 65586 }, { "epoch": 1.27, "learning_rate": 2.879070188718307e-05, "loss": 0.0, "step": 65588 }, { "epoch": 1.27, "learning_rate": 2.8790054926925842e-05, "loss": 0.0, "step": 65590 }, { "epoch": 1.27, "learning_rate": 2.8789407966668607e-05, "loss": 0.0, "step": 65592 }, { "epoch": 1.27, "learning_rate": 2.878876100641138e-05, "loss": 0.0, "step": 65594 }, { "epoch": 1.27, "learning_rate": 2.8788114046154146e-05, "loss": 0.0, "step": 65596 }, { "epoch": 1.27, "learning_rate": 2.8787467085896915e-05, "loss": 0.0, "step": 65598 }, { "epoch": 1.27, "learning_rate": 2.8786820125639684e-05, "loss": 0.0, "step": 65600 }, { "epoch": 1.27, "learning_rate": 2.8786173165382453e-05, "loss": 0.0, "step": 65602 }, { "epoch": 1.27, "learning_rate": 2.878552620512522e-05, "loss": 0.0, "step": 65604 }, { "epoch": 1.27, "learning_rate": 2.878487924486799e-05, "loss": 0.0061, "step": 65606 }, { "epoch": 1.27, "learning_rate": 2.8784232284610757e-05, "loss": 0.0, "step": 65608 }, { "epoch": 1.27, "learning_rate": 2.8783585324353522e-05, "loss": 0.0, "step": 65610 }, { "epoch": 1.27, "learning_rate": 2.8782938364096295e-05, "loss": 0.0, "step": 65612 }, { "epoch": 1.27, "learning_rate": 2.878229140383906e-05, "loss": 0.0, "step": 65614 }, { "epoch": 1.27, "learning_rate": 2.8781644443581833e-05, "loss": 0.0001, "step": 65616 }, { "epoch": 1.27, "learning_rate": 2.8780997483324602e-05, "loss": 0.0, "step": 65618 }, { "epoch": 1.27, "learning_rate": 2.8780350523067368e-05, "loss": 0.0, "step": 65620 }, { "epoch": 1.27, "learning_rate": 2.877970356281014e-05, "loss": 0.0, "step": 65622 }, { "epoch": 1.27, "learning_rate": 2.8779056602552906e-05, "loss": 0.0013, "step": 65624 }, { "epoch": 1.27, "learning_rate": 2.8778409642295672e-05, "loss": 0.0, "step": 65626 }, { "epoch": 1.27, "learning_rate": 2.8777762682038444e-05, "loss": 0.0, "step": 65628 }, { "epoch": 1.27, "learning_rate": 2.877711572178121e-05, "loss": 0.0, "step": 65630 }, { "epoch": 1.27, "learning_rate": 2.8776468761523982e-05, "loss": 0.0, "step": 65632 }, { "epoch": 1.27, "learning_rate": 2.877582180126675e-05, "loss": 0.0, "step": 65634 }, { "epoch": 1.27, "learning_rate": 2.8775174841009517e-05, "loss": 0.0001, "step": 65636 }, { "epoch": 1.27, "learning_rate": 2.877452788075229e-05, "loss": 0.0, "step": 65638 }, { "epoch": 1.27, "learning_rate": 2.8773880920495055e-05, "loss": 0.0, "step": 65640 }, { "epoch": 1.27, "learning_rate": 2.877323396023782e-05, "loss": 0.0045, "step": 65642 }, { "epoch": 1.27, "learning_rate": 2.8772586999980594e-05, "loss": 0.0, "step": 65644 }, { "epoch": 1.27, "learning_rate": 2.877194003972336e-05, "loss": 0.01, "step": 65646 }, { "epoch": 1.27, "learning_rate": 2.8771293079466128e-05, "loss": 0.0, "step": 65648 }, { "epoch": 1.27, "learning_rate": 2.87706461192089e-05, "loss": 0.0, "step": 65650 }, { "epoch": 1.27, "learning_rate": 2.8769999158951666e-05, "loss": 0.0, "step": 65652 }, { "epoch": 1.27, "learning_rate": 2.876935219869444e-05, "loss": 0.0, "step": 65654 }, { "epoch": 1.27, "learning_rate": 2.8768705238437205e-05, "loss": 0.0, "step": 65656 }, { "epoch": 1.27, "learning_rate": 2.876805827817997e-05, "loss": 0.0, "step": 65658 }, { "epoch": 1.27, "learning_rate": 2.8767411317922743e-05, "loss": 0.0, "step": 65660 }, { "epoch": 1.27, "learning_rate": 2.876676435766551e-05, "loss": 0.0001, "step": 65662 }, { "epoch": 1.27, "learning_rate": 2.8766117397408278e-05, "loss": 0.0, "step": 65664 }, { "epoch": 1.27, "learning_rate": 2.876547043715105e-05, "loss": 0.0, "step": 65666 }, { "epoch": 1.27, "learning_rate": 2.8764823476893816e-05, "loss": 0.0095, "step": 65668 }, { "epoch": 1.27, "learning_rate": 2.876417651663658e-05, "loss": 0.0, "step": 65670 }, { "epoch": 1.27, "learning_rate": 2.8763529556379354e-05, "loss": 0.0, "step": 65672 }, { "epoch": 1.27, "learning_rate": 2.876288259612212e-05, "loss": 0.0061, "step": 65674 }, { "epoch": 1.27, "learning_rate": 2.8762235635864892e-05, "loss": 0.0, "step": 65676 }, { "epoch": 1.27, "learning_rate": 2.8761588675607658e-05, "loss": 0.0, "step": 65678 }, { "epoch": 1.27, "learning_rate": 2.8760941715350427e-05, "loss": 0.0, "step": 65680 }, { "epoch": 1.27, "learning_rate": 2.87602947550932e-05, "loss": 0.0, "step": 65682 }, { "epoch": 1.27, "learning_rate": 2.8759647794835965e-05, "loss": 0.0, "step": 65684 }, { "epoch": 1.27, "learning_rate": 2.875900083457873e-05, "loss": 0.0, "step": 65686 }, { "epoch": 1.27, "learning_rate": 2.8758353874321503e-05, "loss": 0.0003, "step": 65688 }, { "epoch": 1.27, "learning_rate": 2.875770691406427e-05, "loss": 0.0, "step": 65690 }, { "epoch": 1.28, "learning_rate": 2.875705995380704e-05, "loss": 0.0, "step": 65692 }, { "epoch": 1.28, "learning_rate": 2.8756412993549807e-05, "loss": 0.0, "step": 65694 }, { "epoch": 1.28, "learning_rate": 2.8755766033292576e-05, "loss": 0.0107, "step": 65696 }, { "epoch": 1.28, "learning_rate": 2.875511907303535e-05, "loss": 0.0001, "step": 65698 }, { "epoch": 1.28, "learning_rate": 2.8754472112778114e-05, "loss": 0.0, "step": 65700 }, { "epoch": 1.28, "learning_rate": 2.875382515252088e-05, "loss": 0.0, "step": 65702 }, { "epoch": 1.28, "learning_rate": 2.8753178192263652e-05, "loss": 0.0, "step": 65704 }, { "epoch": 1.28, "learning_rate": 2.8752531232006418e-05, "loss": 0.0, "step": 65706 }, { "epoch": 1.28, "learning_rate": 2.8751884271749184e-05, "loss": 0.0014, "step": 65708 }, { "epoch": 1.28, "learning_rate": 2.8751237311491956e-05, "loss": 0.0, "step": 65710 }, { "epoch": 1.28, "learning_rate": 2.8750590351234725e-05, "loss": 0.0, "step": 65712 }, { "epoch": 1.28, "learning_rate": 2.8749943390977498e-05, "loss": 0.0, "step": 65714 }, { "epoch": 1.28, "learning_rate": 2.8749296430720264e-05, "loss": 0.0, "step": 65716 }, { "epoch": 1.28, "learning_rate": 2.874864947046303e-05, "loss": 0.0, "step": 65718 }, { "epoch": 1.28, "learning_rate": 2.8748002510205802e-05, "loss": 0.0, "step": 65720 }, { "epoch": 1.28, "learning_rate": 2.8747355549948567e-05, "loss": 0.0, "step": 65722 }, { "epoch": 1.28, "learning_rate": 2.8746708589691333e-05, "loss": 0.0, "step": 65724 }, { "epoch": 1.28, "learning_rate": 2.8746061629434106e-05, "loss": 0.0, "step": 65726 }, { "epoch": 1.28, "learning_rate": 2.8745414669176875e-05, "loss": 0.0, "step": 65728 }, { "epoch": 1.28, "learning_rate": 2.874476770891964e-05, "loss": 0.0, "step": 65730 }, { "epoch": 1.28, "learning_rate": 2.8744120748662413e-05, "loss": 0.0, "step": 65732 }, { "epoch": 1.28, "learning_rate": 2.874347378840518e-05, "loss": 0.0001, "step": 65734 }, { "epoch": 1.28, "learning_rate": 2.874282682814795e-05, "loss": 0.0, "step": 65736 }, { "epoch": 1.28, "learning_rate": 2.8742179867890717e-05, "loss": 0.0, "step": 65738 }, { "epoch": 1.28, "learning_rate": 2.8741532907633482e-05, "loss": 0.0, "step": 65740 }, { "epoch": 1.28, "learning_rate": 2.8740885947376255e-05, "loss": 0.0, "step": 65742 }, { "epoch": 1.28, "learning_rate": 2.8740238987119024e-05, "loss": 0.0001, "step": 65744 }, { "epoch": 1.28, "learning_rate": 2.873959202686179e-05, "loss": 0.0, "step": 65746 }, { "epoch": 1.28, "learning_rate": 2.8738945066604562e-05, "loss": 0.0, "step": 65748 }, { "epoch": 1.28, "learning_rate": 2.8738298106347328e-05, "loss": 0.0, "step": 65750 }, { "epoch": 1.28, "learning_rate": 2.8737651146090094e-05, "loss": 0.0, "step": 65752 }, { "epoch": 1.28, "learning_rate": 2.8737004185832866e-05, "loss": 0.0, "step": 65754 }, { "epoch": 1.28, "learning_rate": 2.8736357225575632e-05, "loss": 0.0001, "step": 65756 }, { "epoch": 1.28, "learning_rate": 2.8735710265318404e-05, "loss": 0.0003, "step": 65758 }, { "epoch": 1.28, "learning_rate": 2.8735063305061173e-05, "loss": 0.0, "step": 65760 }, { "epoch": 1.28, "learning_rate": 2.873441634480394e-05, "loss": 0.0, "step": 65762 }, { "epoch": 1.28, "learning_rate": 2.873376938454671e-05, "loss": 0.0, "step": 65764 }, { "epoch": 1.28, "learning_rate": 2.8733122424289477e-05, "loss": 0.0, "step": 65766 }, { "epoch": 1.28, "learning_rate": 2.8732475464032243e-05, "loss": 0.0, "step": 65768 }, { "epoch": 1.28, "learning_rate": 2.8731828503775015e-05, "loss": 0.0049, "step": 65770 }, { "epoch": 1.28, "learning_rate": 2.873118154351778e-05, "loss": 0.0, "step": 65772 }, { "epoch": 1.28, "learning_rate": 2.8730534583260553e-05, "loss": 0.0, "step": 65774 }, { "epoch": 1.28, "learning_rate": 2.8729887623003323e-05, "loss": 0.0, "step": 65776 }, { "epoch": 1.28, "learning_rate": 2.8729240662746088e-05, "loss": 0.0, "step": 65778 }, { "epoch": 1.28, "learning_rate": 2.872859370248886e-05, "loss": 0.0, "step": 65780 }, { "epoch": 1.28, "learning_rate": 2.8727946742231626e-05, "loss": 0.0, "step": 65782 }, { "epoch": 1.28, "learning_rate": 2.8727299781974392e-05, "loss": 0.0, "step": 65784 }, { "epoch": 1.28, "learning_rate": 2.8726652821717165e-05, "loss": 0.0, "step": 65786 }, { "epoch": 1.28, "learning_rate": 2.872600586145993e-05, "loss": 0.0, "step": 65788 }, { "epoch": 1.28, "learning_rate": 2.87253589012027e-05, "loss": 0.0, "step": 65790 }, { "epoch": 1.28, "learning_rate": 2.872471194094547e-05, "loss": 0.0, "step": 65792 }, { "epoch": 1.28, "learning_rate": 2.8724064980688238e-05, "loss": 0.0, "step": 65794 }, { "epoch": 1.28, "learning_rate": 2.872341802043101e-05, "loss": 0.0015, "step": 65796 }, { "epoch": 1.28, "learning_rate": 2.8722771060173776e-05, "loss": 0.0001, "step": 65798 }, { "epoch": 1.28, "learning_rate": 2.872212409991654e-05, "loss": 0.0, "step": 65800 }, { "epoch": 1.28, "learning_rate": 2.8721477139659314e-05, "loss": 0.0001, "step": 65802 }, { "epoch": 1.28, "learning_rate": 2.872083017940208e-05, "loss": 0.0, "step": 65804 }, { "epoch": 1.28, "learning_rate": 2.872018321914485e-05, "loss": 0.0031, "step": 65806 }, { "epoch": 1.28, "learning_rate": 2.8719536258887618e-05, "loss": 0.0, "step": 65808 }, { "epoch": 1.28, "learning_rate": 2.8718889298630387e-05, "loss": 0.0, "step": 65810 }, { "epoch": 1.28, "learning_rate": 2.8718242338373153e-05, "loss": 0.0001, "step": 65812 }, { "epoch": 1.28, "learning_rate": 2.8717595378115925e-05, "loss": 0.0, "step": 65814 }, { "epoch": 1.28, "learning_rate": 2.871694841785869e-05, "loss": 0.0, "step": 65816 }, { "epoch": 1.28, "learning_rate": 2.8716301457601463e-05, "loss": 0.0, "step": 65818 }, { "epoch": 1.28, "learning_rate": 2.871565449734423e-05, "loss": 0.0, "step": 65820 }, { "epoch": 1.28, "learning_rate": 2.8715007537086998e-05, "loss": 0.0, "step": 65822 }, { "epoch": 1.28, "learning_rate": 2.8714360576829767e-05, "loss": 0.0, "step": 65824 }, { "epoch": 1.28, "learning_rate": 2.8713713616572536e-05, "loss": 0.0043, "step": 65826 }, { "epoch": 1.28, "learning_rate": 2.8713066656315302e-05, "loss": 0.0, "step": 65828 }, { "epoch": 1.28, "learning_rate": 2.8712419696058074e-05, "loss": 0.0, "step": 65830 }, { "epoch": 1.28, "learning_rate": 2.871177273580084e-05, "loss": 0.0, "step": 65832 }, { "epoch": 1.28, "learning_rate": 2.8711125775543612e-05, "loss": 0.0, "step": 65834 }, { "epoch": 1.28, "learning_rate": 2.8710478815286378e-05, "loss": 0.0001, "step": 65836 }, { "epoch": 1.28, "learning_rate": 2.8710155335157763e-05, "loss": 0.0194, "step": 65838 }, { "epoch": 1.28, "learning_rate": 2.870950837490053e-05, "loss": 0.0, "step": 65840 }, { "epoch": 1.28, "learning_rate": 2.87088614146433e-05, "loss": 0.0001, "step": 65842 }, { "epoch": 1.28, "learning_rate": 2.8708214454386067e-05, "loss": 0.0, "step": 65844 }, { "epoch": 1.28, "learning_rate": 2.870756749412884e-05, "loss": 0.0, "step": 65846 }, { "epoch": 1.28, "learning_rate": 2.8706920533871605e-05, "loss": 0.0004, "step": 65848 }, { "epoch": 1.28, "learning_rate": 2.8706273573614374e-05, "loss": 0.0, "step": 65850 }, { "epoch": 1.28, "learning_rate": 2.8705626613357146e-05, "loss": 0.0022, "step": 65852 }, { "epoch": 1.28, "learning_rate": 2.8704979653099912e-05, "loss": 0.0001, "step": 65854 }, { "epoch": 1.28, "learning_rate": 2.8704332692842678e-05, "loss": 0.0, "step": 65856 }, { "epoch": 1.28, "learning_rate": 2.870368573258545e-05, "loss": 0.0004, "step": 65858 }, { "epoch": 1.28, "learning_rate": 2.8703038772328216e-05, "loss": 0.0, "step": 65860 }, { "epoch": 1.28, "learning_rate": 2.8702391812070985e-05, "loss": 0.0017, "step": 65862 }, { "epoch": 1.28, "learning_rate": 2.8701744851813754e-05, "loss": 0.0, "step": 65864 }, { "epoch": 1.28, "learning_rate": 2.8701097891556523e-05, "loss": 0.0, "step": 65866 }, { "epoch": 1.28, "learning_rate": 2.8700450931299296e-05, "loss": 0.0017, "step": 65868 }, { "epoch": 1.28, "learning_rate": 2.869980397104206e-05, "loss": 0.0, "step": 65870 }, { "epoch": 1.28, "learning_rate": 2.8699157010784827e-05, "loss": 0.0, "step": 65872 }, { "epoch": 1.28, "learning_rate": 2.86985100505276e-05, "loss": 0.0005, "step": 65874 }, { "epoch": 1.28, "learning_rate": 2.8697863090270365e-05, "loss": 0.0, "step": 65876 }, { "epoch": 1.28, "learning_rate": 2.8697216130013134e-05, "loss": 0.0, "step": 65878 }, { "epoch": 1.28, "learning_rate": 2.8696569169755903e-05, "loss": 0.0, "step": 65880 }, { "epoch": 1.28, "learning_rate": 2.8695922209498672e-05, "loss": 0.0006, "step": 65882 }, { "epoch": 1.28, "learning_rate": 2.8695275249241438e-05, "loss": 0.0, "step": 65884 }, { "epoch": 1.28, "learning_rate": 2.869462828898421e-05, "loss": 0.0, "step": 65886 }, { "epoch": 1.28, "learning_rate": 2.8693981328726976e-05, "loss": 0.0, "step": 65888 }, { "epoch": 1.28, "learning_rate": 2.869333436846975e-05, "loss": 0.0, "step": 65890 }, { "epoch": 1.28, "learning_rate": 2.8692687408212514e-05, "loss": 0.0, "step": 65892 }, { "epoch": 1.28, "learning_rate": 2.869204044795528e-05, "loss": 0.0, "step": 65894 }, { "epoch": 1.28, "learning_rate": 2.8691393487698053e-05, "loss": 0.0, "step": 65896 }, { "epoch": 1.28, "learning_rate": 2.869074652744082e-05, "loss": 0.0, "step": 65898 }, { "epoch": 1.28, "learning_rate": 2.8690099567183587e-05, "loss": 0.0, "step": 65900 }, { "epoch": 1.28, "learning_rate": 2.868945260692636e-05, "loss": 0.0, "step": 65902 }, { "epoch": 1.28, "learning_rate": 2.8688805646669126e-05, "loss": 0.0, "step": 65904 }, { "epoch": 1.28, "learning_rate": 2.8688158686411898e-05, "loss": 0.0, "step": 65906 }, { "epoch": 1.28, "learning_rate": 2.8687511726154664e-05, "loss": 0.0, "step": 65908 }, { "epoch": 1.28, "learning_rate": 2.868686476589743e-05, "loss": 0.0, "step": 65910 }, { "epoch": 1.28, "learning_rate": 2.8686217805640202e-05, "loss": 0.0, "step": 65912 }, { "epoch": 1.28, "learning_rate": 2.868557084538297e-05, "loss": 0.0, "step": 65914 }, { "epoch": 1.28, "learning_rate": 2.8684923885125737e-05, "loss": 0.0013, "step": 65916 }, { "epoch": 1.28, "learning_rate": 2.868427692486851e-05, "loss": 0.0, "step": 65918 }, { "epoch": 1.28, "learning_rate": 2.8683629964611275e-05, "loss": 0.0, "step": 65920 }, { "epoch": 1.28, "learning_rate": 2.868298300435404e-05, "loss": 0.0, "step": 65922 }, { "epoch": 1.28, "learning_rate": 2.8682336044096813e-05, "loss": 0.0008, "step": 65924 }, { "epoch": 1.28, "learning_rate": 2.868168908383958e-05, "loss": 0.0, "step": 65926 }, { "epoch": 1.28, "learning_rate": 2.868104212358235e-05, "loss": 0.0002, "step": 65928 }, { "epoch": 1.28, "learning_rate": 2.868039516332512e-05, "loss": 0.0, "step": 65930 }, { "epoch": 1.28, "learning_rate": 2.8679748203067886e-05, "loss": 0.0, "step": 65932 }, { "epoch": 1.28, "learning_rate": 2.867910124281066e-05, "loss": 0.0, "step": 65934 }, { "epoch": 1.28, "learning_rate": 2.8678454282553424e-05, "loss": 0.0, "step": 65936 }, { "epoch": 1.28, "learning_rate": 2.867780732229619e-05, "loss": 0.0, "step": 65938 }, { "epoch": 1.28, "learning_rate": 2.8677160362038962e-05, "loss": 0.0041, "step": 65940 }, { "epoch": 1.28, "learning_rate": 2.8676513401781728e-05, "loss": 0.0, "step": 65942 }, { "epoch": 1.28, "learning_rate": 2.8675866441524497e-05, "loss": 0.0, "step": 65944 }, { "epoch": 1.28, "learning_rate": 2.867521948126727e-05, "loss": 0.0, "step": 65946 }, { "epoch": 1.28, "learning_rate": 2.8674572521010035e-05, "loss": 0.0035, "step": 65948 }, { "epoch": 1.28, "learning_rate": 2.8673925560752808e-05, "loss": 0.0, "step": 65950 }, { "epoch": 1.28, "learning_rate": 2.8673278600495573e-05, "loss": 0.0, "step": 65952 }, { "epoch": 1.28, "learning_rate": 2.867263164023834e-05, "loss": 0.002, "step": 65954 }, { "epoch": 1.28, "learning_rate": 2.867198467998111e-05, "loss": 0.0018, "step": 65956 }, { "epoch": 1.28, "learning_rate": 2.8671337719723877e-05, "loss": 0.0, "step": 65958 }, { "epoch": 1.28, "learning_rate": 2.8670690759466646e-05, "loss": 0.0, "step": 65960 }, { "epoch": 1.28, "learning_rate": 2.867004379920942e-05, "loss": 0.0, "step": 65962 }, { "epoch": 1.28, "learning_rate": 2.8669396838952185e-05, "loss": 0.0, "step": 65964 }, { "epoch": 1.28, "learning_rate": 2.866874987869495e-05, "loss": 0.0, "step": 65966 }, { "epoch": 1.28, "learning_rate": 2.8668102918437723e-05, "loss": 0.0, "step": 65968 }, { "epoch": 1.28, "learning_rate": 2.866745595818049e-05, "loss": 0.0, "step": 65970 }, { "epoch": 1.28, "learning_rate": 2.866680899792326e-05, "loss": 0.0, "step": 65972 }, { "epoch": 1.28, "learning_rate": 2.8666162037666027e-05, "loss": 0.0003, "step": 65974 }, { "epoch": 1.28, "learning_rate": 2.8665515077408796e-05, "loss": 0.0, "step": 65976 }, { "epoch": 1.28, "learning_rate": 2.8664868117151568e-05, "loss": 0.0, "step": 65978 }, { "epoch": 1.28, "learning_rate": 2.8664221156894334e-05, "loss": 0.0001, "step": 65980 }, { "epoch": 1.28, "learning_rate": 2.86635741966371e-05, "loss": 0.0, "step": 65982 }, { "epoch": 1.28, "learning_rate": 2.8662927236379872e-05, "loss": 0.0, "step": 65984 }, { "epoch": 1.28, "learning_rate": 2.8662280276122638e-05, "loss": 0.0, "step": 65986 }, { "epoch": 1.28, "learning_rate": 2.866163331586541e-05, "loss": 0.0004, "step": 65988 }, { "epoch": 1.28, "learning_rate": 2.8660986355608176e-05, "loss": 0.0001, "step": 65990 }, { "epoch": 1.28, "learning_rate": 2.8660339395350945e-05, "loss": 0.0002, "step": 65992 }, { "epoch": 1.28, "learning_rate": 2.8659692435093714e-05, "loss": 0.0, "step": 65994 }, { "epoch": 1.28, "learning_rate": 2.8659045474836483e-05, "loss": 0.0, "step": 65996 }, { "epoch": 1.28, "learning_rate": 2.865839851457925e-05, "loss": 0.0, "step": 65998 }, { "epoch": 1.28, "learning_rate": 2.865775155432202e-05, "loss": 0.0, "step": 66000 }, { "epoch": 1.28, "learning_rate": 2.8657104594064787e-05, "loss": 0.0, "step": 66002 }, { "epoch": 1.28, "learning_rate": 2.8656457633807553e-05, "loss": 0.0262, "step": 66004 }, { "epoch": 1.28, "learning_rate": 2.8655810673550325e-05, "loss": 0.0174, "step": 66006 }, { "epoch": 1.28, "learning_rate": 2.8655163713293094e-05, "loss": 0.0072, "step": 66008 }, { "epoch": 1.28, "learning_rate": 2.8654516753035863e-05, "loss": 0.0, "step": 66010 }, { "epoch": 1.28, "learning_rate": 2.8653869792778632e-05, "loss": 0.0, "step": 66012 }, { "epoch": 1.28, "learning_rate": 2.8653222832521398e-05, "loss": 0.0001, "step": 66014 }, { "epoch": 1.28, "learning_rate": 2.865257587226417e-05, "loss": 0.0, "step": 66016 }, { "epoch": 1.28, "learning_rate": 2.8651928912006936e-05, "loss": 0.0, "step": 66018 }, { "epoch": 1.28, "learning_rate": 2.8651281951749702e-05, "loss": 0.0, "step": 66020 }, { "epoch": 1.28, "learning_rate": 2.8650634991492474e-05, "loss": 0.0001, "step": 66022 }, { "epoch": 1.28, "learning_rate": 2.8649988031235244e-05, "loss": 0.0, "step": 66024 }, { "epoch": 1.28, "learning_rate": 2.864934107097801e-05, "loss": 0.0, "step": 66026 }, { "epoch": 1.28, "learning_rate": 2.864869411072078e-05, "loss": 0.0, "step": 66028 }, { "epoch": 1.28, "learning_rate": 2.8648047150463547e-05, "loss": 0.0, "step": 66030 }, { "epoch": 1.28, "learning_rate": 2.864740019020632e-05, "loss": 0.0, "step": 66032 }, { "epoch": 1.28, "learning_rate": 2.8646753229949086e-05, "loss": 0.0, "step": 66034 }, { "epoch": 1.28, "learning_rate": 2.864610626969185e-05, "loss": 0.0, "step": 66036 }, { "epoch": 1.28, "learning_rate": 2.8645459309434624e-05, "loss": 0.0003, "step": 66038 }, { "epoch": 1.28, "learning_rate": 2.8644812349177393e-05, "loss": 0.0, "step": 66040 }, { "epoch": 1.28, "learning_rate": 2.864416538892016e-05, "loss": 0.0, "step": 66042 }, { "epoch": 1.28, "learning_rate": 2.864351842866293e-05, "loss": 0.0, "step": 66044 }, { "epoch": 1.28, "learning_rate": 2.8642871468405697e-05, "loss": 0.0, "step": 66046 }, { "epoch": 1.28, "learning_rate": 2.864222450814847e-05, "loss": 0.0, "step": 66048 }, { "epoch": 1.28, "learning_rate": 2.8641577547891235e-05, "loss": 0.0001, "step": 66050 }, { "epoch": 1.28, "learning_rate": 2.8640930587634e-05, "loss": 0.0001, "step": 66052 }, { "epoch": 1.28, "learning_rate": 2.8640283627376773e-05, "loss": 0.0, "step": 66054 }, { "epoch": 1.28, "learning_rate": 2.8639636667119542e-05, "loss": 0.0, "step": 66056 }, { "epoch": 1.28, "learning_rate": 2.8638989706862308e-05, "loss": 0.0, "step": 66058 }, { "epoch": 1.28, "learning_rate": 2.863834274660508e-05, "loss": 0.0, "step": 66060 }, { "epoch": 1.28, "learning_rate": 2.8637695786347846e-05, "loss": 0.0, "step": 66062 }, { "epoch": 1.28, "learning_rate": 2.863704882609061e-05, "loss": 0.0008, "step": 66064 }, { "epoch": 1.28, "learning_rate": 2.8636401865833384e-05, "loss": 0.0, "step": 66066 }, { "epoch": 1.28, "learning_rate": 2.863575490557615e-05, "loss": 0.0, "step": 66068 }, { "epoch": 1.28, "learning_rate": 2.8635107945318922e-05, "loss": 0.0001, "step": 66070 }, { "epoch": 1.28, "learning_rate": 2.8634460985061688e-05, "loss": 0.0, "step": 66072 }, { "epoch": 1.28, "learning_rate": 2.8633814024804457e-05, "loss": 0.0, "step": 66074 }, { "epoch": 1.28, "learning_rate": 2.863316706454723e-05, "loss": 0.0, "step": 66076 }, { "epoch": 1.28, "learning_rate": 2.8632520104289995e-05, "loss": 0.0003, "step": 66078 }, { "epoch": 1.28, "learning_rate": 2.863187314403276e-05, "loss": 0.0, "step": 66080 }, { "epoch": 1.28, "learning_rate": 2.8631226183775533e-05, "loss": 0.0012, "step": 66082 }, { "epoch": 1.28, "learning_rate": 2.86305792235183e-05, "loss": 0.0, "step": 66084 }, { "epoch": 1.28, "learning_rate": 2.8629932263261068e-05, "loss": 0.0, "step": 66086 }, { "epoch": 1.28, "learning_rate": 2.8629285303003837e-05, "loss": 0.0, "step": 66088 }, { "epoch": 1.28, "learning_rate": 2.8628638342746606e-05, "loss": 0.0, "step": 66090 }, { "epoch": 1.28, "learning_rate": 2.862799138248938e-05, "loss": 0.0, "step": 66092 }, { "epoch": 1.28, "learning_rate": 2.8627344422232145e-05, "loss": 0.0, "step": 66094 }, { "epoch": 1.28, "learning_rate": 2.862669746197491e-05, "loss": 0.0, "step": 66096 }, { "epoch": 1.28, "learning_rate": 2.8626050501717683e-05, "loss": 0.0001, "step": 66098 }, { "epoch": 1.28, "learning_rate": 2.862540354146045e-05, "loss": 0.0008, "step": 66100 }, { "epoch": 1.28, "learning_rate": 2.8624756581203217e-05, "loss": 0.0, "step": 66102 }, { "epoch": 1.28, "learning_rate": 2.8624109620945987e-05, "loss": 0.0, "step": 66104 }, { "epoch": 1.28, "learning_rate": 2.8623462660688756e-05, "loss": 0.0, "step": 66106 }, { "epoch": 1.28, "learning_rate": 2.862281570043152e-05, "loss": 0.0, "step": 66108 }, { "epoch": 1.28, "learning_rate": 2.8622168740174294e-05, "loss": 0.0, "step": 66110 }, { "epoch": 1.28, "learning_rate": 2.862152177991706e-05, "loss": 0.0005, "step": 66112 }, { "epoch": 1.28, "learning_rate": 2.8620874819659832e-05, "loss": 0.0001, "step": 66114 }, { "epoch": 1.28, "learning_rate": 2.8620227859402598e-05, "loss": 0.0, "step": 66116 }, { "epoch": 1.28, "learning_rate": 2.8619580899145367e-05, "loss": 0.0, "step": 66118 }, { "epoch": 1.28, "learning_rate": 2.8618933938888136e-05, "loss": 0.0, "step": 66120 }, { "epoch": 1.28, "learning_rate": 2.8618286978630905e-05, "loss": 0.0003, "step": 66122 }, { "epoch": 1.28, "learning_rate": 2.861764001837367e-05, "loss": 0.0, "step": 66124 }, { "epoch": 1.28, "learning_rate": 2.8616993058116443e-05, "loss": 0.0, "step": 66126 }, { "epoch": 1.28, "learning_rate": 2.861634609785921e-05, "loss": 0.0, "step": 66128 }, { "epoch": 1.28, "learning_rate": 2.861569913760198e-05, "loss": 0.0, "step": 66130 }, { "epoch": 1.28, "learning_rate": 2.8615052177344747e-05, "loss": 0.0, "step": 66132 }, { "epoch": 1.28, "learning_rate": 2.8614405217087513e-05, "loss": 0.0, "step": 66134 }, { "epoch": 1.28, "learning_rate": 2.8613758256830285e-05, "loss": 0.0001, "step": 66136 }, { "epoch": 1.28, "learning_rate": 2.8613111296573054e-05, "loss": 0.0, "step": 66138 }, { "epoch": 1.28, "learning_rate": 2.861246433631582e-05, "loss": 0.0051, "step": 66140 }, { "epoch": 1.28, "learning_rate": 2.8611817376058592e-05, "loss": 0.0, "step": 66142 }, { "epoch": 1.28, "learning_rate": 2.8611170415801358e-05, "loss": 0.0, "step": 66144 }, { "epoch": 1.28, "learning_rate": 2.8610523455544124e-05, "loss": 0.0004, "step": 66146 }, { "epoch": 1.28, "learning_rate": 2.8609876495286896e-05, "loss": 0.0, "step": 66148 }, { "epoch": 1.28, "learning_rate": 2.8609229535029662e-05, "loss": 0.0001, "step": 66150 }, { "epoch": 1.28, "learning_rate": 2.8608582574772434e-05, "loss": 0.0, "step": 66152 }, { "epoch": 1.28, "learning_rate": 2.8607935614515204e-05, "loss": 0.0019, "step": 66154 }, { "epoch": 1.28, "learning_rate": 2.860728865425797e-05, "loss": 0.0, "step": 66156 }, { "epoch": 1.28, "learning_rate": 2.860664169400074e-05, "loss": 0.0, "step": 66158 }, { "epoch": 1.28, "learning_rate": 2.8605994733743507e-05, "loss": 0.0, "step": 66160 }, { "epoch": 1.28, "learning_rate": 2.8605347773486273e-05, "loss": 0.0, "step": 66162 }, { "epoch": 1.28, "learning_rate": 2.8604700813229046e-05, "loss": 0.0, "step": 66164 }, { "epoch": 1.28, "learning_rate": 2.860405385297181e-05, "loss": 0.0, "step": 66166 }, { "epoch": 1.28, "learning_rate": 2.860340689271458e-05, "loss": 0.0, "step": 66168 }, { "epoch": 1.28, "learning_rate": 2.8602759932457353e-05, "loss": 0.0, "step": 66170 }, { "epoch": 1.28, "learning_rate": 2.860211297220012e-05, "loss": 0.0, "step": 66172 }, { "epoch": 1.28, "learning_rate": 2.860146601194289e-05, "loss": 0.0001, "step": 66174 }, { "epoch": 1.28, "learning_rate": 2.8600819051685657e-05, "loss": 0.0, "step": 66176 }, { "epoch": 1.28, "learning_rate": 2.8600172091428422e-05, "loss": 0.0121, "step": 66178 }, { "epoch": 1.28, "learning_rate": 2.8599525131171195e-05, "loss": 0.0, "step": 66180 }, { "epoch": 1.28, "learning_rate": 2.859887817091396e-05, "loss": 0.0, "step": 66182 }, { "epoch": 1.28, "learning_rate": 2.859823121065673e-05, "loss": 0.0032, "step": 66184 }, { "epoch": 1.28, "learning_rate": 2.8597584250399502e-05, "loss": 0.0, "step": 66186 }, { "epoch": 1.28, "learning_rate": 2.8596937290142268e-05, "loss": 0.0, "step": 66188 }, { "epoch": 1.28, "learning_rate": 2.8596290329885033e-05, "loss": 0.0, "step": 66190 }, { "epoch": 1.28, "learning_rate": 2.8595643369627806e-05, "loss": 0.0, "step": 66192 }, { "epoch": 1.28, "learning_rate": 2.859499640937057e-05, "loss": 0.0, "step": 66194 }, { "epoch": 1.28, "learning_rate": 2.8594349449113344e-05, "loss": 0.0, "step": 66196 }, { "epoch": 1.28, "learning_rate": 2.859370248885611e-05, "loss": 0.0175, "step": 66198 }, { "epoch": 1.28, "learning_rate": 2.859305552859888e-05, "loss": 0.0, "step": 66200 }, { "epoch": 1.28, "learning_rate": 2.859240856834165e-05, "loss": 0.0, "step": 66202 }, { "epoch": 1.28, "learning_rate": 2.8591761608084417e-05, "loss": 0.0, "step": 66204 }, { "epoch": 1.28, "learning_rate": 2.8591114647827183e-05, "loss": 0.0, "step": 66206 }, { "epoch": 1.29, "learning_rate": 2.8590467687569955e-05, "loss": 0.0, "step": 66208 }, { "epoch": 1.29, "learning_rate": 2.858982072731272e-05, "loss": 0.0001, "step": 66210 }, { "epoch": 1.29, "learning_rate": 2.8589173767055493e-05, "loss": 0.0, "step": 66212 }, { "epoch": 1.29, "learning_rate": 2.858852680679826e-05, "loss": 0.0, "step": 66214 }, { "epoch": 1.29, "learning_rate": 2.8587879846541028e-05, "loss": 0.0, "step": 66216 }, { "epoch": 1.29, "learning_rate": 2.85872328862838e-05, "loss": 0.0014, "step": 66218 }, { "epoch": 1.29, "learning_rate": 2.8586585926026566e-05, "loss": 0.0, "step": 66220 }, { "epoch": 1.29, "learning_rate": 2.8585938965769332e-05, "loss": 0.0, "step": 66222 }, { "epoch": 1.29, "learning_rate": 2.8585292005512105e-05, "loss": 0.0, "step": 66224 }, { "epoch": 1.29, "learning_rate": 2.858464504525487e-05, "loss": 0.0, "step": 66226 }, { "epoch": 1.29, "learning_rate": 2.8583998084997636e-05, "loss": 0.0, "step": 66228 }, { "epoch": 1.29, "learning_rate": 2.858335112474041e-05, "loss": 0.0001, "step": 66230 }, { "epoch": 1.29, "learning_rate": 2.8582704164483177e-05, "loss": 0.0, "step": 66232 }, { "epoch": 1.29, "learning_rate": 2.8582057204225947e-05, "loss": 0.0, "step": 66234 }, { "epoch": 1.29, "learning_rate": 2.8581410243968716e-05, "loss": 0.0, "step": 66236 }, { "epoch": 1.29, "learning_rate": 2.858076328371148e-05, "loss": 0.0, "step": 66238 }, { "epoch": 1.29, "learning_rate": 2.8580116323454254e-05, "loss": 0.0, "step": 66240 }, { "epoch": 1.29, "learning_rate": 2.857946936319702e-05, "loss": 0.0, "step": 66242 }, { "epoch": 1.29, "learning_rate": 2.8578822402939785e-05, "loss": 0.0001, "step": 66244 }, { "epoch": 1.29, "learning_rate": 2.8578175442682558e-05, "loss": 0.0, "step": 66246 }, { "epoch": 1.29, "learning_rate": 2.8577528482425327e-05, "loss": 0.0, "step": 66248 }, { "epoch": 1.29, "learning_rate": 2.8576881522168092e-05, "loss": 0.0, "step": 66250 }, { "epoch": 1.29, "learning_rate": 2.8576234561910865e-05, "loss": 0.0004, "step": 66252 }, { "epoch": 1.29, "learning_rate": 2.857558760165363e-05, "loss": 0.0, "step": 66254 }, { "epoch": 1.29, "learning_rate": 2.8574940641396403e-05, "loss": 0.0172, "step": 66256 }, { "epoch": 1.29, "learning_rate": 2.857429368113917e-05, "loss": 0.0, "step": 66258 }, { "epoch": 1.29, "learning_rate": 2.8573646720881934e-05, "loss": 0.0, "step": 66260 }, { "epoch": 1.29, "learning_rate": 2.8572999760624707e-05, "loss": 0.0, "step": 66262 }, { "epoch": 1.29, "learning_rate": 2.8572352800367476e-05, "loss": 0.0, "step": 66264 }, { "epoch": 1.29, "learning_rate": 2.8571705840110242e-05, "loss": 0.0, "step": 66266 }, { "epoch": 1.29, "learning_rate": 2.8571058879853014e-05, "loss": 0.0001, "step": 66268 }, { "epoch": 1.29, "learning_rate": 2.857041191959578e-05, "loss": 0.0, "step": 66270 }, { "epoch": 1.29, "learning_rate": 2.8569764959338552e-05, "loss": 0.0012, "step": 66272 }, { "epoch": 1.29, "learning_rate": 2.8569117999081318e-05, "loss": 0.0, "step": 66274 }, { "epoch": 1.29, "learning_rate": 2.8568471038824084e-05, "loss": 0.0, "step": 66276 }, { "epoch": 1.29, "learning_rate": 2.8567824078566856e-05, "loss": 0.0, "step": 66278 }, { "epoch": 1.29, "learning_rate": 2.8567177118309625e-05, "loss": 0.0003, "step": 66280 }, { "epoch": 1.29, "learning_rate": 2.856653015805239e-05, "loss": 0.0, "step": 66282 }, { "epoch": 1.29, "learning_rate": 2.8565883197795163e-05, "loss": 0.0, "step": 66284 }, { "epoch": 1.29, "learning_rate": 2.856523623753793e-05, "loss": 0.0, "step": 66286 }, { "epoch": 1.29, "learning_rate": 2.8564589277280695e-05, "loss": 0.0, "step": 66288 }, { "epoch": 1.29, "learning_rate": 2.8563942317023467e-05, "loss": 0.0, "step": 66290 }, { "epoch": 1.29, "learning_rate": 2.8563295356766233e-05, "loss": 0.0, "step": 66292 }, { "epoch": 1.29, "learning_rate": 2.8562648396509006e-05, "loss": 0.0012, "step": 66294 }, { "epoch": 1.29, "learning_rate": 2.8562001436251775e-05, "loss": 0.0, "step": 66296 }, { "epoch": 1.29, "learning_rate": 2.856135447599454e-05, "loss": 0.0, "step": 66298 }, { "epoch": 1.29, "learning_rate": 2.8560707515737313e-05, "loss": 0.0021, "step": 66300 }, { "epoch": 1.29, "learning_rate": 2.856006055548008e-05, "loss": 0.0, "step": 66302 }, { "epoch": 1.29, "learning_rate": 2.8559413595222844e-05, "loss": 0.0, "step": 66304 }, { "epoch": 1.29, "learning_rate": 2.8558766634965617e-05, "loss": 0.0, "step": 66306 }, { "epoch": 1.29, "learning_rate": 2.8558119674708382e-05, "loss": 0.0, "step": 66308 }, { "epoch": 1.29, "learning_rate": 2.855747271445115e-05, "loss": 0.0002, "step": 66310 }, { "epoch": 1.29, "learning_rate": 2.855682575419392e-05, "loss": 0.0, "step": 66312 }, { "epoch": 1.29, "learning_rate": 2.855617879393669e-05, "loss": 0.0, "step": 66314 }, { "epoch": 1.29, "learning_rate": 2.8555531833679462e-05, "loss": 0.0, "step": 66316 }, { "epoch": 1.29, "learning_rate": 2.8554884873422228e-05, "loss": 0.0, "step": 66318 }, { "epoch": 1.29, "learning_rate": 2.8554237913164993e-05, "loss": 0.0, "step": 66320 }, { "epoch": 1.29, "learning_rate": 2.8553590952907766e-05, "loss": 0.0194, "step": 66322 }, { "epoch": 1.29, "learning_rate": 2.855294399265053e-05, "loss": 0.0, "step": 66324 }, { "epoch": 1.29, "learning_rate": 2.85522970323933e-05, "loss": 0.0016, "step": 66326 }, { "epoch": 1.29, "learning_rate": 2.855165007213607e-05, "loss": 0.0001, "step": 66328 }, { "epoch": 1.29, "learning_rate": 2.855100311187884e-05, "loss": 0.0, "step": 66330 }, { "epoch": 1.29, "learning_rate": 2.8550356151621605e-05, "loss": 0.0, "step": 66332 }, { "epoch": 1.29, "learning_rate": 2.8549709191364377e-05, "loss": 0.0, "step": 66334 }, { "epoch": 1.29, "learning_rate": 2.8549062231107143e-05, "loss": 0.0023, "step": 66336 }, { "epoch": 1.29, "learning_rate": 2.8548415270849915e-05, "loss": 0.0, "step": 66338 }, { "epoch": 1.29, "learning_rate": 2.854776831059268e-05, "loss": 0.0, "step": 66340 }, { "epoch": 1.29, "learning_rate": 2.854712135033545e-05, "loss": 0.0, "step": 66342 }, { "epoch": 1.29, "learning_rate": 2.854647439007822e-05, "loss": 0.001, "step": 66344 }, { "epoch": 1.29, "learning_rate": 2.8545827429820988e-05, "loss": 0.0001, "step": 66346 }, { "epoch": 1.29, "learning_rate": 2.8545180469563754e-05, "loss": 0.0033, "step": 66348 }, { "epoch": 1.29, "learning_rate": 2.8544533509306526e-05, "loss": 0.0, "step": 66350 }, { "epoch": 1.29, "learning_rate": 2.8543886549049292e-05, "loss": 0.0001, "step": 66352 }, { "epoch": 1.29, "learning_rate": 2.8543239588792065e-05, "loss": 0.0002, "step": 66354 }, { "epoch": 1.29, "learning_rate": 2.854259262853483e-05, "loss": 0.0, "step": 66356 }, { "epoch": 1.29, "learning_rate": 2.85419456682776e-05, "loss": 0.0, "step": 66358 }, { "epoch": 1.29, "learning_rate": 2.854129870802037e-05, "loss": 0.0, "step": 66360 }, { "epoch": 1.29, "learning_rate": 2.8540651747763137e-05, "loss": 0.0, "step": 66362 }, { "epoch": 1.29, "learning_rate": 2.8540004787505903e-05, "loss": 0.0, "step": 66364 }, { "epoch": 1.29, "learning_rate": 2.8539357827248676e-05, "loss": 0.0116, "step": 66366 }, { "epoch": 1.29, "learning_rate": 2.853871086699144e-05, "loss": 0.0, "step": 66368 }, { "epoch": 1.29, "learning_rate": 2.8538063906734207e-05, "loss": 0.0, "step": 66370 }, { "epoch": 1.29, "learning_rate": 2.853741694647698e-05, "loss": 0.0001, "step": 66372 }, { "epoch": 1.29, "learning_rate": 2.8536769986219745e-05, "loss": 0.0, "step": 66374 }, { "epoch": 1.29, "learning_rate": 2.8536123025962518e-05, "loss": 0.0001, "step": 66376 }, { "epoch": 1.29, "learning_rate": 2.8535476065705287e-05, "loss": 0.0, "step": 66378 }, { "epoch": 1.29, "learning_rate": 2.8534829105448052e-05, "loss": 0.012, "step": 66380 }, { "epoch": 1.29, "learning_rate": 2.8534182145190825e-05, "loss": 0.0, "step": 66382 }, { "epoch": 1.29, "learning_rate": 2.853353518493359e-05, "loss": 0.0, "step": 66384 }, { "epoch": 1.29, "learning_rate": 2.8532888224676356e-05, "loss": 0.0, "step": 66386 }, { "epoch": 1.29, "learning_rate": 2.853224126441913e-05, "loss": 0.0001, "step": 66388 }, { "epoch": 1.29, "learning_rate": 2.8531594304161894e-05, "loss": 0.0, "step": 66390 }, { "epoch": 1.29, "learning_rate": 2.8530947343904664e-05, "loss": 0.0, "step": 66392 }, { "epoch": 1.29, "learning_rate": 2.8530300383647436e-05, "loss": 0.0, "step": 66394 }, { "epoch": 1.29, "learning_rate": 2.8529653423390202e-05, "loss": 0.004, "step": 66396 }, { "epoch": 1.29, "learning_rate": 2.8529006463132974e-05, "loss": 0.0, "step": 66398 }, { "epoch": 1.29, "learning_rate": 2.852835950287574e-05, "loss": 0.0, "step": 66400 }, { "epoch": 1.29, "learning_rate": 2.8527712542618506e-05, "loss": 0.0, "step": 66402 }, { "epoch": 1.29, "learning_rate": 2.8527065582361278e-05, "loss": 0.0, "step": 66404 }, { "epoch": 1.29, "learning_rate": 2.8526418622104044e-05, "loss": 0.0, "step": 66406 }, { "epoch": 1.29, "learning_rate": 2.8525771661846813e-05, "loss": 0.0001, "step": 66408 }, { "epoch": 1.29, "learning_rate": 2.8525124701589585e-05, "loss": 0.0, "step": 66410 }, { "epoch": 1.29, "learning_rate": 2.852447774133235e-05, "loss": 0.0, "step": 66412 }, { "epoch": 1.29, "learning_rate": 2.8523830781075123e-05, "loss": 0.0, "step": 66414 }, { "epoch": 1.29, "learning_rate": 2.852318382081789e-05, "loss": 0.0028, "step": 66416 }, { "epoch": 1.29, "learning_rate": 2.8522536860560655e-05, "loss": 0.0, "step": 66418 }, { "epoch": 1.29, "learning_rate": 2.8521889900303427e-05, "loss": 0.0, "step": 66420 }, { "epoch": 1.29, "learning_rate": 2.8521242940046193e-05, "loss": 0.0016, "step": 66422 }, { "epoch": 1.29, "learning_rate": 2.8520595979788962e-05, "loss": 0.0, "step": 66424 }, { "epoch": 1.29, "learning_rate": 2.8519949019531735e-05, "loss": 0.0, "step": 66426 }, { "epoch": 1.29, "learning_rate": 2.85193020592745e-05, "loss": 0.0, "step": 66428 }, { "epoch": 1.29, "learning_rate": 2.8518655099017266e-05, "loss": 0.0006, "step": 66430 }, { "epoch": 1.29, "learning_rate": 2.851800813876004e-05, "loss": 0.0, "step": 66432 }, { "epoch": 1.29, "learning_rate": 2.8517361178502804e-05, "loss": 0.0, "step": 66434 }, { "epoch": 1.29, "learning_rate": 2.8516714218245577e-05, "loss": 0.0, "step": 66436 }, { "epoch": 1.29, "learning_rate": 2.8516067257988342e-05, "loss": 0.0, "step": 66438 }, { "epoch": 1.29, "learning_rate": 2.851542029773111e-05, "loss": 0.0, "step": 66440 }, { "epoch": 1.29, "learning_rate": 2.8514773337473884e-05, "loss": 0.0, "step": 66442 }, { "epoch": 1.29, "learning_rate": 2.851412637721665e-05, "loss": 0.0001, "step": 66444 }, { "epoch": 1.29, "learning_rate": 2.8513479416959415e-05, "loss": 0.0, "step": 66446 }, { "epoch": 1.29, "learning_rate": 2.8512832456702188e-05, "loss": 0.0, "step": 66448 }, { "epoch": 1.29, "learning_rate": 2.8512185496444953e-05, "loss": 0.0, "step": 66450 }, { "epoch": 1.29, "learning_rate": 2.851153853618772e-05, "loss": 0.0, "step": 66452 }, { "epoch": 1.29, "learning_rate": 2.851089157593049e-05, "loss": 0.0, "step": 66454 }, { "epoch": 1.29, "learning_rate": 2.851024461567326e-05, "loss": 0.0, "step": 66456 }, { "epoch": 1.29, "learning_rate": 2.8509597655416033e-05, "loss": 0.0, "step": 66458 }, { "epoch": 1.29, "learning_rate": 2.85089506951588e-05, "loss": 0.0, "step": 66460 }, { "epoch": 1.29, "learning_rate": 2.8508303734901565e-05, "loss": 0.0004, "step": 66462 }, { "epoch": 1.29, "learning_rate": 2.8507656774644337e-05, "loss": 0.0, "step": 66464 }, { "epoch": 1.29, "learning_rate": 2.8507009814387103e-05, "loss": 0.0, "step": 66466 }, { "epoch": 1.29, "learning_rate": 2.850636285412987e-05, "loss": 0.0059, "step": 66468 }, { "epoch": 1.29, "learning_rate": 2.850571589387264e-05, "loss": 0.0, "step": 66470 }, { "epoch": 1.29, "learning_rate": 2.850506893361541e-05, "loss": 0.0, "step": 66472 }, { "epoch": 1.29, "learning_rate": 2.8504421973358176e-05, "loss": 0.0001, "step": 66474 }, { "epoch": 1.29, "learning_rate": 2.8503775013100948e-05, "loss": 0.0, "step": 66476 }, { "epoch": 1.29, "learning_rate": 2.8503128052843714e-05, "loss": 0.0, "step": 66478 }, { "epoch": 1.29, "learning_rate": 2.8502481092586486e-05, "loss": 0.0, "step": 66480 }, { "epoch": 1.29, "learning_rate": 2.8501834132329252e-05, "loss": 0.0, "step": 66482 }, { "epoch": 1.29, "learning_rate": 2.8501187172072018e-05, "loss": 0.0, "step": 66484 }, { "epoch": 1.29, "learning_rate": 2.850054021181479e-05, "loss": 0.0, "step": 66486 }, { "epoch": 1.29, "learning_rate": 2.849989325155756e-05, "loss": 0.0, "step": 66488 }, { "epoch": 1.29, "learning_rate": 2.8499246291300325e-05, "loss": 0.0, "step": 66490 }, { "epoch": 1.29, "learning_rate": 2.8498599331043097e-05, "loss": 0.0, "step": 66492 }, { "epoch": 1.29, "learning_rate": 2.8497952370785863e-05, "loss": 0.0, "step": 66494 }, { "epoch": 1.29, "learning_rate": 2.8497305410528636e-05, "loss": 0.0, "step": 66496 }, { "epoch": 1.29, "learning_rate": 2.84966584502714e-05, "loss": 0.0038, "step": 66498 }, { "epoch": 1.29, "learning_rate": 2.8496011490014167e-05, "loss": 0.0, "step": 66500 }, { "epoch": 1.29, "learning_rate": 2.849536452975694e-05, "loss": 0.0, "step": 66502 }, { "epoch": 1.29, "learning_rate": 2.849471756949971e-05, "loss": 0.0001, "step": 66504 }, { "epoch": 1.29, "learning_rate": 2.8494070609242474e-05, "loss": 0.0, "step": 66506 }, { "epoch": 1.29, "learning_rate": 2.8493423648985247e-05, "loss": 0.0, "step": 66508 }, { "epoch": 1.29, "learning_rate": 2.8492776688728012e-05, "loss": 0.0002, "step": 66510 }, { "epoch": 1.29, "learning_rate": 2.8492129728470778e-05, "loss": 0.0, "step": 66512 }, { "epoch": 1.29, "learning_rate": 2.849148276821355e-05, "loss": 0.0, "step": 66514 }, { "epoch": 1.29, "learning_rate": 2.8490835807956316e-05, "loss": 0.0065, "step": 66516 }, { "epoch": 1.29, "learning_rate": 2.849018884769909e-05, "loss": 0.0, "step": 66518 }, { "epoch": 1.29, "learning_rate": 2.8489541887441858e-05, "loss": 0.0037, "step": 66520 }, { "epoch": 1.29, "learning_rate": 2.8488894927184624e-05, "loss": 0.0, "step": 66522 }, { "epoch": 1.29, "learning_rate": 2.8488247966927396e-05, "loss": 0.0, "step": 66524 }, { "epoch": 1.29, "learning_rate": 2.8487601006670162e-05, "loss": 0.0, "step": 66526 }, { "epoch": 1.29, "learning_rate": 2.8486954046412927e-05, "loss": 0.0, "step": 66528 }, { "epoch": 1.29, "learning_rate": 2.84863070861557e-05, "loss": 0.0039, "step": 66530 }, { "epoch": 1.29, "learning_rate": 2.8485660125898466e-05, "loss": 0.0034, "step": 66532 }, { "epoch": 1.29, "learning_rate": 2.8485013165641235e-05, "loss": 0.0, "step": 66534 }, { "epoch": 1.29, "learning_rate": 2.8484366205384007e-05, "loss": 0.0, "step": 66536 }, { "epoch": 1.29, "learning_rate": 2.8483719245126773e-05, "loss": 0.0, "step": 66538 }, { "epoch": 1.29, "learning_rate": 2.8483072284869545e-05, "loss": 0.005, "step": 66540 }, { "epoch": 1.29, "learning_rate": 2.848242532461231e-05, "loss": 0.0, "step": 66542 }, { "epoch": 1.29, "learning_rate": 2.8481778364355077e-05, "loss": 0.003, "step": 66544 }, { "epoch": 1.29, "learning_rate": 2.848113140409785e-05, "loss": 0.0, "step": 66546 }, { "epoch": 1.29, "learning_rate": 2.8480484443840615e-05, "loss": 0.0, "step": 66548 }, { "epoch": 1.29, "learning_rate": 2.8479837483583384e-05, "loss": 0.0001, "step": 66550 }, { "epoch": 1.29, "learning_rate": 2.8479190523326153e-05, "loss": 0.0, "step": 66552 }, { "epoch": 1.29, "learning_rate": 2.8478543563068922e-05, "loss": 0.0005, "step": 66554 }, { "epoch": 1.29, "learning_rate": 2.8477896602811695e-05, "loss": 0.0, "step": 66556 }, { "epoch": 1.29, "learning_rate": 2.847724964255446e-05, "loss": 0.0, "step": 66558 }, { "epoch": 1.29, "learning_rate": 2.8476602682297226e-05, "loss": 0.0153, "step": 66560 }, { "epoch": 1.29, "learning_rate": 2.847595572204e-05, "loss": 0.0, "step": 66562 }, { "epoch": 1.29, "learning_rate": 2.8475308761782764e-05, "loss": 0.0, "step": 66564 }, { "epoch": 1.29, "learning_rate": 2.8474661801525533e-05, "loss": 0.0001, "step": 66566 }, { "epoch": 1.29, "learning_rate": 2.8474014841268302e-05, "loss": 0.0, "step": 66568 }, { "epoch": 1.29, "learning_rate": 2.847336788101107e-05, "loss": 0.0, "step": 66570 }, { "epoch": 1.29, "learning_rate": 2.8472720920753837e-05, "loss": 0.0, "step": 66572 }, { "epoch": 1.29, "learning_rate": 2.847207396049661e-05, "loss": 0.0, "step": 66574 }, { "epoch": 1.29, "learning_rate": 2.8471427000239375e-05, "loss": 0.0001, "step": 66576 }, { "epoch": 1.29, "learning_rate": 2.8470780039982148e-05, "loss": 0.0001, "step": 66578 }, { "epoch": 1.29, "learning_rate": 2.8470133079724913e-05, "loss": 0.0, "step": 66580 }, { "epoch": 1.29, "learning_rate": 2.8469486119467683e-05, "loss": 0.0, "step": 66582 }, { "epoch": 1.29, "learning_rate": 2.846883915921045e-05, "loss": 0.0002, "step": 66584 }, { "epoch": 1.29, "learning_rate": 2.846819219895322e-05, "loss": 0.0001, "step": 66586 }, { "epoch": 1.29, "learning_rate": 2.8467545238695986e-05, "loss": 0.0, "step": 66588 }, { "epoch": 1.29, "learning_rate": 2.846689827843876e-05, "loss": 0.001, "step": 66590 }, { "epoch": 1.29, "learning_rate": 2.8466251318181525e-05, "loss": 0.0, "step": 66592 }, { "epoch": 1.29, "learning_rate": 2.846560435792429e-05, "loss": 0.0, "step": 66594 }, { "epoch": 1.29, "learning_rate": 2.8464957397667063e-05, "loss": 0.0002, "step": 66596 }, { "epoch": 1.29, "learning_rate": 2.8464310437409832e-05, "loss": 0.0002, "step": 66598 }, { "epoch": 1.29, "learning_rate": 2.84636634771526e-05, "loss": 0.001, "step": 66600 }, { "epoch": 1.29, "learning_rate": 2.846301651689537e-05, "loss": 0.0001, "step": 66602 }, { "epoch": 1.29, "learning_rate": 2.8462369556638136e-05, "loss": 0.0, "step": 66604 }, { "epoch": 1.29, "learning_rate": 2.8461722596380908e-05, "loss": 0.0, "step": 66606 }, { "epoch": 1.29, "learning_rate": 2.8461075636123674e-05, "loss": 0.0, "step": 66608 }, { "epoch": 1.29, "learning_rate": 2.846042867586644e-05, "loss": 0.0, "step": 66610 }, { "epoch": 1.29, "learning_rate": 2.8459781715609212e-05, "loss": 0.0, "step": 66612 }, { "epoch": 1.29, "learning_rate": 2.8459134755351978e-05, "loss": 0.0, "step": 66614 }, { "epoch": 1.29, "learning_rate": 2.8458487795094747e-05, "loss": 0.0, "step": 66616 }, { "epoch": 1.29, "learning_rate": 2.845784083483752e-05, "loss": 0.0, "step": 66618 }, { "epoch": 1.29, "learning_rate": 2.8457193874580285e-05, "loss": 0.0025, "step": 66620 }, { "epoch": 1.29, "learning_rate": 2.8456546914323057e-05, "loss": 0.0, "step": 66622 }, { "epoch": 1.29, "learning_rate": 2.8455899954065823e-05, "loss": 0.0, "step": 66624 }, { "epoch": 1.29, "learning_rate": 2.845525299380859e-05, "loss": 0.0, "step": 66626 }, { "epoch": 1.29, "learning_rate": 2.845460603355136e-05, "loss": 0.0, "step": 66628 }, { "epoch": 1.29, "learning_rate": 2.8453959073294127e-05, "loss": 0.0, "step": 66630 }, { "epoch": 1.29, "learning_rate": 2.8453312113036896e-05, "loss": 0.0001, "step": 66632 }, { "epoch": 1.29, "learning_rate": 2.845266515277967e-05, "loss": 0.0204, "step": 66634 }, { "epoch": 1.29, "learning_rate": 2.8452018192522434e-05, "loss": 0.0, "step": 66636 }, { "epoch": 1.29, "learning_rate": 2.8451371232265207e-05, "loss": 0.0, "step": 66638 }, { "epoch": 1.29, "learning_rate": 2.8450724272007972e-05, "loss": 0.0, "step": 66640 }, { "epoch": 1.29, "learning_rate": 2.8450077311750738e-05, "loss": 0.0, "step": 66642 }, { "epoch": 1.29, "learning_rate": 2.844943035149351e-05, "loss": 0.0, "step": 66644 }, { "epoch": 1.29, "learning_rate": 2.8448783391236276e-05, "loss": 0.0015, "step": 66646 }, { "epoch": 1.29, "learning_rate": 2.8448136430979045e-05, "loss": 0.0, "step": 66648 }, { "epoch": 1.29, "learning_rate": 2.8447489470721818e-05, "loss": 0.0, "step": 66650 }, { "epoch": 1.29, "learning_rate": 2.8446842510464584e-05, "loss": 0.0, "step": 66652 }, { "epoch": 1.29, "learning_rate": 2.844619555020735e-05, "loss": 0.0, "step": 66654 }, { "epoch": 1.29, "learning_rate": 2.8445548589950122e-05, "loss": 0.0, "step": 66656 }, { "epoch": 1.29, "learning_rate": 2.8444901629692887e-05, "loss": 0.0, "step": 66658 }, { "epoch": 1.29, "learning_rate": 2.844425466943566e-05, "loss": 0.0, "step": 66660 }, { "epoch": 1.29, "learning_rate": 2.8443607709178426e-05, "loss": 0.0, "step": 66662 }, { "epoch": 1.29, "learning_rate": 2.8442960748921195e-05, "loss": 0.0, "step": 66664 }, { "epoch": 1.29, "learning_rate": 2.8442313788663967e-05, "loss": 0.0, "step": 66666 }, { "epoch": 1.29, "learning_rate": 2.8441666828406733e-05, "loss": 0.0002, "step": 66668 }, { "epoch": 1.29, "learning_rate": 2.84410198681495e-05, "loss": 0.0, "step": 66670 }, { "epoch": 1.29, "learning_rate": 2.844037290789227e-05, "loss": 0.0007, "step": 66672 }, { "epoch": 1.29, "learning_rate": 2.8439725947635037e-05, "loss": 0.0, "step": 66674 }, { "epoch": 1.29, "learning_rate": 2.8439078987377802e-05, "loss": 0.0, "step": 66676 }, { "epoch": 1.29, "learning_rate": 2.8438432027120575e-05, "loss": 0.0008, "step": 66678 }, { "epoch": 1.29, "learning_rate": 2.8437785066863344e-05, "loss": 0.0, "step": 66680 }, { "epoch": 1.29, "learning_rate": 2.8437138106606116e-05, "loss": 0.0, "step": 66682 }, { "epoch": 1.29, "learning_rate": 2.8436491146348882e-05, "loss": 0.0, "step": 66684 }, { "epoch": 1.29, "learning_rate": 2.8435844186091648e-05, "loss": 0.0, "step": 66686 }, { "epoch": 1.29, "learning_rate": 2.843519722583442e-05, "loss": 0.0001, "step": 66688 }, { "epoch": 1.29, "learning_rate": 2.8434550265577186e-05, "loss": 0.0, "step": 66690 }, { "epoch": 1.29, "learning_rate": 2.843390330531995e-05, "loss": 0.0, "step": 66692 }, { "epoch": 1.29, "learning_rate": 2.8433256345062724e-05, "loss": 0.0001, "step": 66694 }, { "epoch": 1.29, "learning_rate": 2.8432609384805493e-05, "loss": 0.0, "step": 66696 }, { "epoch": 1.29, "learning_rate": 2.8431962424548266e-05, "loss": 0.0027, "step": 66698 }, { "epoch": 1.29, "learning_rate": 2.843131546429103e-05, "loss": 0.0001, "step": 66700 }, { "epoch": 1.29, "learning_rate": 2.8430668504033797e-05, "loss": 0.0, "step": 66702 }, { "epoch": 1.29, "learning_rate": 2.843002154377657e-05, "loss": 0.0, "step": 66704 }, { "epoch": 1.29, "learning_rate": 2.8429374583519335e-05, "loss": 0.0102, "step": 66706 }, { "epoch": 1.29, "learning_rate": 2.84287276232621e-05, "loss": 0.0066, "step": 66708 }, { "epoch": 1.29, "learning_rate": 2.8428080663004873e-05, "loss": 0.0, "step": 66710 }, { "epoch": 1.29, "learning_rate": 2.8427433702747643e-05, "loss": 0.0025, "step": 66712 }, { "epoch": 1.29, "learning_rate": 2.8426786742490408e-05, "loss": 0.0, "step": 66714 }, { "epoch": 1.29, "learning_rate": 2.842613978223318e-05, "loss": 0.0, "step": 66716 }, { "epoch": 1.29, "learning_rate": 2.8425492821975946e-05, "loss": 0.0, "step": 66718 }, { "epoch": 1.29, "learning_rate": 2.842484586171872e-05, "loss": 0.0, "step": 66720 }, { "epoch": 1.29, "learning_rate": 2.8424198901461485e-05, "loss": 0.0, "step": 66722 }, { "epoch": 1.3, "learning_rate": 2.842355194120425e-05, "loss": 0.0065, "step": 66724 }, { "epoch": 1.3, "learning_rate": 2.8422904980947023e-05, "loss": 0.0, "step": 66726 }, { "epoch": 1.3, "learning_rate": 2.8422258020689792e-05, "loss": 0.0, "step": 66728 }, { "epoch": 1.3, "learning_rate": 2.8421611060432558e-05, "loss": 0.0, "step": 66730 }, { "epoch": 1.3, "learning_rate": 2.842096410017533e-05, "loss": 0.0, "step": 66732 }, { "epoch": 1.3, "learning_rate": 2.8420317139918096e-05, "loss": 0.0, "step": 66734 }, { "epoch": 1.3, "learning_rate": 2.841967017966086e-05, "loss": 0.0, "step": 66736 }, { "epoch": 1.3, "learning_rate": 2.8419023219403634e-05, "loss": 0.0014, "step": 66738 }, { "epoch": 1.3, "learning_rate": 2.84183762591464e-05, "loss": 0.0, "step": 66740 }, { "epoch": 1.3, "learning_rate": 2.8417729298889172e-05, "loss": 0.0, "step": 66742 }, { "epoch": 1.3, "learning_rate": 2.841708233863194e-05, "loss": 0.009, "step": 66744 }, { "epoch": 1.3, "learning_rate": 2.8416435378374707e-05, "loss": 0.0002, "step": 66746 }, { "epoch": 1.3, "learning_rate": 2.841578841811748e-05, "loss": 0.0, "step": 66748 }, { "epoch": 1.3, "learning_rate": 2.8415141457860245e-05, "loss": 0.0, "step": 66750 }, { "epoch": 1.3, "learning_rate": 2.841449449760301e-05, "loss": 0.0, "step": 66752 }, { "epoch": 1.3, "learning_rate": 2.8413847537345783e-05, "loss": 0.0, "step": 66754 }, { "epoch": 1.3, "learning_rate": 2.841320057708855e-05, "loss": 0.0, "step": 66756 }, { "epoch": 1.3, "learning_rate": 2.8412553616831318e-05, "loss": 0.0, "step": 66758 }, { "epoch": 1.3, "learning_rate": 2.841190665657409e-05, "loss": 0.0, "step": 66760 }, { "epoch": 1.3, "learning_rate": 2.8411259696316856e-05, "loss": 0.0, "step": 66762 }, { "epoch": 1.3, "learning_rate": 2.841061273605963e-05, "loss": 0.0, "step": 66764 }, { "epoch": 1.3, "learning_rate": 2.8409965775802394e-05, "loss": 0.0, "step": 66766 }, { "epoch": 1.3, "learning_rate": 2.840931881554516e-05, "loss": 0.0, "step": 66768 }, { "epoch": 1.3, "learning_rate": 2.8408671855287932e-05, "loss": 0.0, "step": 66770 }, { "epoch": 1.3, "learning_rate": 2.8408024895030698e-05, "loss": 0.0, "step": 66772 }, { "epoch": 1.3, "learning_rate": 2.8407377934773467e-05, "loss": 0.0, "step": 66774 }, { "epoch": 1.3, "learning_rate": 2.840673097451624e-05, "loss": 0.0, "step": 66776 }, { "epoch": 1.3, "learning_rate": 2.8406084014259005e-05, "loss": 0.0, "step": 66778 }, { "epoch": 1.3, "learning_rate": 2.8405437054001778e-05, "loss": 0.0, "step": 66780 }, { "epoch": 1.3, "learning_rate": 2.8404790093744544e-05, "loss": 0.0, "step": 66782 }, { "epoch": 1.3, "learning_rate": 2.840414313348731e-05, "loss": 0.0, "step": 66784 }, { "epoch": 1.3, "learning_rate": 2.8403496173230082e-05, "loss": 0.0, "step": 66786 }, { "epoch": 1.3, "learning_rate": 2.8402849212972847e-05, "loss": 0.0, "step": 66788 }, { "epoch": 1.3, "learning_rate": 2.8402202252715616e-05, "loss": 0.0, "step": 66790 }, { "epoch": 1.3, "learning_rate": 2.8401555292458386e-05, "loss": 0.001, "step": 66792 }, { "epoch": 1.3, "learning_rate": 2.8400908332201155e-05, "loss": 0.0, "step": 66794 }, { "epoch": 1.3, "learning_rate": 2.840026137194392e-05, "loss": 0.0, "step": 66796 }, { "epoch": 1.3, "learning_rate": 2.8399614411686693e-05, "loss": 0.0, "step": 66798 }, { "epoch": 1.3, "learning_rate": 2.839896745142946e-05, "loss": 0.0, "step": 66800 }, { "epoch": 1.3, "learning_rate": 2.839832049117223e-05, "loss": 0.0007, "step": 66802 }, { "epoch": 1.3, "learning_rate": 2.8397673530914997e-05, "loss": 0.0007, "step": 66804 }, { "epoch": 1.3, "learning_rate": 2.8397026570657766e-05, "loss": 0.0, "step": 66806 }, { "epoch": 1.3, "learning_rate": 2.8396379610400535e-05, "loss": 0.0011, "step": 66808 }, { "epoch": 1.3, "learning_rate": 2.8395732650143304e-05, "loss": 0.0, "step": 66810 }, { "epoch": 1.3, "learning_rate": 2.839508568988607e-05, "loss": 0.0, "step": 66812 }, { "epoch": 1.3, "learning_rate": 2.8394438729628842e-05, "loss": 0.0001, "step": 66814 }, { "epoch": 1.3, "learning_rate": 2.8393791769371608e-05, "loss": 0.0044, "step": 66816 }, { "epoch": 1.3, "learning_rate": 2.8393144809114373e-05, "loss": 0.0044, "step": 66818 }, { "epoch": 1.3, "learning_rate": 2.8392497848857146e-05, "loss": 0.0, "step": 66820 }, { "epoch": 1.3, "learning_rate": 2.8391850888599915e-05, "loss": 0.0, "step": 66822 }, { "epoch": 1.3, "learning_rate": 2.8391203928342684e-05, "loss": 0.0, "step": 66824 }, { "epoch": 1.3, "learning_rate": 2.8390556968085453e-05, "loss": 0.0001, "step": 66826 }, { "epoch": 1.3, "learning_rate": 2.838991000782822e-05, "loss": 0.0, "step": 66828 }, { "epoch": 1.3, "learning_rate": 2.838926304757099e-05, "loss": 0.0, "step": 66830 }, { "epoch": 1.3, "learning_rate": 2.8388616087313757e-05, "loss": 0.0, "step": 66832 }, { "epoch": 1.3, "learning_rate": 2.8387969127056523e-05, "loss": 0.0003, "step": 66834 }, { "epoch": 1.3, "learning_rate": 2.8387322166799295e-05, "loss": 0.0432, "step": 66836 }, { "epoch": 1.3, "learning_rate": 2.8386675206542064e-05, "loss": 0.0, "step": 66838 }, { "epoch": 1.3, "learning_rate": 2.8386028246284833e-05, "loss": 0.0001, "step": 66840 }, { "epoch": 1.3, "learning_rate": 2.8385381286027603e-05, "loss": 0.0, "step": 66842 }, { "epoch": 1.3, "learning_rate": 2.8384734325770368e-05, "loss": 0.0, "step": 66844 }, { "epoch": 1.3, "learning_rate": 2.838408736551314e-05, "loss": 0.0, "step": 66846 }, { "epoch": 1.3, "learning_rate": 2.8383440405255906e-05, "loss": 0.0003, "step": 66848 }, { "epoch": 1.3, "learning_rate": 2.8382793444998672e-05, "loss": 0.0, "step": 66850 }, { "epoch": 1.3, "learning_rate": 2.8382146484741445e-05, "loss": 0.0, "step": 66852 }, { "epoch": 1.3, "learning_rate": 2.838149952448421e-05, "loss": 0.0186, "step": 66854 }, { "epoch": 1.3, "learning_rate": 2.838085256422698e-05, "loss": 0.0, "step": 66856 }, { "epoch": 1.3, "learning_rate": 2.8380205603969752e-05, "loss": 0.0102, "step": 66858 }, { "epoch": 1.3, "learning_rate": 2.8379558643712517e-05, "loss": 0.0, "step": 66860 }, { "epoch": 1.3, "learning_rate": 2.837891168345529e-05, "loss": 0.0, "step": 66862 }, { "epoch": 1.3, "learning_rate": 2.8378264723198056e-05, "loss": 0.0004, "step": 66864 }, { "epoch": 1.3, "learning_rate": 2.837761776294082e-05, "loss": 0.0, "step": 66866 }, { "epoch": 1.3, "learning_rate": 2.8376970802683594e-05, "loss": 0.0005, "step": 66868 }, { "epoch": 1.3, "learning_rate": 2.837632384242636e-05, "loss": 0.0, "step": 66870 }, { "epoch": 1.3, "learning_rate": 2.837567688216913e-05, "loss": 0.0, "step": 66872 }, { "epoch": 1.3, "learning_rate": 2.83750299219119e-05, "loss": 0.0, "step": 66874 }, { "epoch": 1.3, "learning_rate": 2.8374382961654667e-05, "loss": 0.0001, "step": 66876 }, { "epoch": 1.3, "learning_rate": 2.8373736001397432e-05, "loss": 0.0, "step": 66878 }, { "epoch": 1.3, "learning_rate": 2.8373089041140205e-05, "loss": 0.0, "step": 66880 }, { "epoch": 1.3, "learning_rate": 2.837244208088297e-05, "loss": 0.0, "step": 66882 }, { "epoch": 1.3, "learning_rate": 2.8371795120625743e-05, "loss": 0.0, "step": 66884 }, { "epoch": 1.3, "learning_rate": 2.837114816036851e-05, "loss": 0.0, "step": 66886 }, { "epoch": 1.3, "learning_rate": 2.8370501200111278e-05, "loss": 0.0, "step": 66888 }, { "epoch": 1.3, "learning_rate": 2.836985423985405e-05, "loss": 0.0, "step": 66890 }, { "epoch": 1.3, "learning_rate": 2.8369207279596816e-05, "loss": 0.0, "step": 66892 }, { "epoch": 1.3, "learning_rate": 2.8368560319339582e-05, "loss": 0.0, "step": 66894 }, { "epoch": 1.3, "learning_rate": 2.8367913359082354e-05, "loss": 0.0001, "step": 66896 }, { "epoch": 1.3, "learning_rate": 2.836726639882512e-05, "loss": 0.0, "step": 66898 }, { "epoch": 1.3, "learning_rate": 2.836661943856789e-05, "loss": 0.0, "step": 66900 }, { "epoch": 1.3, "learning_rate": 2.8365972478310658e-05, "loss": 0.0, "step": 66902 }, { "epoch": 1.3, "learning_rate": 2.8365325518053427e-05, "loss": 0.0007, "step": 66904 }, { "epoch": 1.3, "learning_rate": 2.83646785577962e-05, "loss": 0.0, "step": 66906 }, { "epoch": 1.3, "learning_rate": 2.8364031597538965e-05, "loss": 0.0293, "step": 66908 }, { "epoch": 1.3, "learning_rate": 2.836338463728173e-05, "loss": 0.0, "step": 66910 }, { "epoch": 1.3, "learning_rate": 2.8362737677024504e-05, "loss": 0.0, "step": 66912 }, { "epoch": 1.3, "learning_rate": 2.836209071676727e-05, "loss": 0.0, "step": 66914 }, { "epoch": 1.3, "learning_rate": 2.8361443756510035e-05, "loss": 0.0002, "step": 66916 }, { "epoch": 1.3, "learning_rate": 2.8360796796252807e-05, "loss": 0.0, "step": 66918 }, { "epoch": 1.3, "learning_rate": 2.8360149835995576e-05, "loss": 0.0, "step": 66920 }, { "epoch": 1.3, "learning_rate": 2.835950287573835e-05, "loss": 0.0, "step": 66922 }, { "epoch": 1.3, "learning_rate": 2.8358855915481115e-05, "loss": 0.0, "step": 66924 }, { "epoch": 1.3, "learning_rate": 2.835820895522388e-05, "loss": 0.0, "step": 66926 }, { "epoch": 1.3, "learning_rate": 2.8357561994966653e-05, "loss": 0.0, "step": 66928 }, { "epoch": 1.3, "learning_rate": 2.835691503470942e-05, "loss": 0.0, "step": 66930 }, { "epoch": 1.3, "learning_rate": 2.8356268074452184e-05, "loss": 0.0006, "step": 66932 }, { "epoch": 1.3, "learning_rate": 2.8355621114194957e-05, "loss": 0.0, "step": 66934 }, { "epoch": 1.3, "learning_rate": 2.8354974153937726e-05, "loss": 0.0, "step": 66936 }, { "epoch": 1.3, "learning_rate": 2.835432719368049e-05, "loss": 0.023, "step": 66938 }, { "epoch": 1.3, "learning_rate": 2.8353680233423264e-05, "loss": 0.0, "step": 66940 }, { "epoch": 1.3, "learning_rate": 2.835303327316603e-05, "loss": 0.0, "step": 66942 }, { "epoch": 1.3, "learning_rate": 2.8352386312908802e-05, "loss": 0.0, "step": 66944 }, { "epoch": 1.3, "learning_rate": 2.8351739352651568e-05, "loss": 0.0, "step": 66946 }, { "epoch": 1.3, "learning_rate": 2.8351092392394333e-05, "loss": 0.0, "step": 66948 }, { "epoch": 1.3, "learning_rate": 2.8350445432137106e-05, "loss": 0.0001, "step": 66950 }, { "epoch": 1.3, "learning_rate": 2.8349798471879875e-05, "loss": 0.0006, "step": 66952 }, { "epoch": 1.3, "learning_rate": 2.834915151162264e-05, "loss": 0.0, "step": 66954 }, { "epoch": 1.3, "learning_rate": 2.8348504551365413e-05, "loss": 0.0, "step": 66956 }, { "epoch": 1.3, "learning_rate": 2.834785759110818e-05, "loss": 0.0, "step": 66958 }, { "epoch": 1.3, "learning_rate": 2.8347210630850945e-05, "loss": 0.0, "step": 66960 }, { "epoch": 1.3, "learning_rate": 2.8346563670593717e-05, "loss": 0.0006, "step": 66962 }, { "epoch": 1.3, "learning_rate": 2.8345916710336483e-05, "loss": 0.0, "step": 66964 }, { "epoch": 1.3, "learning_rate": 2.8345269750079255e-05, "loss": 0.0, "step": 66966 }, { "epoch": 1.3, "learning_rate": 2.8344622789822024e-05, "loss": 0.0, "step": 66968 }, { "epoch": 1.3, "learning_rate": 2.834397582956479e-05, "loss": 0.0, "step": 66970 }, { "epoch": 1.3, "learning_rate": 2.8343328869307563e-05, "loss": 0.0, "step": 66972 }, { "epoch": 1.3, "learning_rate": 2.8342681909050328e-05, "loss": 0.0, "step": 66974 }, { "epoch": 1.3, "learning_rate": 2.8342034948793094e-05, "loss": 0.0002, "step": 66976 }, { "epoch": 1.3, "learning_rate": 2.8341387988535866e-05, "loss": 0.0, "step": 66978 }, { "epoch": 1.3, "learning_rate": 2.8340741028278632e-05, "loss": 0.0, "step": 66980 }, { "epoch": 1.3, "learning_rate": 2.8340094068021405e-05, "loss": 0.0049, "step": 66982 }, { "epoch": 1.3, "learning_rate": 2.8339447107764174e-05, "loss": 0.0001, "step": 66984 }, { "epoch": 1.3, "learning_rate": 2.833880014750694e-05, "loss": 0.0006, "step": 66986 }, { "epoch": 1.3, "learning_rate": 2.8338153187249712e-05, "loss": 0.0, "step": 66988 }, { "epoch": 1.3, "learning_rate": 2.8337506226992477e-05, "loss": 0.0, "step": 66990 }, { "epoch": 1.3, "learning_rate": 2.8336859266735243e-05, "loss": 0.0, "step": 66992 }, { "epoch": 1.3, "learning_rate": 2.8336212306478016e-05, "loss": 0.0, "step": 66994 }, { "epoch": 1.3, "learning_rate": 2.833556534622078e-05, "loss": 0.0, "step": 66996 }, { "epoch": 1.3, "learning_rate": 2.833491838596355e-05, "loss": 0.0, "step": 66998 }, { "epoch": 1.3, "learning_rate": 2.8334271425706323e-05, "loss": 0.0002, "step": 67000 }, { "epoch": 1.3, "learning_rate": 2.833362446544909e-05, "loss": 0.0, "step": 67002 }, { "epoch": 1.3, "learning_rate": 2.833297750519186e-05, "loss": 0.0, "step": 67004 }, { "epoch": 1.3, "learning_rate": 2.8332330544934627e-05, "loss": 0.0006, "step": 67006 }, { "epoch": 1.3, "learning_rate": 2.8331683584677392e-05, "loss": 0.0, "step": 67008 }, { "epoch": 1.3, "learning_rate": 2.8331036624420165e-05, "loss": 0.0105, "step": 67010 }, { "epoch": 1.3, "learning_rate": 2.833038966416293e-05, "loss": 0.0, "step": 67012 }, { "epoch": 1.3, "learning_rate": 2.83297427039057e-05, "loss": 0.0, "step": 67014 }, { "epoch": 1.3, "learning_rate": 2.8329095743648472e-05, "loss": 0.0, "step": 67016 }, { "epoch": 1.3, "learning_rate": 2.8328448783391238e-05, "loss": 0.0, "step": 67018 }, { "epoch": 1.3, "learning_rate": 2.8327801823134004e-05, "loss": 0.0, "step": 67020 }, { "epoch": 1.3, "learning_rate": 2.8327154862876776e-05, "loss": 0.0002, "step": 67022 }, { "epoch": 1.3, "learning_rate": 2.8326507902619542e-05, "loss": 0.0003, "step": 67024 }, { "epoch": 1.3, "learning_rate": 2.8325860942362314e-05, "loss": 0.0003, "step": 67026 }, { "epoch": 1.3, "learning_rate": 2.832521398210508e-05, "loss": 0.0397, "step": 67028 }, { "epoch": 1.3, "learning_rate": 2.832456702184785e-05, "loss": 0.0419, "step": 67030 }, { "epoch": 1.3, "learning_rate": 2.8323920061590618e-05, "loss": 0.0, "step": 67032 }, { "epoch": 1.3, "learning_rate": 2.8323273101333387e-05, "loss": 0.0022, "step": 67034 }, { "epoch": 1.3, "learning_rate": 2.8322626141076153e-05, "loss": 0.0002, "step": 67036 }, { "epoch": 1.3, "learning_rate": 2.8321979180818925e-05, "loss": 0.0, "step": 67038 }, { "epoch": 1.3, "learning_rate": 2.832133222056169e-05, "loss": 0.0, "step": 67040 }, { "epoch": 1.3, "learning_rate": 2.8320685260304457e-05, "loss": 0.0, "step": 67042 }, { "epoch": 1.3, "learning_rate": 2.832003830004723e-05, "loss": 0.0001, "step": 67044 }, { "epoch": 1.3, "learning_rate": 2.8319391339789998e-05, "loss": 0.0, "step": 67046 }, { "epoch": 1.3, "learning_rate": 2.8318744379532767e-05, "loss": 0.0, "step": 67048 }, { "epoch": 1.3, "learning_rate": 2.8318097419275536e-05, "loss": 0.0004, "step": 67050 }, { "epoch": 1.3, "learning_rate": 2.8317450459018302e-05, "loss": 0.0, "step": 67052 }, { "epoch": 1.3, "learning_rate": 2.8316803498761075e-05, "loss": 0.0, "step": 67054 }, { "epoch": 1.3, "learning_rate": 2.831615653850384e-05, "loss": 0.0027, "step": 67056 }, { "epoch": 1.3, "learning_rate": 2.8315509578246606e-05, "loss": 0.0, "step": 67058 }, { "epoch": 1.3, "learning_rate": 2.831486261798938e-05, "loss": 0.0, "step": 67060 }, { "epoch": 1.3, "learning_rate": 2.8314215657732148e-05, "loss": 0.0002, "step": 67062 }, { "epoch": 1.3, "learning_rate": 2.8313568697474917e-05, "loss": 0.0, "step": 67064 }, { "epoch": 1.3, "learning_rate": 2.8312921737217686e-05, "loss": 0.0021, "step": 67066 }, { "epoch": 1.3, "learning_rate": 2.831227477696045e-05, "loss": 0.0, "step": 67068 }, { "epoch": 1.3, "learning_rate": 2.8311627816703224e-05, "loss": 0.0, "step": 67070 }, { "epoch": 1.3, "learning_rate": 2.831098085644599e-05, "loss": 0.0, "step": 67072 }, { "epoch": 1.3, "learning_rate": 2.8310333896188755e-05, "loss": 0.0, "step": 67074 }, { "epoch": 1.3, "learning_rate": 2.8309686935931528e-05, "loss": 0.0001, "step": 67076 }, { "epoch": 1.3, "learning_rate": 2.8309039975674297e-05, "loss": 0.0, "step": 67078 }, { "epoch": 1.3, "learning_rate": 2.8308393015417063e-05, "loss": 0.0003, "step": 67080 }, { "epoch": 1.3, "learning_rate": 2.8307746055159835e-05, "loss": 0.0055, "step": 67082 }, { "epoch": 1.3, "learning_rate": 2.83070990949026e-05, "loss": 0.0, "step": 67084 }, { "epoch": 1.3, "learning_rate": 2.8306452134645373e-05, "loss": 0.0, "step": 67086 }, { "epoch": 1.3, "learning_rate": 2.830580517438814e-05, "loss": 0.0, "step": 67088 }, { "epoch": 1.3, "learning_rate": 2.8305158214130905e-05, "loss": 0.0004, "step": 67090 }, { "epoch": 1.3, "learning_rate": 2.8304511253873677e-05, "loss": 0.0001, "step": 67092 }, { "epoch": 1.3, "learning_rate": 2.8303864293616443e-05, "loss": 0.0, "step": 67094 }, { "epoch": 1.3, "learning_rate": 2.8303217333359212e-05, "loss": 0.0, "step": 67096 }, { "epoch": 1.3, "learning_rate": 2.8302570373101984e-05, "loss": 0.0, "step": 67098 }, { "epoch": 1.3, "learning_rate": 2.830192341284475e-05, "loss": 0.0, "step": 67100 }, { "epoch": 1.3, "learning_rate": 2.8301276452587516e-05, "loss": 0.0002, "step": 67102 }, { "epoch": 1.3, "learning_rate": 2.8300629492330288e-05, "loss": 0.0, "step": 67104 }, { "epoch": 1.3, "learning_rate": 2.8299982532073054e-05, "loss": 0.0, "step": 67106 }, { "epoch": 1.3, "learning_rate": 2.8299335571815826e-05, "loss": 0.0, "step": 67108 }, { "epoch": 1.3, "learning_rate": 2.8298688611558592e-05, "loss": 0.0001, "step": 67110 }, { "epoch": 1.3, "learning_rate": 2.829804165130136e-05, "loss": 0.0, "step": 67112 }, { "epoch": 1.3, "learning_rate": 2.8297394691044134e-05, "loss": 0.0003, "step": 67114 }, { "epoch": 1.3, "learning_rate": 2.82967477307869e-05, "loss": 0.0, "step": 67116 }, { "epoch": 1.3, "learning_rate": 2.8296100770529665e-05, "loss": 0.0002, "step": 67118 }, { "epoch": 1.3, "learning_rate": 2.8295453810272437e-05, "loss": 0.0, "step": 67120 }, { "epoch": 1.3, "learning_rate": 2.8294806850015203e-05, "loss": 0.0, "step": 67122 }, { "epoch": 1.3, "learning_rate": 2.8294159889757976e-05, "loss": 0.0, "step": 67124 }, { "epoch": 1.3, "learning_rate": 2.829351292950074e-05, "loss": 0.0, "step": 67126 }, { "epoch": 1.3, "learning_rate": 2.829286596924351e-05, "loss": 0.0, "step": 67128 }, { "epoch": 1.3, "learning_rate": 2.8292219008986283e-05, "loss": 0.0, "step": 67130 }, { "epoch": 1.3, "learning_rate": 2.829157204872905e-05, "loss": 0.0, "step": 67132 }, { "epoch": 1.3, "learning_rate": 2.8290925088471814e-05, "loss": 0.0, "step": 67134 }, { "epoch": 1.3, "learning_rate": 2.8290278128214587e-05, "loss": 0.0, "step": 67136 }, { "epoch": 1.3, "learning_rate": 2.8289631167957352e-05, "loss": 0.0, "step": 67138 }, { "epoch": 1.3, "learning_rate": 2.828898420770012e-05, "loss": 0.0002, "step": 67140 }, { "epoch": 1.3, "learning_rate": 2.828833724744289e-05, "loss": 0.0, "step": 67142 }, { "epoch": 1.3, "learning_rate": 2.828769028718566e-05, "loss": 0.0, "step": 67144 }, { "epoch": 1.3, "learning_rate": 2.8287043326928432e-05, "loss": 0.0059, "step": 67146 }, { "epoch": 1.3, "learning_rate": 2.8286396366671198e-05, "loss": 0.0, "step": 67148 }, { "epoch": 1.3, "learning_rate": 2.8285749406413964e-05, "loss": 0.0, "step": 67150 }, { "epoch": 1.3, "learning_rate": 2.8285102446156736e-05, "loss": 0.0, "step": 67152 }, { "epoch": 1.3, "learning_rate": 2.8284455485899502e-05, "loss": 0.0, "step": 67154 }, { "epoch": 1.3, "learning_rate": 2.8283808525642267e-05, "loss": 0.0005, "step": 67156 }, { "epoch": 1.3, "learning_rate": 2.828316156538504e-05, "loss": 0.0006, "step": 67158 }, { "epoch": 1.3, "learning_rate": 2.828251460512781e-05, "loss": 0.0001, "step": 67160 }, { "epoch": 1.3, "learning_rate": 2.8281867644870575e-05, "loss": 0.0, "step": 67162 }, { "epoch": 1.3, "learning_rate": 2.8281220684613347e-05, "loss": 0.0, "step": 67164 }, { "epoch": 1.3, "learning_rate": 2.8280573724356113e-05, "loss": 0.0, "step": 67166 }, { "epoch": 1.3, "learning_rate": 2.8279926764098885e-05, "loss": 0.0, "step": 67168 }, { "epoch": 1.3, "learning_rate": 2.827927980384165e-05, "loss": 0.0, "step": 67170 }, { "epoch": 1.3, "learning_rate": 2.8278632843584417e-05, "loss": 0.001, "step": 67172 }, { "epoch": 1.3, "learning_rate": 2.827798588332719e-05, "loss": 0.0, "step": 67174 }, { "epoch": 1.3, "learning_rate": 2.8277338923069958e-05, "loss": 0.0115, "step": 67176 }, { "epoch": 1.3, "learning_rate": 2.8276691962812724e-05, "loss": 0.0, "step": 67178 }, { "epoch": 1.3, "learning_rate": 2.8276045002555496e-05, "loss": 0.0, "step": 67180 }, { "epoch": 1.3, "learning_rate": 2.8275398042298262e-05, "loss": 0.0, "step": 67182 }, { "epoch": 1.3, "learning_rate": 2.8274751082041028e-05, "loss": 0.0043, "step": 67184 }, { "epoch": 1.3, "learning_rate": 2.82741041217838e-05, "loss": 0.0, "step": 67186 }, { "epoch": 1.3, "learning_rate": 2.8273457161526566e-05, "loss": 0.0, "step": 67188 }, { "epoch": 1.3, "learning_rate": 2.827281020126934e-05, "loss": 0.0, "step": 67190 }, { "epoch": 1.3, "learning_rate": 2.8272163241012108e-05, "loss": 0.0, "step": 67192 }, { "epoch": 1.3, "learning_rate": 2.8271516280754873e-05, "loss": 0.0, "step": 67194 }, { "epoch": 1.3, "learning_rate": 2.8270869320497646e-05, "loss": 0.0, "step": 67196 }, { "epoch": 1.3, "learning_rate": 2.827022236024041e-05, "loss": 0.0, "step": 67198 }, { "epoch": 1.3, "learning_rate": 2.8269575399983177e-05, "loss": 0.0, "step": 67200 }, { "epoch": 1.3, "learning_rate": 2.826892843972595e-05, "loss": 0.0, "step": 67202 }, { "epoch": 1.3, "learning_rate": 2.8268281479468715e-05, "loss": 0.0121, "step": 67204 }, { "epoch": 1.3, "learning_rate": 2.8267634519211488e-05, "loss": 0.0021, "step": 67206 }, { "epoch": 1.3, "learning_rate": 2.8266987558954257e-05, "loss": 0.0, "step": 67208 }, { "epoch": 1.3, "learning_rate": 2.8266340598697023e-05, "loss": 0.0, "step": 67210 }, { "epoch": 1.3, "learning_rate": 2.8265693638439795e-05, "loss": 0.0, "step": 67212 }, { "epoch": 1.3, "learning_rate": 2.826504667818256e-05, "loss": 0.0, "step": 67214 }, { "epoch": 1.3, "learning_rate": 2.8264399717925326e-05, "loss": 0.0001, "step": 67216 }, { "epoch": 1.3, "learning_rate": 2.82637527576681e-05, "loss": 0.019, "step": 67218 }, { "epoch": 1.3, "learning_rate": 2.8263105797410865e-05, "loss": 0.0, "step": 67220 }, { "epoch": 1.3, "learning_rate": 2.8262458837153634e-05, "loss": 0.0001, "step": 67222 }, { "epoch": 1.3, "learning_rate": 2.8261811876896406e-05, "loss": 0.0, "step": 67224 }, { "epoch": 1.3, "learning_rate": 2.8261164916639172e-05, "loss": 0.0011, "step": 67226 }, { "epoch": 1.3, "learning_rate": 2.8260517956381944e-05, "loss": 0.0, "step": 67228 }, { "epoch": 1.3, "learning_rate": 2.825987099612471e-05, "loss": 0.0, "step": 67230 }, { "epoch": 1.3, "learning_rate": 2.8259224035867476e-05, "loss": 0.0, "step": 67232 }, { "epoch": 1.3, "learning_rate": 2.8258577075610248e-05, "loss": 0.0, "step": 67234 }, { "epoch": 1.3, "learning_rate": 2.8257930115353014e-05, "loss": 0.0, "step": 67236 }, { "epoch": 1.31, "learning_rate": 2.8257283155095783e-05, "loss": 0.0, "step": 67238 }, { "epoch": 1.31, "learning_rate": 2.8256636194838555e-05, "loss": 0.0001, "step": 67240 }, { "epoch": 1.31, "learning_rate": 2.825598923458132e-05, "loss": 0.0, "step": 67242 }, { "epoch": 1.31, "learning_rate": 2.8255342274324087e-05, "loss": 0.0, "step": 67244 }, { "epoch": 1.31, "learning_rate": 2.825469531406686e-05, "loss": 0.0003, "step": 67246 }, { "epoch": 1.31, "learning_rate": 2.8254048353809625e-05, "loss": 0.0, "step": 67248 }, { "epoch": 1.31, "learning_rate": 2.8253401393552397e-05, "loss": 0.0033, "step": 67250 }, { "epoch": 1.31, "learning_rate": 2.8252754433295163e-05, "loss": 0.0, "step": 67252 }, { "epoch": 1.31, "learning_rate": 2.8252107473037932e-05, "loss": 0.01, "step": 67254 }, { "epoch": 1.31, "learning_rate": 2.8251460512780705e-05, "loss": 0.0001, "step": 67256 }, { "epoch": 1.31, "learning_rate": 2.825081355252347e-05, "loss": 0.0, "step": 67258 }, { "epoch": 1.31, "learning_rate": 2.8250166592266236e-05, "loss": 0.0048, "step": 67260 }, { "epoch": 1.31, "learning_rate": 2.824951963200901e-05, "loss": 0.0, "step": 67262 }, { "epoch": 1.31, "learning_rate": 2.8248872671751774e-05, "loss": 0.0, "step": 67264 }, { "epoch": 1.31, "learning_rate": 2.8248225711494547e-05, "loss": 0.0038, "step": 67266 }, { "epoch": 1.31, "learning_rate": 2.8247578751237312e-05, "loss": 0.0, "step": 67268 }, { "epoch": 1.31, "learning_rate": 2.824693179098008e-05, "loss": 0.0022, "step": 67270 }, { "epoch": 1.31, "learning_rate": 2.824628483072285e-05, "loss": 0.0006, "step": 67272 }, { "epoch": 1.31, "learning_rate": 2.824563787046562e-05, "loss": 0.0001, "step": 67274 }, { "epoch": 1.31, "learning_rate": 2.8244990910208385e-05, "loss": 0.0, "step": 67276 }, { "epoch": 1.31, "learning_rate": 2.8244343949951158e-05, "loss": 0.0, "step": 67278 }, { "epoch": 1.31, "learning_rate": 2.8243696989693924e-05, "loss": 0.0, "step": 67280 }, { "epoch": 1.31, "learning_rate": 2.824305002943669e-05, "loss": 0.0025, "step": 67282 }, { "epoch": 1.31, "learning_rate": 2.8242403069179462e-05, "loss": 0.0, "step": 67284 }, { "epoch": 1.31, "learning_rate": 2.824175610892223e-05, "loss": 0.0, "step": 67286 }, { "epoch": 1.31, "learning_rate": 2.8241109148665e-05, "loss": 0.0001, "step": 67288 }, { "epoch": 1.31, "learning_rate": 2.824046218840777e-05, "loss": 0.0, "step": 67290 }, { "epoch": 1.31, "learning_rate": 2.8239815228150535e-05, "loss": 0.0004, "step": 67292 }, { "epoch": 1.31, "learning_rate": 2.8239168267893307e-05, "loss": 0.0, "step": 67294 }, { "epoch": 1.31, "learning_rate": 2.8238521307636073e-05, "loss": 0.0, "step": 67296 }, { "epoch": 1.31, "learning_rate": 2.823787434737884e-05, "loss": 0.0, "step": 67298 }, { "epoch": 1.31, "learning_rate": 2.823722738712161e-05, "loss": 0.0001, "step": 67300 }, { "epoch": 1.31, "learning_rate": 2.823658042686438e-05, "loss": 0.0055, "step": 67302 }, { "epoch": 1.31, "learning_rate": 2.8235933466607146e-05, "loss": 0.0, "step": 67304 }, { "epoch": 1.31, "learning_rate": 2.8235286506349918e-05, "loss": 0.1243, "step": 67306 }, { "epoch": 1.31, "learning_rate": 2.8234639546092684e-05, "loss": 0.0, "step": 67308 }, { "epoch": 1.31, "learning_rate": 2.8233992585835456e-05, "loss": 0.0, "step": 67310 }, { "epoch": 1.31, "learning_rate": 2.8233345625578222e-05, "loss": 0.0, "step": 67312 }, { "epoch": 1.31, "learning_rate": 2.8232698665320988e-05, "loss": 0.0, "step": 67314 }, { "epoch": 1.31, "learning_rate": 2.823205170506376e-05, "loss": 0.0306, "step": 67316 }, { "epoch": 1.31, "learning_rate": 2.823140474480653e-05, "loss": 0.0111, "step": 67318 }, { "epoch": 1.31, "learning_rate": 2.8230757784549295e-05, "loss": 0.0, "step": 67320 }, { "epoch": 1.31, "learning_rate": 2.8230110824292068e-05, "loss": 0.0076, "step": 67322 }, { "epoch": 1.31, "learning_rate": 2.8229463864034833e-05, "loss": 0.0001, "step": 67324 }, { "epoch": 1.31, "learning_rate": 2.82288169037776e-05, "loss": 0.0002, "step": 67326 }, { "epoch": 1.31, "learning_rate": 2.822816994352037e-05, "loss": 0.0, "step": 67328 }, { "epoch": 1.31, "learning_rate": 2.8227522983263137e-05, "loss": 0.0, "step": 67330 }, { "epoch": 1.31, "learning_rate": 2.822687602300591e-05, "loss": 0.0, "step": 67332 }, { "epoch": 1.31, "learning_rate": 2.8226229062748675e-05, "loss": 0.0, "step": 67334 }, { "epoch": 1.31, "learning_rate": 2.8225582102491444e-05, "loss": 0.0, "step": 67336 }, { "epoch": 1.31, "learning_rate": 2.8224935142234217e-05, "loss": 0.0, "step": 67338 }, { "epoch": 1.31, "learning_rate": 2.8224288181976983e-05, "loss": 0.0009, "step": 67340 }, { "epoch": 1.31, "learning_rate": 2.8223641221719748e-05, "loss": 0.0004, "step": 67342 }, { "epoch": 1.31, "learning_rate": 2.822299426146252e-05, "loss": 0.0, "step": 67344 }, { "epoch": 1.31, "learning_rate": 2.8222347301205286e-05, "loss": 0.0, "step": 67346 }, { "epoch": 1.31, "learning_rate": 2.822170034094806e-05, "loss": 0.0111, "step": 67348 }, { "epoch": 1.31, "learning_rate": 2.8221053380690825e-05, "loss": 0.0, "step": 67350 }, { "epoch": 1.31, "learning_rate": 2.8220406420433594e-05, "loss": 0.0001, "step": 67352 }, { "epoch": 1.31, "learning_rate": 2.8219759460176366e-05, "loss": 0.0, "step": 67354 }, { "epoch": 1.31, "learning_rate": 2.8219112499919132e-05, "loss": 0.0001, "step": 67356 }, { "epoch": 1.31, "learning_rate": 2.8218465539661898e-05, "loss": 0.0, "step": 67358 }, { "epoch": 1.31, "learning_rate": 2.821781857940467e-05, "loss": 0.0, "step": 67360 }, { "epoch": 1.31, "learning_rate": 2.8217171619147436e-05, "loss": 0.0, "step": 67362 }, { "epoch": 1.31, "learning_rate": 2.8216524658890205e-05, "loss": 0.0, "step": 67364 }, { "epoch": 1.31, "learning_rate": 2.8215877698632974e-05, "loss": 0.0, "step": 67366 }, { "epoch": 1.31, "learning_rate": 2.8215230738375743e-05, "loss": 0.0, "step": 67368 }, { "epoch": 1.31, "learning_rate": 2.8214583778118515e-05, "loss": 0.0, "step": 67370 }, { "epoch": 1.31, "learning_rate": 2.821393681786128e-05, "loss": 0.0, "step": 67372 }, { "epoch": 1.31, "learning_rate": 2.8213289857604047e-05, "loss": 0.0, "step": 67374 }, { "epoch": 1.31, "learning_rate": 2.821264289734682e-05, "loss": 0.0, "step": 67376 }, { "epoch": 1.31, "learning_rate": 2.8211995937089585e-05, "loss": 0.0, "step": 67378 }, { "epoch": 1.31, "learning_rate": 2.8211348976832354e-05, "loss": 0.0, "step": 67380 }, { "epoch": 1.31, "learning_rate": 2.8210702016575123e-05, "loss": 0.0005, "step": 67382 }, { "epoch": 1.31, "learning_rate": 2.8210055056317892e-05, "loss": 0.0, "step": 67384 }, { "epoch": 1.31, "learning_rate": 2.8209408096060658e-05, "loss": 0.0, "step": 67386 }, { "epoch": 1.31, "learning_rate": 2.820876113580343e-05, "loss": 0.0, "step": 67388 }, { "epoch": 1.31, "learning_rate": 2.8208114175546196e-05, "loss": 0.0, "step": 67390 }, { "epoch": 1.31, "learning_rate": 2.820746721528897e-05, "loss": 0.0, "step": 67392 }, { "epoch": 1.31, "learning_rate": 2.8206820255031734e-05, "loss": 0.0, "step": 67394 }, { "epoch": 1.31, "learning_rate": 2.82061732947745e-05, "loss": 0.0, "step": 67396 }, { "epoch": 1.31, "learning_rate": 2.8205526334517272e-05, "loss": 0.0, "step": 67398 }, { "epoch": 1.31, "learning_rate": 2.820487937426004e-05, "loss": 0.0001, "step": 67400 }, { "epoch": 1.31, "learning_rate": 2.8204232414002807e-05, "loss": 0.0, "step": 67402 }, { "epoch": 1.31, "learning_rate": 2.820358545374558e-05, "loss": 0.0056, "step": 67404 }, { "epoch": 1.31, "learning_rate": 2.8202938493488345e-05, "loss": 0.0, "step": 67406 }, { "epoch": 1.31, "learning_rate": 2.8202291533231118e-05, "loss": 0.0, "step": 67408 }, { "epoch": 1.31, "learning_rate": 2.8201644572973884e-05, "loss": 0.0, "step": 67410 }, { "epoch": 1.31, "learning_rate": 2.820099761271665e-05, "loss": 0.0161, "step": 67412 }, { "epoch": 1.31, "learning_rate": 2.8200350652459422e-05, "loss": 0.0, "step": 67414 }, { "epoch": 1.31, "learning_rate": 2.819970369220219e-05, "loss": 0.0001, "step": 67416 }, { "epoch": 1.31, "learning_rate": 2.8199056731944957e-05, "loss": 0.0, "step": 67418 }, { "epoch": 1.31, "learning_rate": 2.819840977168773e-05, "loss": 0.0, "step": 67420 }, { "epoch": 1.31, "learning_rate": 2.8197762811430495e-05, "loss": 0.0032, "step": 67422 }, { "epoch": 1.31, "learning_rate": 2.819711585117326e-05, "loss": 0.0001, "step": 67424 }, { "epoch": 1.31, "learning_rate": 2.8196468890916033e-05, "loss": 0.0, "step": 67426 }, { "epoch": 1.31, "learning_rate": 2.81958219306588e-05, "loss": 0.0, "step": 67428 }, { "epoch": 1.31, "learning_rate": 2.819517497040157e-05, "loss": 0.0, "step": 67430 }, { "epoch": 1.31, "learning_rate": 2.819452801014434e-05, "loss": 0.0, "step": 67432 }, { "epoch": 1.31, "learning_rate": 2.8193881049887106e-05, "loss": 0.0, "step": 67434 }, { "epoch": 1.31, "learning_rate": 2.8193234089629878e-05, "loss": 0.0, "step": 67436 }, { "epoch": 1.31, "learning_rate": 2.8192587129372644e-05, "loss": 0.0, "step": 67438 }, { "epoch": 1.31, "learning_rate": 2.819194016911541e-05, "loss": 0.0001, "step": 67440 }, { "epoch": 1.31, "learning_rate": 2.8191293208858182e-05, "loss": 0.0001, "step": 67442 }, { "epoch": 1.31, "learning_rate": 2.8190646248600948e-05, "loss": 0.0001, "step": 67444 }, { "epoch": 1.31, "learning_rate": 2.8189999288343717e-05, "loss": 0.0008, "step": 67446 }, { "epoch": 1.31, "learning_rate": 2.818935232808649e-05, "loss": 0.001, "step": 67448 }, { "epoch": 1.31, "learning_rate": 2.8188705367829255e-05, "loss": 0.0178, "step": 67450 }, { "epoch": 1.31, "learning_rate": 2.8188058407572028e-05, "loss": 0.0, "step": 67452 }, { "epoch": 1.31, "learning_rate": 2.8187411447314793e-05, "loss": 0.0, "step": 67454 }, { "epoch": 1.31, "learning_rate": 2.818676448705756e-05, "loss": 0.0, "step": 67456 }, { "epoch": 1.31, "learning_rate": 2.818611752680033e-05, "loss": 0.0, "step": 67458 }, { "epoch": 1.31, "learning_rate": 2.8185470566543097e-05, "loss": 0.0, "step": 67460 }, { "epoch": 1.31, "learning_rate": 2.8184823606285866e-05, "loss": 0.0002, "step": 67462 }, { "epoch": 1.31, "learning_rate": 2.818417664602864e-05, "loss": 0.0, "step": 67464 }, { "epoch": 1.31, "learning_rate": 2.8183529685771404e-05, "loss": 0.0, "step": 67466 }, { "epoch": 1.31, "learning_rate": 2.818288272551417e-05, "loss": 0.0, "step": 67468 }, { "epoch": 1.31, "learning_rate": 2.8182235765256943e-05, "loss": 0.0, "step": 67470 }, { "epoch": 1.31, "learning_rate": 2.8181588804999708e-05, "loss": 0.0, "step": 67472 }, { "epoch": 1.31, "learning_rate": 2.818094184474248e-05, "loss": 0.0, "step": 67474 }, { "epoch": 1.31, "learning_rate": 2.8180294884485246e-05, "loss": 0.0022, "step": 67476 }, { "epoch": 1.31, "learning_rate": 2.8179647924228015e-05, "loss": 0.0, "step": 67478 }, { "epoch": 1.31, "learning_rate": 2.8179000963970788e-05, "loss": 0.0, "step": 67480 }, { "epoch": 1.31, "learning_rate": 2.8178354003713554e-05, "loss": 0.0, "step": 67482 }, { "epoch": 1.31, "learning_rate": 2.817770704345632e-05, "loss": 0.0, "step": 67484 }, { "epoch": 1.31, "learning_rate": 2.8177060083199092e-05, "loss": 0.0034, "step": 67486 }, { "epoch": 1.31, "learning_rate": 2.8176413122941858e-05, "loss": 0.0001, "step": 67488 }, { "epoch": 1.31, "learning_rate": 2.817576616268463e-05, "loss": 0.0, "step": 67490 }, { "epoch": 1.31, "learning_rate": 2.8175119202427396e-05, "loss": 0.0002, "step": 67492 }, { "epoch": 1.31, "learning_rate": 2.8174472242170165e-05, "loss": 0.0, "step": 67494 }, { "epoch": 1.31, "learning_rate": 2.8173825281912937e-05, "loss": 0.0, "step": 67496 }, { "epoch": 1.31, "learning_rate": 2.8173178321655703e-05, "loss": 0.0, "step": 67498 }, { "epoch": 1.31, "learning_rate": 2.817253136139847e-05, "loss": 0.0, "step": 67500 }, { "epoch": 1.31, "learning_rate": 2.817188440114124e-05, "loss": 0.0, "step": 67502 }, { "epoch": 1.31, "learning_rate": 2.8171237440884007e-05, "loss": 0.0181, "step": 67504 }, { "epoch": 1.31, "learning_rate": 2.8170590480626773e-05, "loss": 0.0, "step": 67506 }, { "epoch": 1.31, "learning_rate": 2.8169943520369545e-05, "loss": 0.0, "step": 67508 }, { "epoch": 1.31, "learning_rate": 2.8169296560112314e-05, "loss": 0.0, "step": 67510 }, { "epoch": 1.31, "learning_rate": 2.8168649599855083e-05, "loss": 0.0002, "step": 67512 }, { "epoch": 1.31, "learning_rate": 2.8168002639597852e-05, "loss": 0.0, "step": 67514 }, { "epoch": 1.31, "learning_rate": 2.8167355679340618e-05, "loss": 0.0001, "step": 67516 }, { "epoch": 1.31, "learning_rate": 2.816670871908339e-05, "loss": 0.0001, "step": 67518 }, { "epoch": 1.31, "learning_rate": 2.8166061758826156e-05, "loss": 0.0, "step": 67520 }, { "epoch": 1.31, "learning_rate": 2.8165414798568922e-05, "loss": 0.0, "step": 67522 }, { "epoch": 1.31, "learning_rate": 2.8164767838311694e-05, "loss": 0.0001, "step": 67524 }, { "epoch": 1.31, "learning_rate": 2.8164120878054463e-05, "loss": 0.0, "step": 67526 }, { "epoch": 1.31, "learning_rate": 2.816347391779723e-05, "loss": 0.0, "step": 67528 }, { "epoch": 1.31, "learning_rate": 2.816282695754e-05, "loss": 0.0, "step": 67530 }, { "epoch": 1.31, "learning_rate": 2.8162179997282767e-05, "loss": 0.0, "step": 67532 }, { "epoch": 1.31, "learning_rate": 2.816153303702554e-05, "loss": 0.0, "step": 67534 }, { "epoch": 1.31, "learning_rate": 2.8160886076768305e-05, "loss": 0.0, "step": 67536 }, { "epoch": 1.31, "learning_rate": 2.816023911651107e-05, "loss": 0.0, "step": 67538 }, { "epoch": 1.31, "learning_rate": 2.8159592156253844e-05, "loss": 0.0, "step": 67540 }, { "epoch": 1.31, "learning_rate": 2.8158945195996613e-05, "loss": 0.0, "step": 67542 }, { "epoch": 1.31, "learning_rate": 2.815829823573938e-05, "loss": 0.0001, "step": 67544 }, { "epoch": 1.31, "learning_rate": 2.815765127548215e-05, "loss": 0.0, "step": 67546 }, { "epoch": 1.31, "learning_rate": 2.8157004315224916e-05, "loss": 0.0005, "step": 67548 }, { "epoch": 1.31, "learning_rate": 2.815635735496769e-05, "loss": 0.0, "step": 67550 }, { "epoch": 1.31, "learning_rate": 2.8155710394710455e-05, "loss": 0.0, "step": 67552 }, { "epoch": 1.31, "learning_rate": 2.815506343445322e-05, "loss": 0.0, "step": 67554 }, { "epoch": 1.31, "learning_rate": 2.8154416474195993e-05, "loss": 0.0, "step": 67556 }, { "epoch": 1.31, "learning_rate": 2.8153769513938762e-05, "loss": 0.0, "step": 67558 }, { "epoch": 1.31, "learning_rate": 2.8153122553681528e-05, "loss": 0.0, "step": 67560 }, { "epoch": 1.31, "learning_rate": 2.81524755934243e-05, "loss": 0.0, "step": 67562 }, { "epoch": 1.31, "learning_rate": 2.8151828633167066e-05, "loss": 0.0, "step": 67564 }, { "epoch": 1.31, "learning_rate": 2.815118167290983e-05, "loss": 0.0, "step": 67566 }, { "epoch": 1.31, "learning_rate": 2.8150534712652604e-05, "loss": 0.0, "step": 67568 }, { "epoch": 1.31, "learning_rate": 2.814988775239537e-05, "loss": 0.0009, "step": 67570 }, { "epoch": 1.31, "learning_rate": 2.8149240792138142e-05, "loss": 0.0, "step": 67572 }, { "epoch": 1.31, "learning_rate": 2.8148593831880908e-05, "loss": 0.0, "step": 67574 }, { "epoch": 1.31, "learning_rate": 2.8147946871623677e-05, "loss": 0.0, "step": 67576 }, { "epoch": 1.31, "learning_rate": 2.814729991136645e-05, "loss": 0.0, "step": 67578 }, { "epoch": 1.31, "learning_rate": 2.8146652951109215e-05, "loss": 0.0, "step": 67580 }, { "epoch": 1.31, "learning_rate": 2.814600599085198e-05, "loss": 0.0, "step": 67582 }, { "epoch": 1.31, "learning_rate": 2.8145359030594753e-05, "loss": 0.0, "step": 67584 }, { "epoch": 1.31, "learning_rate": 2.814471207033752e-05, "loss": 0.0, "step": 67586 }, { "epoch": 1.31, "learning_rate": 2.8144065110080288e-05, "loss": 0.0, "step": 67588 }, { "epoch": 1.31, "learning_rate": 2.8143418149823057e-05, "loss": 0.0, "step": 67590 }, { "epoch": 1.31, "learning_rate": 2.8142771189565826e-05, "loss": 0.0, "step": 67592 }, { "epoch": 1.31, "learning_rate": 2.81421242293086e-05, "loss": 0.0, "step": 67594 }, { "epoch": 1.31, "learning_rate": 2.8141477269051364e-05, "loss": 0.0, "step": 67596 }, { "epoch": 1.31, "learning_rate": 2.814083030879413e-05, "loss": 0.0, "step": 67598 }, { "epoch": 1.31, "learning_rate": 2.8140183348536903e-05, "loss": 0.0068, "step": 67600 }, { "epoch": 1.31, "learning_rate": 2.8139536388279668e-05, "loss": 0.0, "step": 67602 }, { "epoch": 1.31, "learning_rate": 2.8138889428022437e-05, "loss": 0.0, "step": 67604 }, { "epoch": 1.31, "learning_rate": 2.8138242467765206e-05, "loss": 0.0006, "step": 67606 }, { "epoch": 1.31, "learning_rate": 2.8137595507507975e-05, "loss": 0.0, "step": 67608 }, { "epoch": 1.31, "learning_rate": 2.813694854725074e-05, "loss": 0.0, "step": 67610 }, { "epoch": 1.31, "learning_rate": 2.8136301586993514e-05, "loss": 0.0, "step": 67612 }, { "epoch": 1.31, "learning_rate": 2.813565462673628e-05, "loss": 0.0, "step": 67614 }, { "epoch": 1.31, "learning_rate": 2.8135007666479052e-05, "loss": 0.0008, "step": 67616 }, { "epoch": 1.31, "learning_rate": 2.8134360706221818e-05, "loss": 0.0, "step": 67618 }, { "epoch": 1.31, "learning_rate": 2.8133713745964587e-05, "loss": 0.0004, "step": 67620 }, { "epoch": 1.31, "learning_rate": 2.8133066785707356e-05, "loss": 0.0, "step": 67622 }, { "epoch": 1.31, "learning_rate": 2.8132419825450125e-05, "loss": 0.0, "step": 67624 }, { "epoch": 1.31, "learning_rate": 2.813177286519289e-05, "loss": 0.0, "step": 67626 }, { "epoch": 1.31, "learning_rate": 2.8131125904935663e-05, "loss": 0.0, "step": 67628 }, { "epoch": 1.31, "learning_rate": 2.813047894467843e-05, "loss": 0.0, "step": 67630 }, { "epoch": 1.31, "learning_rate": 2.81298319844212e-05, "loss": 0.0, "step": 67632 }, { "epoch": 1.31, "learning_rate": 2.8129185024163967e-05, "loss": 0.0, "step": 67634 }, { "epoch": 1.31, "learning_rate": 2.8128538063906732e-05, "loss": 0.0, "step": 67636 }, { "epoch": 1.31, "learning_rate": 2.8127891103649505e-05, "loss": 0.0, "step": 67638 }, { "epoch": 1.31, "learning_rate": 2.8127244143392274e-05, "loss": 0.0, "step": 67640 }, { "epoch": 1.31, "learning_rate": 2.812659718313504e-05, "loss": 0.0, "step": 67642 }, { "epoch": 1.31, "learning_rate": 2.8125950222877812e-05, "loss": 0.0, "step": 67644 }, { "epoch": 1.31, "learning_rate": 2.8125303262620578e-05, "loss": 0.0, "step": 67646 }, { "epoch": 1.31, "learning_rate": 2.8124656302363344e-05, "loss": 0.0, "step": 67648 }, { "epoch": 1.31, "learning_rate": 2.8124009342106116e-05, "loss": 0.0, "step": 67650 }, { "epoch": 1.31, "learning_rate": 2.8123362381848882e-05, "loss": 0.0, "step": 67652 }, { "epoch": 1.31, "learning_rate": 2.8122715421591654e-05, "loss": 0.0, "step": 67654 }, { "epoch": 1.31, "learning_rate": 2.8122068461334423e-05, "loss": 0.0001, "step": 67656 }, { "epoch": 1.31, "learning_rate": 2.812142150107719e-05, "loss": 0.0, "step": 67658 }, { "epoch": 1.31, "learning_rate": 2.812077454081996e-05, "loss": 0.0, "step": 67660 }, { "epoch": 1.31, "learning_rate": 2.8120127580562727e-05, "loss": 0.0, "step": 67662 }, { "epoch": 1.31, "learning_rate": 2.8119480620305493e-05, "loss": 0.0, "step": 67664 }, { "epoch": 1.31, "learning_rate": 2.8118833660048265e-05, "loss": 0.0, "step": 67666 }, { "epoch": 1.31, "learning_rate": 2.811818669979103e-05, "loss": 0.0001, "step": 67668 }, { "epoch": 1.31, "learning_rate": 2.81175397395338e-05, "loss": 0.0, "step": 67670 }, { "epoch": 1.31, "learning_rate": 2.8116892779276573e-05, "loss": 0.0003, "step": 67672 }, { "epoch": 1.31, "learning_rate": 2.811624581901934e-05, "loss": 0.0, "step": 67674 }, { "epoch": 1.31, "learning_rate": 2.811559885876211e-05, "loss": 0.0, "step": 67676 }, { "epoch": 1.31, "learning_rate": 2.8114951898504876e-05, "loss": 0.0, "step": 67678 }, { "epoch": 1.31, "learning_rate": 2.8114304938247642e-05, "loss": 0.0, "step": 67680 }, { "epoch": 1.31, "learning_rate": 2.8113657977990415e-05, "loss": 0.0, "step": 67682 }, { "epoch": 1.31, "learning_rate": 2.811301101773318e-05, "loss": 0.0, "step": 67684 }, { "epoch": 1.31, "learning_rate": 2.811236405747595e-05, "loss": 0.0, "step": 67686 }, { "epoch": 1.31, "learning_rate": 2.8111717097218722e-05, "loss": 0.0, "step": 67688 }, { "epoch": 1.31, "learning_rate": 2.8111070136961488e-05, "loss": 0.0, "step": 67690 }, { "epoch": 1.31, "learning_rate": 2.811042317670426e-05, "loss": 0.0, "step": 67692 }, { "epoch": 1.31, "learning_rate": 2.8109776216447026e-05, "loss": 0.0, "step": 67694 }, { "epoch": 1.31, "learning_rate": 2.810912925618979e-05, "loss": 0.0, "step": 67696 }, { "epoch": 1.31, "learning_rate": 2.8108482295932564e-05, "loss": 0.0, "step": 67698 }, { "epoch": 1.31, "learning_rate": 2.810783533567533e-05, "loss": 0.0, "step": 67700 }, { "epoch": 1.31, "learning_rate": 2.81071883754181e-05, "loss": 0.0, "step": 67702 }, { "epoch": 1.31, "learning_rate": 2.810654141516087e-05, "loss": 0.0, "step": 67704 }, { "epoch": 1.31, "learning_rate": 2.8105894454903637e-05, "loss": 0.0001, "step": 67706 }, { "epoch": 1.31, "learning_rate": 2.8105247494646403e-05, "loss": 0.0476, "step": 67708 }, { "epoch": 1.31, "learning_rate": 2.8104600534389175e-05, "loss": 0.0, "step": 67710 }, { "epoch": 1.31, "learning_rate": 2.810395357413194e-05, "loss": 0.0, "step": 67712 }, { "epoch": 1.31, "learning_rate": 2.8103306613874713e-05, "loss": 0.0, "step": 67714 }, { "epoch": 1.31, "learning_rate": 2.810265965361748e-05, "loss": 0.0, "step": 67716 }, { "epoch": 1.31, "learning_rate": 2.8102012693360248e-05, "loss": 0.0, "step": 67718 }, { "epoch": 1.31, "learning_rate": 2.810136573310302e-05, "loss": 0.0, "step": 67720 }, { "epoch": 1.31, "learning_rate": 2.8100718772845786e-05, "loss": 0.0001, "step": 67722 }, { "epoch": 1.31, "learning_rate": 2.8100071812588552e-05, "loss": 0.0, "step": 67724 }, { "epoch": 1.31, "learning_rate": 2.8099424852331324e-05, "loss": 0.0, "step": 67726 }, { "epoch": 1.31, "learning_rate": 2.809877789207409e-05, "loss": 0.0011, "step": 67728 }, { "epoch": 1.31, "learning_rate": 2.8098130931816856e-05, "loss": 0.0023, "step": 67730 }, { "epoch": 1.31, "learning_rate": 2.8097483971559628e-05, "loss": 0.0, "step": 67732 }, { "epoch": 1.31, "learning_rate": 2.8096837011302397e-05, "loss": 0.0, "step": 67734 }, { "epoch": 1.31, "learning_rate": 2.809619005104517e-05, "loss": 0.0, "step": 67736 }, { "epoch": 1.31, "learning_rate": 2.8095543090787935e-05, "loss": 0.0, "step": 67738 }, { "epoch": 1.31, "learning_rate": 2.80948961305307e-05, "loss": 0.0, "step": 67740 }, { "epoch": 1.31, "learning_rate": 2.8094249170273474e-05, "loss": 0.0007, "step": 67742 }, { "epoch": 1.31, "learning_rate": 2.809360221001624e-05, "loss": 0.0001, "step": 67744 }, { "epoch": 1.31, "learning_rate": 2.8092955249759005e-05, "loss": 0.0, "step": 67746 }, { "epoch": 1.31, "learning_rate": 2.8092308289501778e-05, "loss": 0.0001, "step": 67748 }, { "epoch": 1.31, "learning_rate": 2.8091661329244547e-05, "loss": 0.0, "step": 67750 }, { "epoch": 1.31, "learning_rate": 2.8091014368987312e-05, "loss": 0.0301, "step": 67752 }, { "epoch": 1.32, "learning_rate": 2.8090367408730085e-05, "loss": 0.0001, "step": 67754 }, { "epoch": 1.32, "learning_rate": 2.808972044847285e-05, "loss": 0.0, "step": 67756 }, { "epoch": 1.32, "learning_rate": 2.8089073488215623e-05, "loss": 0.0, "step": 67758 }, { "epoch": 1.32, "learning_rate": 2.808842652795839e-05, "loss": 0.0, "step": 67760 }, { "epoch": 1.32, "learning_rate": 2.8087779567701154e-05, "loss": 0.0, "step": 67762 }, { "epoch": 1.32, "learning_rate": 2.8087132607443927e-05, "loss": 0.0, "step": 67764 }, { "epoch": 1.32, "learning_rate": 2.8086485647186696e-05, "loss": 0.002, "step": 67766 }, { "epoch": 1.32, "learning_rate": 2.808583868692946e-05, "loss": 0.0, "step": 67768 }, { "epoch": 1.32, "learning_rate": 2.8085191726672234e-05, "loss": 0.0, "step": 67770 }, { "epoch": 1.32, "learning_rate": 2.8084544766415e-05, "loss": 0.0, "step": 67772 }, { "epoch": 1.32, "learning_rate": 2.8083897806157772e-05, "loss": 0.0014, "step": 67774 }, { "epoch": 1.32, "learning_rate": 2.8083250845900538e-05, "loss": 0.0, "step": 67776 }, { "epoch": 1.32, "learning_rate": 2.8082603885643304e-05, "loss": 0.0069, "step": 67778 }, { "epoch": 1.32, "learning_rate": 2.8081956925386076e-05, "loss": 0.0, "step": 67780 }, { "epoch": 1.32, "learning_rate": 2.8081309965128845e-05, "loss": 0.0, "step": 67782 }, { "epoch": 1.32, "learning_rate": 2.808066300487161e-05, "loss": 0.0, "step": 67784 }, { "epoch": 1.32, "learning_rate": 2.8080016044614383e-05, "loss": 0.0021, "step": 67786 }, { "epoch": 1.32, "learning_rate": 2.807936908435715e-05, "loss": 0.0, "step": 67788 }, { "epoch": 1.32, "learning_rate": 2.8078722124099915e-05, "loss": 0.0, "step": 67790 }, { "epoch": 1.32, "learning_rate": 2.8078075163842687e-05, "loss": 0.0, "step": 67792 }, { "epoch": 1.32, "learning_rate": 2.8077428203585453e-05, "loss": 0.0001, "step": 67794 }, { "epoch": 1.32, "learning_rate": 2.8076781243328225e-05, "loss": 0.0049, "step": 67796 }, { "epoch": 1.32, "learning_rate": 2.8076134283070994e-05, "loss": 0.0006, "step": 67798 }, { "epoch": 1.32, "learning_rate": 2.807548732281376e-05, "loss": 0.0, "step": 67800 }, { "epoch": 1.32, "learning_rate": 2.8074840362556533e-05, "loss": 0.0001, "step": 67802 }, { "epoch": 1.32, "learning_rate": 2.80741934022993e-05, "loss": 0.0, "step": 67804 }, { "epoch": 1.32, "learning_rate": 2.8073546442042064e-05, "loss": 0.0, "step": 67806 }, { "epoch": 1.32, "learning_rate": 2.8072899481784836e-05, "loss": 0.0, "step": 67808 }, { "epoch": 1.32, "learning_rate": 2.8072252521527602e-05, "loss": 0.0, "step": 67810 }, { "epoch": 1.32, "learning_rate": 2.807160556127037e-05, "loss": 0.0, "step": 67812 }, { "epoch": 1.32, "learning_rate": 2.807095860101314e-05, "loss": 0.0, "step": 67814 }, { "epoch": 1.32, "learning_rate": 2.807031164075591e-05, "loss": 0.0, "step": 67816 }, { "epoch": 1.32, "learning_rate": 2.8069664680498682e-05, "loss": 0.0001, "step": 67818 }, { "epoch": 1.32, "learning_rate": 2.8069017720241448e-05, "loss": 0.0002, "step": 67820 }, { "epoch": 1.32, "learning_rate": 2.8068370759984213e-05, "loss": 0.0521, "step": 67822 }, { "epoch": 1.32, "learning_rate": 2.8067723799726986e-05, "loss": 0.0, "step": 67824 }, { "epoch": 1.32, "learning_rate": 2.806707683946975e-05, "loss": 0.0, "step": 67826 }, { "epoch": 1.32, "learning_rate": 2.806642987921252e-05, "loss": 0.007, "step": 67828 }, { "epoch": 1.32, "learning_rate": 2.806578291895529e-05, "loss": 0.0, "step": 67830 }, { "epoch": 1.32, "learning_rate": 2.806513595869806e-05, "loss": 0.0, "step": 67832 }, { "epoch": 1.32, "learning_rate": 2.8064488998440824e-05, "loss": 0.0, "step": 67834 }, { "epoch": 1.32, "learning_rate": 2.8063842038183597e-05, "loss": 0.0, "step": 67836 }, { "epoch": 1.32, "learning_rate": 2.8063195077926363e-05, "loss": 0.0022, "step": 67838 }, { "epoch": 1.32, "learning_rate": 2.8062548117669135e-05, "loss": 0.0003, "step": 67840 }, { "epoch": 1.32, "learning_rate": 2.80619011574119e-05, "loss": 0.0, "step": 67842 }, { "epoch": 1.32, "learning_rate": 2.806125419715467e-05, "loss": 0.0, "step": 67844 }, { "epoch": 1.32, "learning_rate": 2.806060723689744e-05, "loss": 0.0, "step": 67846 }, { "epoch": 1.32, "learning_rate": 2.8059960276640208e-05, "loss": 0.0, "step": 67848 }, { "epoch": 1.32, "learning_rate": 2.8059313316382974e-05, "loss": 0.0, "step": 67850 }, { "epoch": 1.32, "learning_rate": 2.8058666356125746e-05, "loss": 0.0, "step": 67852 }, { "epoch": 1.32, "learning_rate": 2.8058019395868512e-05, "loss": 0.0001, "step": 67854 }, { "epoch": 1.32, "learning_rate": 2.8057372435611284e-05, "loss": 0.0, "step": 67856 }, { "epoch": 1.32, "learning_rate": 2.805672547535405e-05, "loss": 0.0001, "step": 67858 }, { "epoch": 1.32, "learning_rate": 2.805607851509682e-05, "loss": 0.0002, "step": 67860 }, { "epoch": 1.32, "learning_rate": 2.8055431554839588e-05, "loss": 0.0, "step": 67862 }, { "epoch": 1.32, "learning_rate": 2.8054784594582357e-05, "loss": 0.0, "step": 67864 }, { "epoch": 1.32, "learning_rate": 2.8054137634325123e-05, "loss": 0.0024, "step": 67866 }, { "epoch": 1.32, "learning_rate": 2.8053490674067895e-05, "loss": 0.0002, "step": 67868 }, { "epoch": 1.32, "learning_rate": 2.805284371381066e-05, "loss": 0.0, "step": 67870 }, { "epoch": 1.32, "learning_rate": 2.8052196753553427e-05, "loss": 0.0, "step": 67872 }, { "epoch": 1.32, "learning_rate": 2.80515497932962e-05, "loss": 0.0, "step": 67874 }, { "epoch": 1.32, "learning_rate": 2.8050902833038965e-05, "loss": 0.003, "step": 67876 }, { "epoch": 1.32, "learning_rate": 2.8050255872781737e-05, "loss": 0.0, "step": 67878 }, { "epoch": 1.32, "learning_rate": 2.8049608912524507e-05, "loss": 0.0, "step": 67880 }, { "epoch": 1.32, "learning_rate": 2.8048961952267272e-05, "loss": 0.0, "step": 67882 }, { "epoch": 1.32, "learning_rate": 2.8048314992010045e-05, "loss": 0.0, "step": 67884 }, { "epoch": 1.32, "learning_rate": 2.804766803175281e-05, "loss": 0.0001, "step": 67886 }, { "epoch": 1.32, "learning_rate": 2.8047021071495576e-05, "loss": 0.0, "step": 67888 }, { "epoch": 1.32, "learning_rate": 2.804637411123835e-05, "loss": 0.0, "step": 67890 }, { "epoch": 1.32, "learning_rate": 2.8045727150981114e-05, "loss": 0.0036, "step": 67892 }, { "epoch": 1.32, "learning_rate": 2.8045080190723883e-05, "loss": 0.0, "step": 67894 }, { "epoch": 1.32, "learning_rate": 2.8044433230466656e-05, "loss": 0.0, "step": 67896 }, { "epoch": 1.32, "learning_rate": 2.804378627020942e-05, "loss": 0.0, "step": 67898 }, { "epoch": 1.32, "learning_rate": 2.8043139309952194e-05, "loss": 0.0, "step": 67900 }, { "epoch": 1.32, "learning_rate": 2.804249234969496e-05, "loss": 0.002, "step": 67902 }, { "epoch": 1.32, "learning_rate": 2.8041845389437725e-05, "loss": 0.0, "step": 67904 }, { "epoch": 1.32, "learning_rate": 2.8041198429180498e-05, "loss": 0.0005, "step": 67906 }, { "epoch": 1.32, "learning_rate": 2.8040551468923264e-05, "loss": 0.0, "step": 67908 }, { "epoch": 1.32, "learning_rate": 2.8039904508666033e-05, "loss": 0.0, "step": 67910 }, { "epoch": 1.32, "learning_rate": 2.8039257548408805e-05, "loss": 0.0206, "step": 67912 }, { "epoch": 1.32, "learning_rate": 2.803861058815157e-05, "loss": 0.0, "step": 67914 }, { "epoch": 1.32, "learning_rate": 2.8037963627894343e-05, "loss": 0.0, "step": 67916 }, { "epoch": 1.32, "learning_rate": 2.803731666763711e-05, "loss": 0.0089, "step": 67918 }, { "epoch": 1.32, "learning_rate": 2.8036669707379875e-05, "loss": 0.0, "step": 67920 }, { "epoch": 1.32, "learning_rate": 2.8036022747122647e-05, "loss": 0.0001, "step": 67922 }, { "epoch": 1.32, "learning_rate": 2.8035375786865413e-05, "loss": 0.0, "step": 67924 }, { "epoch": 1.32, "learning_rate": 2.8034728826608182e-05, "loss": 0.0002, "step": 67926 }, { "epoch": 1.32, "learning_rate": 2.8034081866350954e-05, "loss": 0.0, "step": 67928 }, { "epoch": 1.32, "learning_rate": 2.803343490609372e-05, "loss": 0.0, "step": 67930 }, { "epoch": 1.32, "learning_rate": 2.8032787945836486e-05, "loss": 0.0, "step": 67932 }, { "epoch": 1.32, "learning_rate": 2.8032140985579258e-05, "loss": 0.0042, "step": 67934 }, { "epoch": 1.32, "learning_rate": 2.8031494025322024e-05, "loss": 0.0, "step": 67936 }, { "epoch": 1.32, "learning_rate": 2.8030847065064796e-05, "loss": 0.0006, "step": 67938 }, { "epoch": 1.32, "learning_rate": 2.8030200104807562e-05, "loss": 0.0, "step": 67940 }, { "epoch": 1.32, "learning_rate": 2.802955314455033e-05, "loss": 0.0, "step": 67942 }, { "epoch": 1.32, "learning_rate": 2.8028906184293104e-05, "loss": 0.0, "step": 67944 }, { "epoch": 1.32, "learning_rate": 2.802825922403587e-05, "loss": 0.0, "step": 67946 }, { "epoch": 1.32, "learning_rate": 2.8027612263778635e-05, "loss": 0.0, "step": 67948 }, { "epoch": 1.32, "learning_rate": 2.8026965303521408e-05, "loss": 0.0, "step": 67950 }, { "epoch": 1.32, "learning_rate": 2.8026318343264173e-05, "loss": 0.0, "step": 67952 }, { "epoch": 1.32, "learning_rate": 2.802567138300694e-05, "loss": 0.0095, "step": 67954 }, { "epoch": 1.32, "learning_rate": 2.802502442274971e-05, "loss": 0.0002, "step": 67956 }, { "epoch": 1.32, "learning_rate": 2.802437746249248e-05, "loss": 0.0, "step": 67958 }, { "epoch": 1.32, "learning_rate": 2.8023730502235253e-05, "loss": 0.0, "step": 67960 }, { "epoch": 1.32, "learning_rate": 2.802308354197802e-05, "loss": 0.0, "step": 67962 }, { "epoch": 1.32, "learning_rate": 2.8022436581720784e-05, "loss": 0.0002, "step": 67964 }, { "epoch": 1.32, "learning_rate": 2.8021789621463557e-05, "loss": 0.0, "step": 67966 }, { "epoch": 1.32, "learning_rate": 2.8021142661206323e-05, "loss": 0.0, "step": 67968 }, { "epoch": 1.32, "learning_rate": 2.8020495700949088e-05, "loss": 0.0, "step": 67970 }, { "epoch": 1.32, "learning_rate": 2.801984874069186e-05, "loss": 0.0, "step": 67972 }, { "epoch": 1.32, "learning_rate": 2.801920178043463e-05, "loss": 0.0, "step": 67974 }, { "epoch": 1.32, "learning_rate": 2.8018554820177396e-05, "loss": 0.0, "step": 67976 }, { "epoch": 1.32, "learning_rate": 2.8017907859920168e-05, "loss": 0.006, "step": 67978 }, { "epoch": 1.32, "learning_rate": 2.8017260899662934e-05, "loss": 0.0, "step": 67980 }, { "epoch": 1.32, "learning_rate": 2.8016613939405706e-05, "loss": 0.0, "step": 67982 }, { "epoch": 1.32, "learning_rate": 2.8015966979148472e-05, "loss": 0.0, "step": 67984 }, { "epoch": 1.32, "learning_rate": 2.8015320018891238e-05, "loss": 0.0, "step": 67986 }, { "epoch": 1.32, "learning_rate": 2.801467305863401e-05, "loss": 0.0, "step": 67988 }, { "epoch": 1.32, "learning_rate": 2.801402609837678e-05, "loss": 0.0, "step": 67990 }, { "epoch": 1.32, "learning_rate": 2.8013379138119545e-05, "loss": 0.0, "step": 67992 }, { "epoch": 1.32, "learning_rate": 2.8012732177862317e-05, "loss": 0.0043, "step": 67994 }, { "epoch": 1.32, "learning_rate": 2.8012085217605083e-05, "loss": 0.0, "step": 67996 }, { "epoch": 1.32, "learning_rate": 2.8011438257347855e-05, "loss": 0.0128, "step": 67998 }, { "epoch": 1.32, "learning_rate": 2.801079129709062e-05, "loss": 0.0021, "step": 68000 }, { "epoch": 1.32, "learning_rate": 2.8010144336833387e-05, "loss": 0.0, "step": 68002 }, { "epoch": 1.32, "learning_rate": 2.800949737657616e-05, "loss": 0.0, "step": 68004 }, { "epoch": 1.32, "learning_rate": 2.800885041631893e-05, "loss": 0.0003, "step": 68006 }, { "epoch": 1.32, "learning_rate": 2.8008203456061694e-05, "loss": 0.0, "step": 68008 }, { "epoch": 1.32, "learning_rate": 2.8007556495804467e-05, "loss": 0.0, "step": 68010 }, { "epoch": 1.32, "learning_rate": 2.8006909535547232e-05, "loss": 0.0001, "step": 68012 }, { "epoch": 1.32, "learning_rate": 2.8006262575289998e-05, "loss": 0.0, "step": 68014 }, { "epoch": 1.32, "learning_rate": 2.800561561503277e-05, "loss": 0.0007, "step": 68016 }, { "epoch": 1.32, "learning_rate": 2.8004968654775536e-05, "loss": 0.0, "step": 68018 }, { "epoch": 1.32, "learning_rate": 2.800432169451831e-05, "loss": 0.0001, "step": 68020 }, { "epoch": 1.32, "learning_rate": 2.8003674734261078e-05, "loss": 0.0002, "step": 68022 }, { "epoch": 1.32, "learning_rate": 2.8003027774003843e-05, "loss": 0.0, "step": 68024 }, { "epoch": 1.32, "learning_rate": 2.8002380813746616e-05, "loss": 0.0, "step": 68026 }, { "epoch": 1.32, "learning_rate": 2.800173385348938e-05, "loss": 0.0, "step": 68028 }, { "epoch": 1.32, "learning_rate": 2.8001086893232147e-05, "loss": 0.0003, "step": 68030 }, { "epoch": 1.32, "learning_rate": 2.800043993297492e-05, "loss": 0.0006, "step": 68032 }, { "epoch": 1.32, "learning_rate": 2.7999792972717685e-05, "loss": 0.0, "step": 68034 }, { "epoch": 1.32, "learning_rate": 2.7999146012460455e-05, "loss": 0.0, "step": 68036 }, { "epoch": 1.32, "learning_rate": 2.7998499052203227e-05, "loss": 0.0, "step": 68038 }, { "epoch": 1.32, "learning_rate": 2.7997852091945993e-05, "loss": 0.0, "step": 68040 }, { "epoch": 1.32, "learning_rate": 2.7997205131688765e-05, "loss": 0.0, "step": 68042 }, { "epoch": 1.32, "learning_rate": 2.799655817143153e-05, "loss": 0.0, "step": 68044 }, { "epoch": 1.32, "learning_rate": 2.7995911211174297e-05, "loss": 0.0006, "step": 68046 }, { "epoch": 1.32, "learning_rate": 2.799526425091707e-05, "loss": 0.0, "step": 68048 }, { "epoch": 1.32, "learning_rate": 2.7994617290659835e-05, "loss": 0.0, "step": 68050 }, { "epoch": 1.32, "learning_rate": 2.7993970330402604e-05, "loss": 0.0042, "step": 68052 }, { "epoch": 1.32, "learning_rate": 2.7993323370145373e-05, "loss": 0.0002, "step": 68054 }, { "epoch": 1.32, "learning_rate": 2.7992676409888142e-05, "loss": 0.0, "step": 68056 }, { "epoch": 1.32, "learning_rate": 2.7992029449630914e-05, "loss": 0.0, "step": 68058 }, { "epoch": 1.32, "learning_rate": 2.799138248937368e-05, "loss": 0.0003, "step": 68060 }, { "epoch": 1.32, "learning_rate": 2.7990735529116446e-05, "loss": 0.0, "step": 68062 }, { "epoch": 1.32, "learning_rate": 2.7990088568859218e-05, "loss": 0.0, "step": 68064 }, { "epoch": 1.32, "learning_rate": 2.7989441608601984e-05, "loss": 0.0036, "step": 68066 }, { "epoch": 1.32, "learning_rate": 2.7988794648344753e-05, "loss": 0.0, "step": 68068 }, { "epoch": 1.32, "learning_rate": 2.7988147688087522e-05, "loss": 0.0, "step": 68070 }, { "epoch": 1.32, "learning_rate": 2.798750072783029e-05, "loss": 0.0, "step": 68072 }, { "epoch": 1.32, "learning_rate": 2.7986853767573057e-05, "loss": 0.0, "step": 68074 }, { "epoch": 1.32, "learning_rate": 2.798620680731583e-05, "loss": 0.0, "step": 68076 }, { "epoch": 1.32, "learning_rate": 2.7985559847058595e-05, "loss": 0.0, "step": 68078 }, { "epoch": 1.32, "learning_rate": 2.7984912886801368e-05, "loss": 0.0, "step": 68080 }, { "epoch": 1.32, "learning_rate": 2.7984265926544133e-05, "loss": 0.0, "step": 68082 }, { "epoch": 1.32, "learning_rate": 2.7983618966286902e-05, "loss": 0.0, "step": 68084 }, { "epoch": 1.32, "learning_rate": 2.798297200602967e-05, "loss": 0.0, "step": 68086 }, { "epoch": 1.32, "learning_rate": 2.798232504577244e-05, "loss": 0.0, "step": 68088 }, { "epoch": 1.32, "learning_rate": 2.7981678085515206e-05, "loss": 0.0, "step": 68090 }, { "epoch": 1.32, "learning_rate": 2.798103112525798e-05, "loss": 0.0, "step": 68092 }, { "epoch": 1.32, "learning_rate": 2.7980384165000744e-05, "loss": 0.0, "step": 68094 }, { "epoch": 1.32, "learning_rate": 2.797973720474351e-05, "loss": 0.0, "step": 68096 }, { "epoch": 1.32, "learning_rate": 2.7979090244486283e-05, "loss": 0.0101, "step": 68098 }, { "epoch": 1.32, "learning_rate": 2.797844328422905e-05, "loss": 0.0, "step": 68100 }, { "epoch": 1.32, "learning_rate": 2.797779632397182e-05, "loss": 0.0, "step": 68102 }, { "epoch": 1.32, "learning_rate": 2.797714936371459e-05, "loss": 0.0, "step": 68104 }, { "epoch": 1.32, "learning_rate": 2.7976502403457356e-05, "loss": 0.0, "step": 68106 }, { "epoch": 1.32, "learning_rate": 2.7975855443200128e-05, "loss": 0.0009, "step": 68108 }, { "epoch": 1.32, "learning_rate": 2.7975208482942894e-05, "loss": 0.0, "step": 68110 }, { "epoch": 1.32, "learning_rate": 2.797456152268566e-05, "loss": 0.0, "step": 68112 }, { "epoch": 1.32, "learning_rate": 2.7973914562428432e-05, "loss": 0.0048, "step": 68114 }, { "epoch": 1.32, "learning_rate": 2.7973267602171198e-05, "loss": 0.0001, "step": 68116 }, { "epoch": 1.32, "learning_rate": 2.7972620641913967e-05, "loss": 0.0005, "step": 68118 }, { "epoch": 1.32, "learning_rate": 2.797197368165674e-05, "loss": 0.0, "step": 68120 }, { "epoch": 1.32, "learning_rate": 2.7971326721399505e-05, "loss": 0.0, "step": 68122 }, { "epoch": 1.32, "learning_rate": 2.7970679761142277e-05, "loss": 0.0, "step": 68124 }, { "epoch": 1.32, "learning_rate": 2.7970032800885043e-05, "loss": 0.0002, "step": 68126 }, { "epoch": 1.32, "learning_rate": 2.796938584062781e-05, "loss": 0.0, "step": 68128 }, { "epoch": 1.32, "learning_rate": 2.796873888037058e-05, "loss": 0.0, "step": 68130 }, { "epoch": 1.32, "learning_rate": 2.7968091920113347e-05, "loss": 0.0, "step": 68132 }, { "epoch": 1.32, "learning_rate": 2.7967444959856116e-05, "loss": 0.0042, "step": 68134 }, { "epoch": 1.32, "learning_rate": 2.796679799959889e-05, "loss": 0.0, "step": 68136 }, { "epoch": 1.32, "learning_rate": 2.7966151039341654e-05, "loss": 0.0, "step": 68138 }, { "epoch": 1.32, "learning_rate": 2.7965504079084427e-05, "loss": 0.0, "step": 68140 }, { "epoch": 1.32, "learning_rate": 2.7964857118827192e-05, "loss": 0.0, "step": 68142 }, { "epoch": 1.32, "learning_rate": 2.7964210158569958e-05, "loss": 0.0, "step": 68144 }, { "epoch": 1.32, "learning_rate": 2.796356319831273e-05, "loss": 0.0, "step": 68146 }, { "epoch": 1.32, "learning_rate": 2.7962916238055496e-05, "loss": 0.0001, "step": 68148 }, { "epoch": 1.32, "learning_rate": 2.7962269277798265e-05, "loss": 0.0, "step": 68150 }, { "epoch": 1.32, "learning_rate": 2.7961622317541038e-05, "loss": 0.0003, "step": 68152 }, { "epoch": 1.32, "learning_rate": 2.7960975357283803e-05, "loss": 0.0, "step": 68154 }, { "epoch": 1.32, "learning_rate": 2.796032839702657e-05, "loss": 0.0, "step": 68156 }, { "epoch": 1.32, "learning_rate": 2.795968143676934e-05, "loss": 0.0008, "step": 68158 }, { "epoch": 1.32, "learning_rate": 2.7959034476512107e-05, "loss": 0.0012, "step": 68160 }, { "epoch": 1.32, "learning_rate": 2.795838751625488e-05, "loss": 0.0, "step": 68162 }, { "epoch": 1.32, "learning_rate": 2.7957740555997645e-05, "loss": 0.0, "step": 68164 }, { "epoch": 1.32, "learning_rate": 2.7957093595740414e-05, "loss": 0.0004, "step": 68166 }, { "epoch": 1.32, "learning_rate": 2.7956446635483187e-05, "loss": 0.0025, "step": 68168 }, { "epoch": 1.32, "learning_rate": 2.7955799675225953e-05, "loss": 0.0, "step": 68170 }, { "epoch": 1.32, "learning_rate": 2.795515271496872e-05, "loss": 0.0, "step": 68172 }, { "epoch": 1.32, "learning_rate": 2.795450575471149e-05, "loss": 0.0, "step": 68174 }, { "epoch": 1.32, "learning_rate": 2.7953858794454257e-05, "loss": 0.0, "step": 68176 }, { "epoch": 1.32, "learning_rate": 2.7953211834197026e-05, "loss": 0.0, "step": 68178 }, { "epoch": 1.32, "learning_rate": 2.7952564873939795e-05, "loss": 0.0, "step": 68180 }, { "epoch": 1.32, "learning_rate": 2.7951917913682564e-05, "loss": 0.0022, "step": 68182 }, { "epoch": 1.32, "learning_rate": 2.7951270953425336e-05, "loss": 0.0, "step": 68184 }, { "epoch": 1.32, "learning_rate": 2.7950623993168102e-05, "loss": 0.0, "step": 68186 }, { "epoch": 1.32, "learning_rate": 2.7949977032910868e-05, "loss": 0.0001, "step": 68188 }, { "epoch": 1.32, "learning_rate": 2.794933007265364e-05, "loss": 0.0, "step": 68190 }, { "epoch": 1.32, "learning_rate": 2.7948683112396406e-05, "loss": 0.0005, "step": 68192 }, { "epoch": 1.32, "learning_rate": 2.794803615213917e-05, "loss": 0.0, "step": 68194 }, { "epoch": 1.32, "learning_rate": 2.7947389191881944e-05, "loss": 0.0, "step": 68196 }, { "epoch": 1.32, "learning_rate": 2.7946742231624713e-05, "loss": 0.0, "step": 68198 }, { "epoch": 1.32, "learning_rate": 2.7946095271367486e-05, "loss": 0.0005, "step": 68200 }, { "epoch": 1.32, "learning_rate": 2.794544831111025e-05, "loss": 0.0, "step": 68202 }, { "epoch": 1.32, "learning_rate": 2.7944801350853017e-05, "loss": 0.0, "step": 68204 }, { "epoch": 1.32, "learning_rate": 2.794415439059579e-05, "loss": 0.0, "step": 68206 }, { "epoch": 1.32, "learning_rate": 2.7943507430338555e-05, "loss": 0.0, "step": 68208 }, { "epoch": 1.32, "learning_rate": 2.794286047008132e-05, "loss": 0.0, "step": 68210 }, { "epoch": 1.32, "learning_rate": 2.7942213509824093e-05, "loss": 0.0, "step": 68212 }, { "epoch": 1.32, "learning_rate": 2.7941566549566862e-05, "loss": 0.0001, "step": 68214 }, { "epoch": 1.32, "learning_rate": 2.7940919589309628e-05, "loss": 0.0, "step": 68216 }, { "epoch": 1.32, "learning_rate": 2.79402726290524e-05, "loss": 0.0, "step": 68218 }, { "epoch": 1.32, "learning_rate": 2.7939625668795166e-05, "loss": 0.0, "step": 68220 }, { "epoch": 1.32, "learning_rate": 2.793897870853794e-05, "loss": 0.0026, "step": 68222 }, { "epoch": 1.32, "learning_rate": 2.7938331748280704e-05, "loss": 0.0005, "step": 68224 }, { "epoch": 1.32, "learning_rate": 2.793768478802347e-05, "loss": 0.0, "step": 68226 }, { "epoch": 1.32, "learning_rate": 2.7937037827766243e-05, "loss": 0.0, "step": 68228 }, { "epoch": 1.32, "learning_rate": 2.793639086750901e-05, "loss": 0.0, "step": 68230 }, { "epoch": 1.32, "learning_rate": 2.7935743907251777e-05, "loss": 0.0, "step": 68232 }, { "epoch": 1.32, "learning_rate": 2.793509694699455e-05, "loss": 0.0005, "step": 68234 }, { "epoch": 1.32, "learning_rate": 2.7934449986737316e-05, "loss": 0.0, "step": 68236 }, { "epoch": 1.32, "learning_rate": 2.793380302648008e-05, "loss": 0.0, "step": 68238 }, { "epoch": 1.32, "learning_rate": 2.7933156066222854e-05, "loss": 0.0, "step": 68240 }, { "epoch": 1.32, "learning_rate": 2.793250910596562e-05, "loss": 0.0002, "step": 68242 }, { "epoch": 1.32, "learning_rate": 2.7931862145708392e-05, "loss": 0.0014, "step": 68244 }, { "epoch": 1.32, "learning_rate": 2.793121518545116e-05, "loss": 0.0, "step": 68246 }, { "epoch": 1.32, "learning_rate": 2.7930568225193927e-05, "loss": 0.0003, "step": 68248 }, { "epoch": 1.32, "learning_rate": 2.79299212649367e-05, "loss": 0.0, "step": 68250 }, { "epoch": 1.32, "learning_rate": 2.7929274304679465e-05, "loss": 0.0, "step": 68252 }, { "epoch": 1.32, "learning_rate": 2.792862734442223e-05, "loss": 0.0, "step": 68254 }, { "epoch": 1.32, "learning_rate": 2.7927980384165003e-05, "loss": 0.0, "step": 68256 }, { "epoch": 1.32, "learning_rate": 2.792733342390777e-05, "loss": 0.0011, "step": 68258 }, { "epoch": 1.32, "learning_rate": 2.7926686463650538e-05, "loss": 0.0, "step": 68260 }, { "epoch": 1.32, "learning_rate": 2.792603950339331e-05, "loss": 0.0, "step": 68262 }, { "epoch": 1.32, "learning_rate": 2.7925392543136076e-05, "loss": 0.0, "step": 68264 }, { "epoch": 1.32, "learning_rate": 2.792474558287885e-05, "loss": 0.0, "step": 68266 }, { "epoch": 1.33, "learning_rate": 2.7924098622621614e-05, "loss": 0.0018, "step": 68268 }, { "epoch": 1.33, "learning_rate": 2.792345166236438e-05, "loss": 0.0, "step": 68270 }, { "epoch": 1.33, "learning_rate": 2.7922804702107152e-05, "loss": 0.0, "step": 68272 }, { "epoch": 1.33, "learning_rate": 2.7922157741849918e-05, "loss": 0.0, "step": 68274 }, { "epoch": 1.33, "learning_rate": 2.7921510781592687e-05, "loss": 0.0, "step": 68276 }, { "epoch": 1.33, "learning_rate": 2.792086382133546e-05, "loss": 0.0003, "step": 68278 }, { "epoch": 1.33, "learning_rate": 2.7920216861078225e-05, "loss": 0.0, "step": 68280 }, { "epoch": 1.33, "learning_rate": 2.7919569900820998e-05, "loss": 0.0, "step": 68282 }, { "epoch": 1.33, "learning_rate": 2.7918922940563763e-05, "loss": 0.0001, "step": 68284 }, { "epoch": 1.33, "learning_rate": 2.791827598030653e-05, "loss": 0.0, "step": 68286 }, { "epoch": 1.33, "learning_rate": 2.79176290200493e-05, "loss": 0.0, "step": 68288 }, { "epoch": 1.33, "learning_rate": 2.7916982059792067e-05, "loss": 0.0, "step": 68290 }, { "epoch": 1.33, "learning_rate": 2.7916335099534836e-05, "loss": 0.0, "step": 68292 }, { "epoch": 1.33, "learning_rate": 2.7915688139277605e-05, "loss": 0.0, "step": 68294 }, { "epoch": 1.33, "learning_rate": 2.7915041179020374e-05, "loss": 0.0, "step": 68296 }, { "epoch": 1.33, "learning_rate": 2.791439421876314e-05, "loss": 0.0, "step": 68298 }, { "epoch": 1.33, "learning_rate": 2.7913747258505913e-05, "loss": 0.0, "step": 68300 }, { "epoch": 1.33, "learning_rate": 2.791310029824868e-05, "loss": 0.0, "step": 68302 }, { "epoch": 1.33, "learning_rate": 2.791245333799145e-05, "loss": 0.0, "step": 68304 }, { "epoch": 1.33, "learning_rate": 2.7911806377734217e-05, "loss": 0.0, "step": 68306 }, { "epoch": 1.33, "learning_rate": 2.7911159417476986e-05, "loss": 0.0, "step": 68308 }, { "epoch": 1.33, "learning_rate": 2.7910512457219755e-05, "loss": 0.0, "step": 68310 }, { "epoch": 1.33, "learning_rate": 2.7909865496962524e-05, "loss": 0.0, "step": 68312 }, { "epoch": 1.33, "learning_rate": 2.790921853670529e-05, "loss": 0.0, "step": 68314 }, { "epoch": 1.33, "learning_rate": 2.7908571576448062e-05, "loss": 0.0, "step": 68316 }, { "epoch": 1.33, "learning_rate": 2.7907924616190828e-05, "loss": 0.0, "step": 68318 }, { "epoch": 1.33, "learning_rate": 2.7907277655933593e-05, "loss": 0.0, "step": 68320 }, { "epoch": 1.33, "learning_rate": 2.7906630695676366e-05, "loss": 0.0, "step": 68322 }, { "epoch": 1.33, "learning_rate": 2.7905983735419135e-05, "loss": 0.0004, "step": 68324 }, { "epoch": 1.33, "learning_rate": 2.7905336775161904e-05, "loss": 0.0, "step": 68326 }, { "epoch": 1.33, "learning_rate": 2.7904689814904673e-05, "loss": 0.0001, "step": 68328 }, { "epoch": 1.33, "learning_rate": 2.790404285464744e-05, "loss": 0.0, "step": 68330 }, { "epoch": 1.33, "learning_rate": 2.790339589439021e-05, "loss": 0.0, "step": 68332 }, { "epoch": 1.33, "learning_rate": 2.7902748934132977e-05, "loss": 0.0, "step": 68334 }, { "epoch": 1.33, "learning_rate": 2.7902101973875743e-05, "loss": 0.0032, "step": 68336 }, { "epoch": 1.33, "learning_rate": 2.7901455013618515e-05, "loss": 0.0, "step": 68338 }, { "epoch": 1.33, "learning_rate": 2.7900808053361284e-05, "loss": 0.0004, "step": 68340 }, { "epoch": 1.33, "learning_rate": 2.7900161093104053e-05, "loss": 0.0, "step": 68342 }, { "epoch": 1.33, "learning_rate": 2.7899514132846822e-05, "loss": 0.0001, "step": 68344 }, { "epoch": 1.33, "learning_rate": 2.7898867172589588e-05, "loss": 0.0, "step": 68346 }, { "epoch": 1.33, "learning_rate": 2.789822021233236e-05, "loss": 0.0, "step": 68348 }, { "epoch": 1.33, "learning_rate": 2.7897573252075126e-05, "loss": 0.0, "step": 68350 }, { "epoch": 1.33, "learning_rate": 2.7896926291817892e-05, "loss": 0.0169, "step": 68352 }, { "epoch": 1.33, "learning_rate": 2.7896279331560664e-05, "loss": 0.0019, "step": 68354 }, { "epoch": 1.33, "learning_rate": 2.7895632371303433e-05, "loss": 0.0, "step": 68356 }, { "epoch": 1.33, "learning_rate": 2.78949854110462e-05, "loss": 0.0, "step": 68358 }, { "epoch": 1.33, "learning_rate": 2.789433845078897e-05, "loss": 0.0, "step": 68360 }, { "epoch": 1.33, "learning_rate": 2.7893691490531737e-05, "loss": 0.0074, "step": 68362 }, { "epoch": 1.33, "learning_rate": 2.789304453027451e-05, "loss": 0.0, "step": 68364 }, { "epoch": 1.33, "learning_rate": 2.7892397570017275e-05, "loss": 0.0, "step": 68366 }, { "epoch": 1.33, "learning_rate": 2.789175060976004e-05, "loss": 0.0, "step": 68368 }, { "epoch": 1.33, "learning_rate": 2.7891103649502814e-05, "loss": 0.0, "step": 68370 }, { "epoch": 1.33, "learning_rate": 2.789045668924558e-05, "loss": 0.0023, "step": 68372 }, { "epoch": 1.33, "learning_rate": 2.788980972898835e-05, "loss": 0.0, "step": 68374 }, { "epoch": 1.33, "learning_rate": 2.788916276873112e-05, "loss": 0.0, "step": 68376 }, { "epoch": 1.33, "learning_rate": 2.7888515808473887e-05, "loss": 0.025, "step": 68378 }, { "epoch": 1.33, "learning_rate": 2.7887868848216652e-05, "loss": 0.0, "step": 68380 }, { "epoch": 1.33, "learning_rate": 2.7887221887959425e-05, "loss": 0.0062, "step": 68382 }, { "epoch": 1.33, "learning_rate": 2.788657492770219e-05, "loss": 0.0, "step": 68384 }, { "epoch": 1.33, "learning_rate": 2.7885927967444963e-05, "loss": 0.0, "step": 68386 }, { "epoch": 1.33, "learning_rate": 2.788528100718773e-05, "loss": 0.0, "step": 68388 }, { "epoch": 1.33, "learning_rate": 2.7884634046930498e-05, "loss": 0.0, "step": 68390 }, { "epoch": 1.33, "learning_rate": 2.788398708667327e-05, "loss": 0.0, "step": 68392 }, { "epoch": 1.33, "learning_rate": 2.7883340126416036e-05, "loss": 0.0, "step": 68394 }, { "epoch": 1.33, "learning_rate": 2.78826931661588e-05, "loss": 0.0003, "step": 68396 }, { "epoch": 1.33, "learning_rate": 2.7882046205901574e-05, "loss": 0.0, "step": 68398 }, { "epoch": 1.33, "learning_rate": 2.788139924564434e-05, "loss": 0.004, "step": 68400 }, { "epoch": 1.33, "learning_rate": 2.788075228538711e-05, "loss": 0.0003, "step": 68402 }, { "epoch": 1.33, "learning_rate": 2.7880105325129878e-05, "loss": 0.0, "step": 68404 }, { "epoch": 1.33, "learning_rate": 2.7879458364872647e-05, "loss": 0.0, "step": 68406 }, { "epoch": 1.33, "learning_rate": 2.787881140461542e-05, "loss": 0.0, "step": 68408 }, { "epoch": 1.33, "learning_rate": 2.7878164444358185e-05, "loss": 0.0, "step": 68410 }, { "epoch": 1.33, "learning_rate": 2.787751748410095e-05, "loss": 0.0015, "step": 68412 }, { "epoch": 1.33, "learning_rate": 2.7876870523843723e-05, "loss": 0.0, "step": 68414 }, { "epoch": 1.33, "learning_rate": 2.787622356358649e-05, "loss": 0.0, "step": 68416 }, { "epoch": 1.33, "learning_rate": 2.7875576603329258e-05, "loss": 0.0, "step": 68418 }, { "epoch": 1.33, "learning_rate": 2.7874929643072027e-05, "loss": 0.0, "step": 68420 }, { "epoch": 1.33, "learning_rate": 2.7874282682814796e-05, "loss": 0.0, "step": 68422 }, { "epoch": 1.33, "learning_rate": 2.787363572255757e-05, "loss": 0.0, "step": 68424 }, { "epoch": 1.33, "learning_rate": 2.7872988762300334e-05, "loss": 0.0045, "step": 68426 }, { "epoch": 1.33, "learning_rate": 2.78723418020431e-05, "loss": 0.0, "step": 68428 }, { "epoch": 1.33, "learning_rate": 2.7871694841785873e-05, "loss": 0.0007, "step": 68430 }, { "epoch": 1.33, "learning_rate": 2.787104788152864e-05, "loss": 0.0, "step": 68432 }, { "epoch": 1.33, "learning_rate": 2.7870400921271404e-05, "loss": 0.0, "step": 68434 }, { "epoch": 1.33, "learning_rate": 2.7869753961014177e-05, "loss": 0.0, "step": 68436 }, { "epoch": 1.33, "learning_rate": 2.7869107000756946e-05, "loss": 0.0, "step": 68438 }, { "epoch": 1.33, "learning_rate": 2.786846004049971e-05, "loss": 0.0, "step": 68440 }, { "epoch": 1.33, "learning_rate": 2.7867813080242484e-05, "loss": 0.0, "step": 68442 }, { "epoch": 1.33, "learning_rate": 2.786716611998525e-05, "loss": 0.0, "step": 68444 }, { "epoch": 1.33, "learning_rate": 2.7866519159728022e-05, "loss": 0.0, "step": 68446 }, { "epoch": 1.33, "learning_rate": 2.7865872199470788e-05, "loss": 0.0002, "step": 68448 }, { "epoch": 1.33, "learning_rate": 2.7865225239213553e-05, "loss": 0.0, "step": 68450 }, { "epoch": 1.33, "learning_rate": 2.7864578278956326e-05, "loss": 0.0, "step": 68452 }, { "epoch": 1.33, "learning_rate": 2.7863931318699095e-05, "loss": 0.0, "step": 68454 }, { "epoch": 1.33, "learning_rate": 2.786328435844186e-05, "loss": 0.0021, "step": 68456 }, { "epoch": 1.33, "learning_rate": 2.7862637398184633e-05, "loss": 0.0, "step": 68458 }, { "epoch": 1.33, "learning_rate": 2.78619904379274e-05, "loss": 0.0043, "step": 68460 }, { "epoch": 1.33, "learning_rate": 2.7861343477670164e-05, "loss": 0.0, "step": 68462 }, { "epoch": 1.33, "learning_rate": 2.7860696517412937e-05, "loss": 0.0007, "step": 68464 }, { "epoch": 1.33, "learning_rate": 2.7860049557155703e-05, "loss": 0.009, "step": 68466 }, { "epoch": 1.33, "learning_rate": 2.7859402596898475e-05, "loss": 0.0018, "step": 68468 }, { "epoch": 1.33, "learning_rate": 2.7858755636641244e-05, "loss": 0.0, "step": 68470 }, { "epoch": 1.33, "learning_rate": 2.785810867638401e-05, "loss": 0.0, "step": 68472 }, { "epoch": 1.33, "learning_rate": 2.7857461716126782e-05, "loss": 0.0, "step": 68474 }, { "epoch": 1.33, "learning_rate": 2.7856814755869548e-05, "loss": 0.0036, "step": 68476 }, { "epoch": 1.33, "learning_rate": 2.7856167795612314e-05, "loss": 0.0, "step": 68478 }, { "epoch": 1.33, "learning_rate": 2.7855520835355086e-05, "loss": 0.0, "step": 68480 }, { "epoch": 1.33, "learning_rate": 2.7854873875097852e-05, "loss": 0.0, "step": 68482 }, { "epoch": 1.33, "learning_rate": 2.7854226914840624e-05, "loss": 0.0, "step": 68484 }, { "epoch": 1.33, "learning_rate": 2.7853579954583393e-05, "loss": 0.0, "step": 68486 }, { "epoch": 1.33, "learning_rate": 2.785293299432616e-05, "loss": 0.0, "step": 68488 }, { "epoch": 1.33, "learning_rate": 2.785228603406893e-05, "loss": 0.0, "step": 68490 }, { "epoch": 1.33, "learning_rate": 2.7851639073811697e-05, "loss": 0.0, "step": 68492 }, { "epoch": 1.33, "learning_rate": 2.7850992113554463e-05, "loss": 0.0001, "step": 68494 }, { "epoch": 1.33, "learning_rate": 2.7850345153297235e-05, "loss": 0.0, "step": 68496 }, { "epoch": 1.33, "learning_rate": 2.784969819304e-05, "loss": 0.0, "step": 68498 }, { "epoch": 1.33, "learning_rate": 2.784905123278277e-05, "loss": 0.0025, "step": 68500 }, { "epoch": 1.33, "learning_rate": 2.7848404272525543e-05, "loss": 0.0, "step": 68502 }, { "epoch": 1.33, "learning_rate": 2.784775731226831e-05, "loss": 0.0, "step": 68504 }, { "epoch": 1.33, "learning_rate": 2.784711035201108e-05, "loss": 0.0, "step": 68506 }, { "epoch": 1.33, "learning_rate": 2.7846463391753847e-05, "loss": 0.0016, "step": 68508 }, { "epoch": 1.33, "learning_rate": 2.7845816431496612e-05, "loss": 0.0, "step": 68510 }, { "epoch": 1.33, "learning_rate": 2.7845169471239385e-05, "loss": 0.0, "step": 68512 }, { "epoch": 1.33, "learning_rate": 2.784452251098215e-05, "loss": 0.0067, "step": 68514 }, { "epoch": 1.33, "learning_rate": 2.784387555072492e-05, "loss": 0.0, "step": 68516 }, { "epoch": 1.33, "learning_rate": 2.7843228590467692e-05, "loss": 0.0, "step": 68518 }, { "epoch": 1.33, "learning_rate": 2.7842581630210458e-05, "loss": 0.0, "step": 68520 }, { "epoch": 1.33, "learning_rate": 2.7841934669953223e-05, "loss": 0.0039, "step": 68522 }, { "epoch": 1.33, "learning_rate": 2.7841287709695996e-05, "loss": 0.0001, "step": 68524 }, { "epoch": 1.33, "learning_rate": 2.784064074943876e-05, "loss": 0.0, "step": 68526 }, { "epoch": 1.33, "learning_rate": 2.7839993789181534e-05, "loss": 0.0, "step": 68528 }, { "epoch": 1.33, "learning_rate": 2.78393468289243e-05, "loss": 0.0133, "step": 68530 }, { "epoch": 1.33, "learning_rate": 2.783869986866707e-05, "loss": 0.0, "step": 68532 }, { "epoch": 1.33, "learning_rate": 2.7838052908409838e-05, "loss": 0.0, "step": 68534 }, { "epoch": 1.33, "learning_rate": 2.7837405948152607e-05, "loss": 0.0, "step": 68536 }, { "epoch": 1.33, "learning_rate": 2.7836758987895373e-05, "loss": 0.0083, "step": 68538 }, { "epoch": 1.33, "learning_rate": 2.7836112027638145e-05, "loss": 0.0, "step": 68540 }, { "epoch": 1.33, "learning_rate": 2.783546506738091e-05, "loss": 0.0023, "step": 68542 }, { "epoch": 1.33, "learning_rate": 2.7834818107123677e-05, "loss": 0.0, "step": 68544 }, { "epoch": 1.33, "learning_rate": 2.783417114686645e-05, "loss": 0.0, "step": 68546 }, { "epoch": 1.33, "learning_rate": 2.7833524186609218e-05, "loss": 0.0, "step": 68548 }, { "epoch": 1.33, "learning_rate": 2.7832877226351987e-05, "loss": 0.0, "step": 68550 }, { "epoch": 1.33, "learning_rate": 2.7832230266094756e-05, "loss": 0.0, "step": 68552 }, { "epoch": 1.33, "learning_rate": 2.7831583305837522e-05, "loss": 0.0, "step": 68554 }, { "epoch": 1.33, "learning_rate": 2.7830936345580294e-05, "loss": 0.0, "step": 68556 }, { "epoch": 1.33, "learning_rate": 2.783028938532306e-05, "loss": 0.0, "step": 68558 }, { "epoch": 1.33, "learning_rate": 2.7829642425065826e-05, "loss": 0.0, "step": 68560 }, { "epoch": 1.33, "learning_rate": 2.78289954648086e-05, "loss": 0.0001, "step": 68562 }, { "epoch": 1.33, "learning_rate": 2.7828348504551367e-05, "loss": 0.0, "step": 68564 }, { "epoch": 1.33, "learning_rate": 2.7827701544294136e-05, "loss": 0.0001, "step": 68566 }, { "epoch": 1.33, "learning_rate": 2.7827054584036906e-05, "loss": 0.0001, "step": 68568 }, { "epoch": 1.33, "learning_rate": 2.782640762377967e-05, "loss": 0.0, "step": 68570 }, { "epoch": 1.33, "learning_rate": 2.7825760663522444e-05, "loss": 0.0, "step": 68572 }, { "epoch": 1.33, "learning_rate": 2.782511370326521e-05, "loss": 0.0, "step": 68574 }, { "epoch": 1.33, "learning_rate": 2.7824466743007975e-05, "loss": 0.0036, "step": 68576 }, { "epoch": 1.33, "learning_rate": 2.7823819782750748e-05, "loss": 0.0, "step": 68578 }, { "epoch": 1.33, "learning_rate": 2.7823172822493517e-05, "loss": 0.0, "step": 68580 }, { "epoch": 1.33, "learning_rate": 2.7822525862236282e-05, "loss": 0.0, "step": 68582 }, { "epoch": 1.33, "learning_rate": 2.7821878901979055e-05, "loss": 0.0, "step": 68584 }, { "epoch": 1.33, "learning_rate": 2.782123194172182e-05, "loss": 0.0, "step": 68586 }, { "epoch": 1.33, "learning_rate": 2.7820584981464593e-05, "loss": 0.0009, "step": 68588 }, { "epoch": 1.33, "learning_rate": 2.781993802120736e-05, "loss": 0.0002, "step": 68590 }, { "epoch": 1.33, "learning_rate": 2.7819291060950124e-05, "loss": 0.0073, "step": 68592 }, { "epoch": 1.33, "learning_rate": 2.7818644100692897e-05, "loss": 0.02, "step": 68594 }, { "epoch": 1.33, "learning_rate": 2.7817997140435666e-05, "loss": 0.0001, "step": 68596 }, { "epoch": 1.33, "learning_rate": 2.781735018017843e-05, "loss": 0.0, "step": 68598 }, { "epoch": 1.33, "learning_rate": 2.7816703219921204e-05, "loss": 0.0, "step": 68600 }, { "epoch": 1.33, "learning_rate": 2.781605625966397e-05, "loss": 0.0032, "step": 68602 }, { "epoch": 1.33, "learning_rate": 2.7815409299406736e-05, "loss": 0.0, "step": 68604 }, { "epoch": 1.33, "learning_rate": 2.7814762339149508e-05, "loss": 0.0, "step": 68606 }, { "epoch": 1.33, "learning_rate": 2.7814115378892274e-05, "loss": 0.0, "step": 68608 }, { "epoch": 1.33, "learning_rate": 2.7813468418635046e-05, "loss": 0.0, "step": 68610 }, { "epoch": 1.33, "learning_rate": 2.7812821458377812e-05, "loss": 0.0001, "step": 68612 }, { "epoch": 1.33, "learning_rate": 2.781217449812058e-05, "loss": 0.0, "step": 68614 }, { "epoch": 1.33, "learning_rate": 2.7811527537863353e-05, "loss": 0.0, "step": 68616 }, { "epoch": 1.33, "learning_rate": 2.781088057760612e-05, "loss": 0.0, "step": 68618 }, { "epoch": 1.33, "learning_rate": 2.7810233617348885e-05, "loss": 0.0021, "step": 68620 }, { "epoch": 1.33, "learning_rate": 2.7809586657091657e-05, "loss": 0.0, "step": 68622 }, { "epoch": 1.33, "learning_rate": 2.7808939696834423e-05, "loss": 0.0, "step": 68624 }, { "epoch": 1.33, "learning_rate": 2.7808292736577195e-05, "loss": 0.0, "step": 68626 }, { "epoch": 1.33, "learning_rate": 2.780764577631996e-05, "loss": 0.0, "step": 68628 }, { "epoch": 1.33, "learning_rate": 2.780699881606273e-05, "loss": 0.0, "step": 68630 }, { "epoch": 1.33, "learning_rate": 2.7806351855805503e-05, "loss": 0.0, "step": 68632 }, { "epoch": 1.33, "learning_rate": 2.780570489554827e-05, "loss": 0.0021, "step": 68634 }, { "epoch": 1.33, "learning_rate": 2.7805057935291034e-05, "loss": 0.0, "step": 68636 }, { "epoch": 1.33, "learning_rate": 2.7804410975033807e-05, "loss": 0.0, "step": 68638 }, { "epoch": 1.33, "learning_rate": 2.7803764014776572e-05, "loss": 0.0103, "step": 68640 }, { "epoch": 1.33, "learning_rate": 2.780311705451934e-05, "loss": 0.0, "step": 68642 }, { "epoch": 1.33, "learning_rate": 2.780247009426211e-05, "loss": 0.0003, "step": 68644 }, { "epoch": 1.33, "learning_rate": 2.780182313400488e-05, "loss": 0.0017, "step": 68646 }, { "epoch": 1.33, "learning_rate": 2.7801176173747652e-05, "loss": 0.0, "step": 68648 }, { "epoch": 1.33, "learning_rate": 2.7800529213490418e-05, "loss": 0.0, "step": 68650 }, { "epoch": 1.33, "learning_rate": 2.7799882253233183e-05, "loss": 0.0149, "step": 68652 }, { "epoch": 1.33, "learning_rate": 2.7799235292975956e-05, "loss": 0.0, "step": 68654 }, { "epoch": 1.33, "learning_rate": 2.779858833271872e-05, "loss": 0.0002, "step": 68656 }, { "epoch": 1.33, "learning_rate": 2.779794137246149e-05, "loss": 0.0, "step": 68658 }, { "epoch": 1.33, "learning_rate": 2.779729441220426e-05, "loss": 0.0, "step": 68660 }, { "epoch": 1.33, "learning_rate": 2.779664745194703e-05, "loss": 0.0018, "step": 68662 }, { "epoch": 1.33, "learning_rate": 2.7796000491689795e-05, "loss": 0.0, "step": 68664 }, { "epoch": 1.33, "learning_rate": 2.7795353531432567e-05, "loss": 0.0, "step": 68666 }, { "epoch": 1.33, "learning_rate": 2.7794706571175333e-05, "loss": 0.0004, "step": 68668 }, { "epoch": 1.33, "learning_rate": 2.7794059610918105e-05, "loss": 0.0, "step": 68670 }, { "epoch": 1.33, "learning_rate": 2.779341265066087e-05, "loss": 0.0, "step": 68672 }, { "epoch": 1.33, "learning_rate": 2.7792765690403637e-05, "loss": 0.0001, "step": 68674 }, { "epoch": 1.33, "learning_rate": 2.779211873014641e-05, "loss": 0.0, "step": 68676 }, { "epoch": 1.33, "learning_rate": 2.7791471769889178e-05, "loss": 0.0, "step": 68678 }, { "epoch": 1.33, "learning_rate": 2.7790824809631944e-05, "loss": 0.0, "step": 68680 }, { "epoch": 1.33, "learning_rate": 2.7790177849374716e-05, "loss": 0.0, "step": 68682 }, { "epoch": 1.33, "learning_rate": 2.7789530889117482e-05, "loss": 0.0053, "step": 68684 }, { "epoch": 1.33, "learning_rate": 2.7788883928860248e-05, "loss": 0.0075, "step": 68686 }, { "epoch": 1.33, "learning_rate": 2.778823696860302e-05, "loss": 0.0007, "step": 68688 }, { "epoch": 1.33, "learning_rate": 2.7787590008345786e-05, "loss": 0.0, "step": 68690 }, { "epoch": 1.33, "learning_rate": 2.778694304808856e-05, "loss": 0.0, "step": 68692 }, { "epoch": 1.33, "learning_rate": 2.7786296087831327e-05, "loss": 0.0003, "step": 68694 }, { "epoch": 1.33, "learning_rate": 2.7785649127574093e-05, "loss": 0.0, "step": 68696 }, { "epoch": 1.33, "learning_rate": 2.7785002167316866e-05, "loss": 0.0, "step": 68698 }, { "epoch": 1.33, "learning_rate": 2.778435520705963e-05, "loss": 0.0, "step": 68700 }, { "epoch": 1.33, "learning_rate": 2.7783708246802397e-05, "loss": 0.0, "step": 68702 }, { "epoch": 1.33, "learning_rate": 2.778306128654517e-05, "loss": 0.0, "step": 68704 }, { "epoch": 1.33, "learning_rate": 2.7782414326287935e-05, "loss": 0.0, "step": 68706 }, { "epoch": 1.33, "learning_rate": 2.7781767366030708e-05, "loss": 0.0001, "step": 68708 }, { "epoch": 1.33, "learning_rate": 2.7781120405773477e-05, "loss": 0.0, "step": 68710 }, { "epoch": 1.33, "learning_rate": 2.7780473445516242e-05, "loss": 0.0, "step": 68712 }, { "epoch": 1.33, "learning_rate": 2.7779826485259015e-05, "loss": 0.0, "step": 68714 }, { "epoch": 1.33, "learning_rate": 2.777917952500178e-05, "loss": 0.0, "step": 68716 }, { "epoch": 1.33, "learning_rate": 2.7778532564744546e-05, "loss": 0.0, "step": 68718 }, { "epoch": 1.33, "learning_rate": 2.777788560448732e-05, "loss": 0.0001, "step": 68720 }, { "epoch": 1.33, "learning_rate": 2.7777238644230084e-05, "loss": 0.0, "step": 68722 }, { "epoch": 1.33, "learning_rate": 2.7776591683972854e-05, "loss": 0.0, "step": 68724 }, { "epoch": 1.33, "learning_rate": 2.7775944723715626e-05, "loss": 0.0, "step": 68726 }, { "epoch": 1.33, "learning_rate": 2.777529776345839e-05, "loss": 0.0, "step": 68728 }, { "epoch": 1.33, "learning_rate": 2.7774650803201164e-05, "loss": 0.0, "step": 68730 }, { "epoch": 1.33, "learning_rate": 2.777400384294393e-05, "loss": 0.0, "step": 68732 }, { "epoch": 1.33, "learning_rate": 2.7773356882686696e-05, "loss": 0.0, "step": 68734 }, { "epoch": 1.33, "learning_rate": 2.7772709922429468e-05, "loss": 0.0, "step": 68736 }, { "epoch": 1.33, "learning_rate": 2.7772062962172234e-05, "loss": 0.0, "step": 68738 }, { "epoch": 1.33, "learning_rate": 2.7771416001915003e-05, "loss": 0.0, "step": 68740 }, { "epoch": 1.33, "learning_rate": 2.7770769041657775e-05, "loss": 0.0319, "step": 68742 }, { "epoch": 1.33, "learning_rate": 2.777012208140054e-05, "loss": 0.0, "step": 68744 }, { "epoch": 1.33, "learning_rate": 2.7769475121143307e-05, "loss": 0.0, "step": 68746 }, { "epoch": 1.33, "learning_rate": 2.776882816088608e-05, "loss": 0.0, "step": 68748 }, { "epoch": 1.33, "learning_rate": 2.7768181200628845e-05, "loss": 0.0, "step": 68750 }, { "epoch": 1.33, "learning_rate": 2.7767534240371617e-05, "loss": 0.0011, "step": 68752 }, { "epoch": 1.33, "learning_rate": 2.7766887280114383e-05, "loss": 0.0, "step": 68754 }, { "epoch": 1.33, "learning_rate": 2.7766240319857152e-05, "loss": 0.0, "step": 68756 }, { "epoch": 1.33, "learning_rate": 2.7765593359599925e-05, "loss": 0.0, "step": 68758 }, { "epoch": 1.33, "learning_rate": 2.776494639934269e-05, "loss": 0.0, "step": 68760 }, { "epoch": 1.33, "learning_rate": 2.7764299439085456e-05, "loss": 0.0, "step": 68762 }, { "epoch": 1.33, "learning_rate": 2.776365247882823e-05, "loss": 0.0005, "step": 68764 }, { "epoch": 1.33, "learning_rate": 2.7763005518570994e-05, "loss": 0.0, "step": 68766 }, { "epoch": 1.33, "learning_rate": 2.7762358558313767e-05, "loss": 0.0, "step": 68768 }, { "epoch": 1.33, "learning_rate": 2.7761711598056532e-05, "loss": 0.0, "step": 68770 }, { "epoch": 1.33, "learning_rate": 2.77610646377993e-05, "loss": 0.0, "step": 68772 }, { "epoch": 1.33, "learning_rate": 2.776041767754207e-05, "loss": 0.0001, "step": 68774 }, { "epoch": 1.33, "learning_rate": 2.775977071728484e-05, "loss": 0.0, "step": 68776 }, { "epoch": 1.33, "learning_rate": 2.7759123757027605e-05, "loss": 0.0, "step": 68778 }, { "epoch": 1.33, "learning_rate": 2.7758476796770378e-05, "loss": 0.0, "step": 68780 }, { "epoch": 1.33, "learning_rate": 2.7757829836513143e-05, "loss": 0.0, "step": 68782 }, { "epoch": 1.34, "learning_rate": 2.775718287625591e-05, "loss": 0.0, "step": 68784 }, { "epoch": 1.34, "learning_rate": 2.775653591599868e-05, "loss": 0.0, "step": 68786 }, { "epoch": 1.34, "learning_rate": 2.775588895574145e-05, "loss": 0.0, "step": 68788 }, { "epoch": 1.34, "learning_rate": 2.775524199548422e-05, "loss": 0.0, "step": 68790 }, { "epoch": 1.34, "learning_rate": 2.775459503522699e-05, "loss": 0.0248, "step": 68792 }, { "epoch": 1.34, "learning_rate": 2.7753948074969755e-05, "loss": 0.0034, "step": 68794 }, { "epoch": 1.34, "learning_rate": 2.7753301114712527e-05, "loss": 0.0, "step": 68796 }, { "epoch": 1.34, "learning_rate": 2.7752654154455293e-05, "loss": 0.0, "step": 68798 }, { "epoch": 1.34, "learning_rate": 2.775200719419806e-05, "loss": 0.0, "step": 68800 }, { "epoch": 1.34, "learning_rate": 2.775136023394083e-05, "loss": 0.0136, "step": 68802 }, { "epoch": 1.34, "learning_rate": 2.77507132736836e-05, "loss": 0.0, "step": 68804 }, { "epoch": 1.34, "learning_rate": 2.7750066313426366e-05, "loss": 0.0001, "step": 68806 }, { "epoch": 1.34, "learning_rate": 2.7749419353169138e-05, "loss": 0.0, "step": 68808 }, { "epoch": 1.34, "learning_rate": 2.7748772392911904e-05, "loss": 0.0, "step": 68810 }, { "epoch": 1.34, "learning_rate": 2.7748125432654676e-05, "loss": 0.006, "step": 68812 }, { "epoch": 1.34, "learning_rate": 2.7747478472397442e-05, "loss": 0.0, "step": 68814 }, { "epoch": 1.34, "learning_rate": 2.7746831512140208e-05, "loss": 0.0, "step": 68816 }, { "epoch": 1.34, "learning_rate": 2.774618455188298e-05, "loss": 0.0, "step": 68818 }, { "epoch": 1.34, "learning_rate": 2.774553759162575e-05, "loss": 0.0, "step": 68820 }, { "epoch": 1.34, "learning_rate": 2.7744890631368515e-05, "loss": 0.0, "step": 68822 }, { "epoch": 1.34, "learning_rate": 2.7744243671111287e-05, "loss": 0.0, "step": 68824 }, { "epoch": 1.34, "learning_rate": 2.7743596710854053e-05, "loss": 0.0, "step": 68826 }, { "epoch": 1.34, "learning_rate": 2.774294975059682e-05, "loss": 0.0001, "step": 68828 }, { "epoch": 1.34, "learning_rate": 2.774230279033959e-05, "loss": 0.0064, "step": 68830 }, { "epoch": 1.34, "learning_rate": 2.7741655830082357e-05, "loss": 0.0, "step": 68832 }, { "epoch": 1.34, "learning_rate": 2.774100886982513e-05, "loss": 0.0, "step": 68834 }, { "epoch": 1.34, "learning_rate": 2.77403619095679e-05, "loss": 0.0, "step": 68836 }, { "epoch": 1.34, "learning_rate": 2.7739714949310664e-05, "loss": 0.003, "step": 68838 }, { "epoch": 1.34, "learning_rate": 2.7739067989053437e-05, "loss": 0.0, "step": 68840 }, { "epoch": 1.34, "learning_rate": 2.7738421028796202e-05, "loss": 0.0, "step": 68842 }, { "epoch": 1.34, "learning_rate": 2.7737774068538968e-05, "loss": 0.0013, "step": 68844 }, { "epoch": 1.34, "learning_rate": 2.773712710828174e-05, "loss": 0.0, "step": 68846 }, { "epoch": 1.34, "learning_rate": 2.7736480148024506e-05, "loss": 0.0, "step": 68848 }, { "epoch": 1.34, "learning_rate": 2.773583318776728e-05, "loss": 0.0, "step": 68850 }, { "epoch": 1.34, "learning_rate": 2.7735186227510044e-05, "loss": 0.0, "step": 68852 }, { "epoch": 1.34, "learning_rate": 2.7734539267252813e-05, "loss": 0.0, "step": 68854 }, { "epoch": 1.34, "learning_rate": 2.7733892306995586e-05, "loss": 0.0, "step": 68856 }, { "epoch": 1.34, "learning_rate": 2.773324534673835e-05, "loss": 0.0, "step": 68858 }, { "epoch": 1.34, "learning_rate": 2.7732598386481117e-05, "loss": 0.0, "step": 68860 }, { "epoch": 1.34, "learning_rate": 2.773195142622389e-05, "loss": 0.0, "step": 68862 }, { "epoch": 1.34, "learning_rate": 2.7731304465966656e-05, "loss": 0.0, "step": 68864 }, { "epoch": 1.34, "learning_rate": 2.7730657505709425e-05, "loss": 0.0, "step": 68866 }, { "epoch": 1.34, "learning_rate": 2.7730010545452194e-05, "loss": 0.0, "step": 68868 }, { "epoch": 1.34, "learning_rate": 2.7729363585194963e-05, "loss": 0.0001, "step": 68870 }, { "epoch": 1.34, "learning_rate": 2.7728716624937735e-05, "loss": 0.0, "step": 68872 }, { "epoch": 1.34, "learning_rate": 2.77280696646805e-05, "loss": 0.0, "step": 68874 }, { "epoch": 1.34, "learning_rate": 2.7727422704423267e-05, "loss": 0.0007, "step": 68876 }, { "epoch": 1.34, "learning_rate": 2.772677574416604e-05, "loss": 0.0, "step": 68878 }, { "epoch": 1.34, "learning_rate": 2.7726128783908805e-05, "loss": 0.0, "step": 68880 }, { "epoch": 1.34, "learning_rate": 2.7725481823651574e-05, "loss": 0.0, "step": 68882 }, { "epoch": 1.34, "learning_rate": 2.7724834863394343e-05, "loss": 0.0, "step": 68884 }, { "epoch": 1.34, "learning_rate": 2.7724187903137112e-05, "loss": 0.0, "step": 68886 }, { "epoch": 1.34, "learning_rate": 2.7723540942879878e-05, "loss": 0.0002, "step": 68888 }, { "epoch": 1.34, "learning_rate": 2.772289398262265e-05, "loss": 0.001, "step": 68890 }, { "epoch": 1.34, "learning_rate": 2.7722247022365416e-05, "loss": 0.0055, "step": 68892 }, { "epoch": 1.34, "learning_rate": 2.772160006210819e-05, "loss": 0.0024, "step": 68894 }, { "epoch": 1.34, "learning_rate": 2.7720953101850954e-05, "loss": 0.0, "step": 68896 }, { "epoch": 1.34, "learning_rate": 2.7720306141593723e-05, "loss": 0.0, "step": 68898 }, { "epoch": 1.34, "learning_rate": 2.7719659181336492e-05, "loss": 0.0, "step": 68900 }, { "epoch": 1.34, "learning_rate": 2.771901222107926e-05, "loss": 0.0, "step": 68902 }, { "epoch": 1.34, "learning_rate": 2.7718365260822027e-05, "loss": 0.0038, "step": 68904 }, { "epoch": 1.34, "learning_rate": 2.77177183005648e-05, "loss": 0.0002, "step": 68906 }, { "epoch": 1.34, "learning_rate": 2.7717071340307565e-05, "loss": 0.0001, "step": 68908 }, { "epoch": 1.34, "learning_rate": 2.7716424380050338e-05, "loss": 0.0, "step": 68910 }, { "epoch": 1.34, "learning_rate": 2.7715777419793103e-05, "loss": 0.0, "step": 68912 }, { "epoch": 1.34, "learning_rate": 2.771513045953587e-05, "loss": 0.0001, "step": 68914 }, { "epoch": 1.34, "learning_rate": 2.771448349927864e-05, "loss": 0.0, "step": 68916 }, { "epoch": 1.34, "learning_rate": 2.771383653902141e-05, "loss": 0.0, "step": 68918 }, { "epoch": 1.34, "learning_rate": 2.7713189578764176e-05, "loss": 0.0002, "step": 68920 }, { "epoch": 1.34, "learning_rate": 2.771254261850695e-05, "loss": 0.0, "step": 68922 }, { "epoch": 1.34, "learning_rate": 2.7711895658249715e-05, "loss": 0.0, "step": 68924 }, { "epoch": 1.34, "learning_rate": 2.771124869799248e-05, "loss": 0.0, "step": 68926 }, { "epoch": 1.34, "learning_rate": 2.7710601737735253e-05, "loss": 0.0008, "step": 68928 }, { "epoch": 1.34, "learning_rate": 2.770995477747802e-05, "loss": 0.0, "step": 68930 }, { "epoch": 1.34, "learning_rate": 2.770930781722079e-05, "loss": 0.0, "step": 68932 }, { "epoch": 1.34, "learning_rate": 2.770866085696356e-05, "loss": 0.0001, "step": 68934 }, { "epoch": 1.34, "learning_rate": 2.7708013896706326e-05, "loss": 0.0, "step": 68936 }, { "epoch": 1.34, "learning_rate": 2.7707366936449098e-05, "loss": 0.0002, "step": 68938 }, { "epoch": 1.34, "learning_rate": 2.7706719976191864e-05, "loss": 0.0, "step": 68940 }, { "epoch": 1.34, "learning_rate": 2.770607301593463e-05, "loss": 0.0, "step": 68942 }, { "epoch": 1.34, "learning_rate": 2.7705426055677402e-05, "loss": 0.0, "step": 68944 }, { "epoch": 1.34, "learning_rate": 2.7704779095420168e-05, "loss": 0.0, "step": 68946 }, { "epoch": 1.34, "learning_rate": 2.7704132135162937e-05, "loss": 0.0043, "step": 68948 }, { "epoch": 1.34, "learning_rate": 2.770348517490571e-05, "loss": 0.0, "step": 68950 }, { "epoch": 1.34, "learning_rate": 2.7702838214648475e-05, "loss": 0.0, "step": 68952 }, { "epoch": 1.34, "learning_rate": 2.7702191254391247e-05, "loss": 0.0, "step": 68954 }, { "epoch": 1.34, "learning_rate": 2.7701544294134013e-05, "loss": 0.0, "step": 68956 }, { "epoch": 1.34, "learning_rate": 2.770089733387678e-05, "loss": 0.0, "step": 68958 }, { "epoch": 1.34, "learning_rate": 2.770025037361955e-05, "loss": 0.0, "step": 68960 }, { "epoch": 1.34, "learning_rate": 2.7699603413362317e-05, "loss": 0.0, "step": 68962 }, { "epoch": 1.34, "learning_rate": 2.7698956453105086e-05, "loss": 0.0, "step": 68964 }, { "epoch": 1.34, "learning_rate": 2.769830949284786e-05, "loss": 0.0072, "step": 68966 }, { "epoch": 1.34, "learning_rate": 2.7697662532590624e-05, "loss": 0.0, "step": 68968 }, { "epoch": 1.34, "learning_rate": 2.769701557233339e-05, "loss": 0.0, "step": 68970 }, { "epoch": 1.34, "learning_rate": 2.7696368612076162e-05, "loss": 0.0, "step": 68972 }, { "epoch": 1.34, "learning_rate": 2.7695721651818928e-05, "loss": 0.0, "step": 68974 }, { "epoch": 1.34, "learning_rate": 2.76950746915617e-05, "loss": 0.0027, "step": 68976 }, { "epoch": 1.34, "learning_rate": 2.7694427731304466e-05, "loss": 0.0, "step": 68978 }, { "epoch": 1.34, "learning_rate": 2.7693780771047235e-05, "loss": 0.0, "step": 68980 }, { "epoch": 1.34, "learning_rate": 2.7693133810790008e-05, "loss": 0.0, "step": 68982 }, { "epoch": 1.34, "learning_rate": 2.7692486850532773e-05, "loss": 0.0001, "step": 68984 }, { "epoch": 1.34, "learning_rate": 2.769183989027554e-05, "loss": 0.0007, "step": 68986 }, { "epoch": 1.34, "learning_rate": 2.769119293001831e-05, "loss": 0.0, "step": 68988 }, { "epoch": 1.34, "learning_rate": 2.7690545969761077e-05, "loss": 0.0, "step": 68990 }, { "epoch": 1.34, "learning_rate": 2.768989900950385e-05, "loss": 0.0, "step": 68992 }, { "epoch": 1.34, "learning_rate": 2.7689252049246616e-05, "loss": 0.0, "step": 68994 }, { "epoch": 1.34, "learning_rate": 2.7688605088989385e-05, "loss": 0.0065, "step": 68996 }, { "epoch": 1.34, "learning_rate": 2.7687958128732157e-05, "loss": 0.0006, "step": 68998 }, { "epoch": 1.34, "learning_rate": 2.7687311168474923e-05, "loss": 0.0, "step": 69000 }, { "epoch": 1.34, "learning_rate": 2.768666420821769e-05, "loss": 0.0, "step": 69002 }, { "epoch": 1.34, "learning_rate": 2.768601724796046e-05, "loss": 0.0, "step": 69004 }, { "epoch": 1.34, "learning_rate": 2.7685370287703227e-05, "loss": 0.0007, "step": 69006 }, { "epoch": 1.34, "learning_rate": 2.7684723327445992e-05, "loss": 0.0001, "step": 69008 }, { "epoch": 1.34, "learning_rate": 2.7684076367188765e-05, "loss": 0.0, "step": 69010 }, { "epoch": 1.34, "learning_rate": 2.7683429406931534e-05, "loss": 0.0, "step": 69012 }, { "epoch": 1.34, "learning_rate": 2.7682782446674303e-05, "loss": 0.0, "step": 69014 }, { "epoch": 1.34, "learning_rate": 2.7682135486417072e-05, "loss": 0.0, "step": 69016 }, { "epoch": 1.34, "learning_rate": 2.7681488526159838e-05, "loss": 0.0, "step": 69018 }, { "epoch": 1.34, "learning_rate": 2.768084156590261e-05, "loss": 0.0, "step": 69020 }, { "epoch": 1.34, "learning_rate": 2.7680194605645376e-05, "loss": 0.0, "step": 69022 }, { "epoch": 1.34, "learning_rate": 2.767954764538814e-05, "loss": 0.0, "step": 69024 }, { "epoch": 1.34, "learning_rate": 2.7678900685130914e-05, "loss": 0.0, "step": 69026 }, { "epoch": 1.34, "learning_rate": 2.7678253724873683e-05, "loss": 0.0, "step": 69028 }, { "epoch": 1.34, "learning_rate": 2.767760676461645e-05, "loss": 0.0, "step": 69030 }, { "epoch": 1.34, "learning_rate": 2.767695980435922e-05, "loss": 0.0, "step": 69032 }, { "epoch": 1.34, "learning_rate": 2.7676312844101987e-05, "loss": 0.0, "step": 69034 }, { "epoch": 1.34, "learning_rate": 2.767566588384476e-05, "loss": 0.0, "step": 69036 }, { "epoch": 1.34, "learning_rate": 2.7675018923587525e-05, "loss": 0.0, "step": 69038 }, { "epoch": 1.34, "learning_rate": 2.767437196333029e-05, "loss": 0.0, "step": 69040 }, { "epoch": 1.34, "learning_rate": 2.7673725003073063e-05, "loss": 0.0, "step": 69042 }, { "epoch": 1.34, "learning_rate": 2.7673078042815832e-05, "loss": 0.0001, "step": 69044 }, { "epoch": 1.34, "learning_rate": 2.7672431082558598e-05, "loss": 0.0069, "step": 69046 }, { "epoch": 1.34, "learning_rate": 2.767178412230137e-05, "loss": 0.0, "step": 69048 }, { "epoch": 1.34, "learning_rate": 2.7671137162044136e-05, "loss": 0.0, "step": 69050 }, { "epoch": 1.34, "learning_rate": 2.767049020178691e-05, "loss": 0.0, "step": 69052 }, { "epoch": 1.34, "learning_rate": 2.7669843241529674e-05, "loss": 0.0001, "step": 69054 }, { "epoch": 1.34, "learning_rate": 2.766919628127244e-05, "loss": 0.0, "step": 69056 }, { "epoch": 1.34, "learning_rate": 2.7668549321015213e-05, "loss": 0.004, "step": 69058 }, { "epoch": 1.34, "learning_rate": 2.7667902360757982e-05, "loss": 0.001, "step": 69060 }, { "epoch": 1.34, "learning_rate": 2.7667255400500747e-05, "loss": 0.0002, "step": 69062 }, { "epoch": 1.34, "learning_rate": 2.766660844024352e-05, "loss": 0.0, "step": 69064 }, { "epoch": 1.34, "learning_rate": 2.7665961479986286e-05, "loss": 0.0, "step": 69066 }, { "epoch": 1.34, "learning_rate": 2.766531451972905e-05, "loss": 0.0, "step": 69068 }, { "epoch": 1.34, "learning_rate": 2.7664667559471824e-05, "loss": 0.0, "step": 69070 }, { "epoch": 1.34, "learning_rate": 2.766402059921459e-05, "loss": 0.0, "step": 69072 }, { "epoch": 1.34, "learning_rate": 2.7663373638957362e-05, "loss": 0.0, "step": 69074 }, { "epoch": 1.34, "learning_rate": 2.766272667870013e-05, "loss": 0.0, "step": 69076 }, { "epoch": 1.34, "learning_rate": 2.7662079718442897e-05, "loss": 0.0, "step": 69078 }, { "epoch": 1.34, "learning_rate": 2.766143275818567e-05, "loss": 0.0, "step": 69080 }, { "epoch": 1.34, "learning_rate": 2.7660785797928435e-05, "loss": 0.0, "step": 69082 }, { "epoch": 1.34, "learning_rate": 2.76601388376712e-05, "loss": 0.0, "step": 69084 }, { "epoch": 1.34, "learning_rate": 2.7659491877413973e-05, "loss": 0.0, "step": 69086 }, { "epoch": 1.34, "learning_rate": 2.765884491715674e-05, "loss": 0.0, "step": 69088 }, { "epoch": 1.34, "learning_rate": 2.7658197956899508e-05, "loss": 0.0, "step": 69090 }, { "epoch": 1.34, "learning_rate": 2.7657550996642277e-05, "loss": 0.0179, "step": 69092 }, { "epoch": 1.34, "learning_rate": 2.7656904036385046e-05, "loss": 0.0, "step": 69094 }, { "epoch": 1.34, "learning_rate": 2.765625707612782e-05, "loss": 0.0, "step": 69096 }, { "epoch": 1.34, "learning_rate": 2.7655610115870584e-05, "loss": 0.0, "step": 69098 }, { "epoch": 1.34, "learning_rate": 2.765496315561335e-05, "loss": 0.0, "step": 69100 }, { "epoch": 1.34, "learning_rate": 2.7654316195356122e-05, "loss": 0.001, "step": 69102 }, { "epoch": 1.34, "learning_rate": 2.7653669235098888e-05, "loss": 0.0008, "step": 69104 }, { "epoch": 1.34, "learning_rate": 2.7653022274841657e-05, "loss": 0.0, "step": 69106 }, { "epoch": 1.34, "learning_rate": 2.7652375314584426e-05, "loss": 0.0, "step": 69108 }, { "epoch": 1.34, "learning_rate": 2.7651728354327195e-05, "loss": 0.0022, "step": 69110 }, { "epoch": 1.34, "learning_rate": 2.765108139406996e-05, "loss": 0.0, "step": 69112 }, { "epoch": 1.34, "learning_rate": 2.7650434433812733e-05, "loss": 0.0, "step": 69114 }, { "epoch": 1.34, "learning_rate": 2.76497874735555e-05, "loss": 0.002, "step": 69116 }, { "epoch": 1.34, "learning_rate": 2.764914051329827e-05, "loss": 0.0, "step": 69118 }, { "epoch": 1.34, "learning_rate": 2.7648493553041037e-05, "loss": 0.0, "step": 69120 }, { "epoch": 1.34, "learning_rate": 2.7647846592783806e-05, "loss": 0.0, "step": 69122 }, { "epoch": 1.34, "learning_rate": 2.7647199632526576e-05, "loss": 0.0, "step": 69124 }, { "epoch": 1.34, "learning_rate": 2.7646552672269345e-05, "loss": 0.0, "step": 69126 }, { "epoch": 1.34, "learning_rate": 2.764590571201211e-05, "loss": 0.0, "step": 69128 }, { "epoch": 1.34, "learning_rate": 2.7645258751754883e-05, "loss": 0.0008, "step": 69130 }, { "epoch": 1.34, "learning_rate": 2.764461179149765e-05, "loss": 0.0054, "step": 69132 }, { "epoch": 1.34, "learning_rate": 2.764396483124042e-05, "loss": 0.0064, "step": 69134 }, { "epoch": 1.34, "learning_rate": 2.7643317870983187e-05, "loss": 0.0, "step": 69136 }, { "epoch": 1.34, "learning_rate": 2.7642670910725956e-05, "loss": 0.0, "step": 69138 }, { "epoch": 1.34, "learning_rate": 2.7642023950468725e-05, "loss": 0.0179, "step": 69140 }, { "epoch": 1.34, "learning_rate": 2.7641376990211494e-05, "loss": 0.0029, "step": 69142 }, { "epoch": 1.34, "learning_rate": 2.764073002995426e-05, "loss": 0.0, "step": 69144 }, { "epoch": 1.34, "learning_rate": 2.7640083069697032e-05, "loss": 0.002, "step": 69146 }, { "epoch": 1.34, "learning_rate": 2.7639436109439798e-05, "loss": 0.0007, "step": 69148 }, { "epoch": 1.34, "learning_rate": 2.7638789149182563e-05, "loss": 0.0, "step": 69150 }, { "epoch": 1.34, "learning_rate": 2.7638142188925336e-05, "loss": 0.0016, "step": 69152 }, { "epoch": 1.34, "learning_rate": 2.76374952286681e-05, "loss": 0.0, "step": 69154 }, { "epoch": 1.34, "learning_rate": 2.7636848268410874e-05, "loss": 0.0001, "step": 69156 }, { "epoch": 1.34, "learning_rate": 2.7636201308153643e-05, "loss": 0.0037, "step": 69158 }, { "epoch": 1.34, "learning_rate": 2.763555434789641e-05, "loss": 0.0, "step": 69160 }, { "epoch": 1.34, "learning_rate": 2.763490738763918e-05, "loss": 0.0, "step": 69162 }, { "epoch": 1.34, "learning_rate": 2.7634260427381947e-05, "loss": 0.0002, "step": 69164 }, { "epoch": 1.34, "learning_rate": 2.7633613467124713e-05, "loss": 0.0, "step": 69166 }, { "epoch": 1.34, "learning_rate": 2.7632966506867485e-05, "loss": 0.0, "step": 69168 }, { "epoch": 1.34, "learning_rate": 2.763231954661025e-05, "loss": 0.0, "step": 69170 }, { "epoch": 1.34, "learning_rate": 2.763167258635302e-05, "loss": 0.0017, "step": 69172 }, { "epoch": 1.34, "learning_rate": 2.7631025626095792e-05, "loss": 0.0001, "step": 69174 }, { "epoch": 1.34, "learning_rate": 2.7630378665838558e-05, "loss": 0.0004, "step": 69176 }, { "epoch": 1.34, "learning_rate": 2.762973170558133e-05, "loss": 0.0, "step": 69178 }, { "epoch": 1.34, "learning_rate": 2.7629084745324096e-05, "loss": 0.0, "step": 69180 }, { "epoch": 1.34, "learning_rate": 2.7628437785066862e-05, "loss": 0.0001, "step": 69182 }, { "epoch": 1.34, "learning_rate": 2.7627790824809634e-05, "loss": 0.0, "step": 69184 }, { "epoch": 1.34, "learning_rate": 2.76271438645524e-05, "loss": 0.0, "step": 69186 }, { "epoch": 1.34, "learning_rate": 2.762649690429517e-05, "loss": 0.0, "step": 69188 }, { "epoch": 1.34, "learning_rate": 2.7625849944037942e-05, "loss": 0.0001, "step": 69190 }, { "epoch": 1.34, "learning_rate": 2.7625202983780707e-05, "loss": 0.0, "step": 69192 }, { "epoch": 1.34, "learning_rate": 2.762455602352348e-05, "loss": 0.0, "step": 69194 }, { "epoch": 1.34, "learning_rate": 2.7623909063266246e-05, "loss": 0.0, "step": 69196 }, { "epoch": 1.34, "learning_rate": 2.762326210300901e-05, "loss": 0.0, "step": 69198 }, { "epoch": 1.34, "learning_rate": 2.7622615142751784e-05, "loss": 0.0005, "step": 69200 }, { "epoch": 1.34, "learning_rate": 2.762196818249455e-05, "loss": 0.0002, "step": 69202 }, { "epoch": 1.34, "learning_rate": 2.762132122223732e-05, "loss": 0.0045, "step": 69204 }, { "epoch": 1.34, "learning_rate": 2.762067426198009e-05, "loss": 0.0002, "step": 69206 }, { "epoch": 1.34, "learning_rate": 2.7620027301722857e-05, "loss": 0.0001, "step": 69208 }, { "epoch": 1.34, "learning_rate": 2.7619380341465622e-05, "loss": 0.0, "step": 69210 }, { "epoch": 1.34, "learning_rate": 2.7618733381208395e-05, "loss": 0.0, "step": 69212 }, { "epoch": 1.34, "learning_rate": 2.761808642095116e-05, "loss": 0.0, "step": 69214 }, { "epoch": 1.34, "learning_rate": 2.7617439460693933e-05, "loss": 0.0, "step": 69216 }, { "epoch": 1.34, "learning_rate": 2.76167925004367e-05, "loss": 0.0, "step": 69218 }, { "epoch": 1.34, "learning_rate": 2.7616145540179468e-05, "loss": 0.0, "step": 69220 }, { "epoch": 1.34, "learning_rate": 2.761549857992224e-05, "loss": 0.0095, "step": 69222 }, { "epoch": 1.34, "learning_rate": 2.7614851619665006e-05, "loss": 0.0, "step": 69224 }, { "epoch": 1.34, "learning_rate": 2.7614204659407772e-05, "loss": 0.0007, "step": 69226 }, { "epoch": 1.34, "learning_rate": 2.7613557699150544e-05, "loss": 0.0, "step": 69228 }, { "epoch": 1.34, "learning_rate": 2.761291073889331e-05, "loss": 0.0, "step": 69230 }, { "epoch": 1.34, "learning_rate": 2.7612263778636076e-05, "loss": 0.0162, "step": 69232 }, { "epoch": 1.34, "learning_rate": 2.7611616818378848e-05, "loss": 0.0, "step": 69234 }, { "epoch": 1.34, "learning_rate": 2.7610969858121617e-05, "loss": 0.0, "step": 69236 }, { "epoch": 1.34, "learning_rate": 2.761032289786439e-05, "loss": 0.0001, "step": 69238 }, { "epoch": 1.34, "learning_rate": 2.7609675937607155e-05, "loss": 0.0116, "step": 69240 }, { "epoch": 1.34, "learning_rate": 2.760902897734992e-05, "loss": 0.0, "step": 69242 }, { "epoch": 1.34, "learning_rate": 2.7608382017092693e-05, "loss": 0.0218, "step": 69244 }, { "epoch": 1.34, "learning_rate": 2.760773505683546e-05, "loss": 0.0, "step": 69246 }, { "epoch": 1.34, "learning_rate": 2.7607088096578225e-05, "loss": 0.0, "step": 69248 }, { "epoch": 1.34, "learning_rate": 2.7606441136320997e-05, "loss": 0.0, "step": 69250 }, { "epoch": 1.34, "learning_rate": 2.7605794176063766e-05, "loss": 0.0, "step": 69252 }, { "epoch": 1.34, "learning_rate": 2.7605147215806532e-05, "loss": 0.0001, "step": 69254 }, { "epoch": 1.34, "learning_rate": 2.7604500255549305e-05, "loss": 0.0, "step": 69256 }, { "epoch": 1.34, "learning_rate": 2.760385329529207e-05, "loss": 0.0, "step": 69258 }, { "epoch": 1.34, "learning_rate": 2.7603206335034843e-05, "loss": 0.0, "step": 69260 }, { "epoch": 1.34, "learning_rate": 2.760255937477761e-05, "loss": 0.0, "step": 69262 }, { "epoch": 1.34, "learning_rate": 2.7601912414520374e-05, "loss": 0.0001, "step": 69264 }, { "epoch": 1.34, "learning_rate": 2.7601265454263147e-05, "loss": 0.0, "step": 69266 }, { "epoch": 1.34, "learning_rate": 2.7600618494005916e-05, "loss": 0.0, "step": 69268 }, { "epoch": 1.34, "learning_rate": 2.759997153374868e-05, "loss": 0.0, "step": 69270 }, { "epoch": 1.34, "learning_rate": 2.7599324573491454e-05, "loss": 0.0, "step": 69272 }, { "epoch": 1.34, "learning_rate": 2.759867761323422e-05, "loss": 0.0001, "step": 69274 }, { "epoch": 1.34, "learning_rate": 2.7598030652976992e-05, "loss": 0.0001, "step": 69276 }, { "epoch": 1.34, "learning_rate": 2.7597383692719758e-05, "loss": 0.0, "step": 69278 }, { "epoch": 1.34, "learning_rate": 2.7596736732462523e-05, "loss": 0.0, "step": 69280 }, { "epoch": 1.34, "learning_rate": 2.7596089772205296e-05, "loss": 0.0, "step": 69282 }, { "epoch": 1.34, "learning_rate": 2.7595442811948065e-05, "loss": 0.0001, "step": 69284 }, { "epoch": 1.34, "learning_rate": 2.759479585169083e-05, "loss": 0.0, "step": 69286 }, { "epoch": 1.34, "learning_rate": 2.7594148891433603e-05, "loss": 0.0012, "step": 69288 }, { "epoch": 1.34, "learning_rate": 2.759350193117637e-05, "loss": 0.0, "step": 69290 }, { "epoch": 1.34, "learning_rate": 2.7592854970919135e-05, "loss": 0.0, "step": 69292 }, { "epoch": 1.34, "learning_rate": 2.7592208010661907e-05, "loss": 0.0013, "step": 69294 }, { "epoch": 1.34, "learning_rate": 2.7591561050404673e-05, "loss": 0.0001, "step": 69296 }, { "epoch": 1.34, "learning_rate": 2.7590914090147445e-05, "loss": 0.0, "step": 69298 }, { "epoch": 1.35, "learning_rate": 2.7590267129890214e-05, "loss": 0.0017, "step": 69300 }, { "epoch": 1.35, "learning_rate": 2.758962016963298e-05, "loss": 0.0, "step": 69302 }, { "epoch": 1.35, "learning_rate": 2.7588973209375752e-05, "loss": 0.0, "step": 69304 }, { "epoch": 1.35, "learning_rate": 2.7588326249118518e-05, "loss": 0.0, "step": 69306 }, { "epoch": 1.35, "learning_rate": 2.7587679288861284e-05, "loss": 0.0, "step": 69308 }, { "epoch": 1.35, "learning_rate": 2.7587032328604056e-05, "loss": 0.0003, "step": 69310 }, { "epoch": 1.35, "learning_rate": 2.7586385368346822e-05, "loss": 0.0107, "step": 69312 }, { "epoch": 1.35, "learning_rate": 2.758573840808959e-05, "loss": 0.0, "step": 69314 }, { "epoch": 1.35, "learning_rate": 2.7585091447832364e-05, "loss": 0.0, "step": 69316 }, { "epoch": 1.35, "learning_rate": 2.758444448757513e-05, "loss": 0.0058, "step": 69318 }, { "epoch": 1.35, "learning_rate": 2.7583797527317902e-05, "loss": 0.0, "step": 69320 }, { "epoch": 1.35, "learning_rate": 2.7583150567060667e-05, "loss": 0.0, "step": 69322 }, { "epoch": 1.35, "learning_rate": 2.7582503606803433e-05, "loss": 0.0, "step": 69324 }, { "epoch": 1.35, "learning_rate": 2.7581856646546206e-05, "loss": 0.0, "step": 69326 }, { "epoch": 1.35, "learning_rate": 2.758120968628897e-05, "loss": 0.0001, "step": 69328 }, { "epoch": 1.35, "learning_rate": 2.758056272603174e-05, "loss": 0.0083, "step": 69330 }, { "epoch": 1.35, "learning_rate": 2.757991576577451e-05, "loss": 0.0004, "step": 69332 }, { "epoch": 1.35, "learning_rate": 2.757926880551728e-05, "loss": 0.0, "step": 69334 }, { "epoch": 1.35, "learning_rate": 2.757862184526005e-05, "loss": 0.0, "step": 69336 }, { "epoch": 1.35, "learning_rate": 2.7577974885002817e-05, "loss": 0.0, "step": 69338 }, { "epoch": 1.35, "learning_rate": 2.7577327924745582e-05, "loss": 0.0, "step": 69340 }, { "epoch": 1.35, "learning_rate": 2.7576680964488355e-05, "loss": 0.0, "step": 69342 }, { "epoch": 1.35, "learning_rate": 2.757603400423112e-05, "loss": 0.0003, "step": 69344 }, { "epoch": 1.35, "learning_rate": 2.757538704397389e-05, "loss": 0.0, "step": 69346 }, { "epoch": 1.35, "learning_rate": 2.757474008371666e-05, "loss": 0.0001, "step": 69348 }, { "epoch": 1.35, "learning_rate": 2.7574093123459428e-05, "loss": 0.0002, "step": 69350 }, { "epoch": 1.35, "learning_rate": 2.7573446163202194e-05, "loss": 0.0149, "step": 69352 }, { "epoch": 1.35, "learning_rate": 2.7572799202944966e-05, "loss": 0.0, "step": 69354 }, { "epoch": 1.35, "learning_rate": 2.7572152242687732e-05, "loss": 0.0, "step": 69356 }, { "epoch": 1.35, "learning_rate": 2.7571505282430504e-05, "loss": 0.0, "step": 69358 }, { "epoch": 1.35, "learning_rate": 2.757085832217327e-05, "loss": 0.0, "step": 69360 }, { "epoch": 1.35, "learning_rate": 2.757021136191604e-05, "loss": 0.0, "step": 69362 }, { "epoch": 1.35, "learning_rate": 2.7569564401658808e-05, "loss": 0.0, "step": 69364 }, { "epoch": 1.35, "learning_rate": 2.7568917441401577e-05, "loss": 0.0, "step": 69366 }, { "epoch": 1.35, "learning_rate": 2.7568270481144343e-05, "loss": 0.0, "step": 69368 }, { "epoch": 1.35, "learning_rate": 2.7567623520887115e-05, "loss": 0.0248, "step": 69370 }, { "epoch": 1.35, "learning_rate": 2.756697656062988e-05, "loss": 0.0001, "step": 69372 }, { "epoch": 1.35, "learning_rate": 2.7566329600372647e-05, "loss": 0.0021, "step": 69374 }, { "epoch": 1.35, "learning_rate": 2.756568264011542e-05, "loss": 0.0001, "step": 69376 }, { "epoch": 1.35, "learning_rate": 2.7565035679858188e-05, "loss": 0.0, "step": 69378 }, { "epoch": 1.35, "learning_rate": 2.7564388719600957e-05, "loss": 0.0, "step": 69380 }, { "epoch": 1.35, "learning_rate": 2.7563741759343726e-05, "loss": 0.0004, "step": 69382 }, { "epoch": 1.35, "learning_rate": 2.7563094799086492e-05, "loss": 0.0, "step": 69384 }, { "epoch": 1.35, "learning_rate": 2.7562447838829265e-05, "loss": 0.0001, "step": 69386 }, { "epoch": 1.35, "learning_rate": 2.756180087857203e-05, "loss": 0.0, "step": 69388 }, { "epoch": 1.35, "learning_rate": 2.7561153918314796e-05, "loss": 0.0043, "step": 69390 }, { "epoch": 1.35, "learning_rate": 2.756050695805757e-05, "loss": 0.0, "step": 69392 }, { "epoch": 1.35, "learning_rate": 2.7559859997800334e-05, "loss": 0.0, "step": 69394 }, { "epoch": 1.35, "learning_rate": 2.7559213037543103e-05, "loss": 0.0007, "step": 69396 }, { "epoch": 1.35, "learning_rate": 2.7558566077285876e-05, "loss": 0.0, "step": 69398 }, { "epoch": 1.35, "learning_rate": 2.755791911702864e-05, "loss": 0.0002, "step": 69400 }, { "epoch": 1.35, "learning_rate": 2.7557272156771414e-05, "loss": 0.0006, "step": 69402 }, { "epoch": 1.35, "learning_rate": 2.755662519651418e-05, "loss": 0.0, "step": 69404 }, { "epoch": 1.35, "learning_rate": 2.7555978236256945e-05, "loss": 0.0, "step": 69406 }, { "epoch": 1.35, "learning_rate": 2.7555331275999718e-05, "loss": 0.0, "step": 69408 }, { "epoch": 1.35, "learning_rate": 2.7554684315742483e-05, "loss": 0.0, "step": 69410 }, { "epoch": 1.35, "learning_rate": 2.7554037355485253e-05, "loss": 0.0, "step": 69412 }, { "epoch": 1.35, "learning_rate": 2.7553390395228025e-05, "loss": 0.0, "step": 69414 }, { "epoch": 1.35, "learning_rate": 2.755274343497079e-05, "loss": 0.0, "step": 69416 }, { "epoch": 1.35, "learning_rate": 2.7552096474713563e-05, "loss": 0.0, "step": 69418 }, { "epoch": 1.35, "learning_rate": 2.755144951445633e-05, "loss": 0.0, "step": 69420 }, { "epoch": 1.35, "learning_rate": 2.7550802554199095e-05, "loss": 0.0029, "step": 69422 }, { "epoch": 1.35, "learning_rate": 2.7550155593941867e-05, "loss": 0.0009, "step": 69424 }, { "epoch": 1.35, "learning_rate": 2.7549508633684633e-05, "loss": 0.0, "step": 69426 }, { "epoch": 1.35, "learning_rate": 2.7548861673427402e-05, "loss": 0.0001, "step": 69428 }, { "epoch": 1.35, "learning_rate": 2.7548214713170174e-05, "loss": 0.0, "step": 69430 }, { "epoch": 1.35, "learning_rate": 2.754756775291294e-05, "loss": 0.0, "step": 69432 }, { "epoch": 1.35, "learning_rate": 2.7546920792655706e-05, "loss": 0.0, "step": 69434 }, { "epoch": 1.35, "learning_rate": 2.7546273832398478e-05, "loss": 0.0, "step": 69436 }, { "epoch": 1.35, "learning_rate": 2.7545626872141244e-05, "loss": 0.0069, "step": 69438 }, { "epoch": 1.35, "learning_rate": 2.7544979911884016e-05, "loss": 0.0094, "step": 69440 }, { "epoch": 1.35, "learning_rate": 2.7544332951626782e-05, "loss": 0.0, "step": 69442 }, { "epoch": 1.35, "learning_rate": 2.754368599136955e-05, "loss": 0.0003, "step": 69444 }, { "epoch": 1.35, "learning_rate": 2.7543039031112324e-05, "loss": 0.0, "step": 69446 }, { "epoch": 1.35, "learning_rate": 2.754239207085509e-05, "loss": 0.0, "step": 69448 }, { "epoch": 1.35, "learning_rate": 2.7541745110597855e-05, "loss": 0.0, "step": 69450 }, { "epoch": 1.35, "learning_rate": 2.7541098150340627e-05, "loss": 0.0, "step": 69452 }, { "epoch": 1.35, "learning_rate": 2.7540451190083393e-05, "loss": 0.0, "step": 69454 }, { "epoch": 1.35, "learning_rate": 2.753980422982616e-05, "loss": 0.0, "step": 69456 }, { "epoch": 1.35, "learning_rate": 2.753915726956893e-05, "loss": 0.0002, "step": 69458 }, { "epoch": 1.35, "learning_rate": 2.75385103093117e-05, "loss": 0.0083, "step": 69460 }, { "epoch": 1.35, "learning_rate": 2.7537863349054473e-05, "loss": 0.0, "step": 69462 }, { "epoch": 1.35, "learning_rate": 2.753721638879724e-05, "loss": 0.0, "step": 69464 }, { "epoch": 1.35, "learning_rate": 2.7536569428540004e-05, "loss": 0.0, "step": 69466 }, { "epoch": 1.35, "learning_rate": 2.7535922468282777e-05, "loss": 0.0, "step": 69468 }, { "epoch": 1.35, "learning_rate": 2.7535275508025542e-05, "loss": 0.0, "step": 69470 }, { "epoch": 1.35, "learning_rate": 2.7534628547768308e-05, "loss": 0.0001, "step": 69472 }, { "epoch": 1.35, "learning_rate": 2.753398158751108e-05, "loss": 0.0, "step": 69474 }, { "epoch": 1.35, "learning_rate": 2.753333462725385e-05, "loss": 0.0, "step": 69476 }, { "epoch": 1.35, "learning_rate": 2.7532687666996615e-05, "loss": 0.0, "step": 69478 }, { "epoch": 1.35, "learning_rate": 2.7532040706739388e-05, "loss": 0.0, "step": 69480 }, { "epoch": 1.35, "learning_rate": 2.7531393746482154e-05, "loss": 0.0, "step": 69482 }, { "epoch": 1.35, "learning_rate": 2.7530746786224926e-05, "loss": 0.0, "step": 69484 }, { "epoch": 1.35, "learning_rate": 2.753009982596769e-05, "loss": 0.0, "step": 69486 }, { "epoch": 1.35, "learning_rate": 2.7529452865710457e-05, "loss": 0.0016, "step": 69488 }, { "epoch": 1.35, "learning_rate": 2.752880590545323e-05, "loss": 0.0, "step": 69490 }, { "epoch": 1.35, "learning_rate": 2.7528158945196e-05, "loss": 0.0, "step": 69492 }, { "epoch": 1.35, "learning_rate": 2.7527511984938765e-05, "loss": 0.0, "step": 69494 }, { "epoch": 1.35, "learning_rate": 2.7526865024681537e-05, "loss": 0.0, "step": 69496 }, { "epoch": 1.35, "learning_rate": 2.7526218064424303e-05, "loss": 0.0, "step": 69498 }, { "epoch": 1.35, "learning_rate": 2.7525571104167075e-05, "loss": 0.0, "step": 69500 }, { "epoch": 1.35, "learning_rate": 2.752492414390984e-05, "loss": 0.0, "step": 69502 }, { "epoch": 1.35, "learning_rate": 2.7524277183652607e-05, "loss": 0.0, "step": 69504 }, { "epoch": 1.35, "learning_rate": 2.752363022339538e-05, "loss": 0.0, "step": 69506 }, { "epoch": 1.35, "learning_rate": 2.7522983263138148e-05, "loss": 0.0002, "step": 69508 }, { "epoch": 1.35, "learning_rate": 2.7522336302880914e-05, "loss": 0.0139, "step": 69510 }, { "epoch": 1.35, "learning_rate": 2.7521689342623686e-05, "loss": 0.0, "step": 69512 }, { "epoch": 1.35, "learning_rate": 2.7521042382366452e-05, "loss": 0.0, "step": 69514 }, { "epoch": 1.35, "learning_rate": 2.7520395422109218e-05, "loss": 0.0, "step": 69516 }, { "epoch": 1.35, "learning_rate": 2.751974846185199e-05, "loss": 0.0, "step": 69518 }, { "epoch": 1.35, "learning_rate": 2.7519101501594756e-05, "loss": 0.0, "step": 69520 }, { "epoch": 1.35, "learning_rate": 2.751845454133753e-05, "loss": 0.0, "step": 69522 }, { "epoch": 1.35, "learning_rate": 2.7517807581080298e-05, "loss": 0.0, "step": 69524 }, { "epoch": 1.35, "learning_rate": 2.7517160620823063e-05, "loss": 0.0, "step": 69526 }, { "epoch": 1.35, "learning_rate": 2.7516513660565836e-05, "loss": 0.0, "step": 69528 }, { "epoch": 1.35, "learning_rate": 2.75158667003086e-05, "loss": 0.0001, "step": 69530 }, { "epoch": 1.35, "learning_rate": 2.7515219740051367e-05, "loss": 0.0, "step": 69532 }, { "epoch": 1.35, "learning_rate": 2.751457277979414e-05, "loss": 0.0, "step": 69534 }, { "epoch": 1.35, "learning_rate": 2.7513925819536905e-05, "loss": 0.0, "step": 69536 }, { "epoch": 1.35, "learning_rate": 2.7513278859279674e-05, "loss": 0.0, "step": 69538 }, { "epoch": 1.35, "learning_rate": 2.7512631899022447e-05, "loss": 0.0001, "step": 69540 }, { "epoch": 1.35, "learning_rate": 2.7511984938765213e-05, "loss": 0.0, "step": 69542 }, { "epoch": 1.35, "learning_rate": 2.7511337978507985e-05, "loss": 0.0042, "step": 69544 }, { "epoch": 1.35, "learning_rate": 2.751069101825075e-05, "loss": 0.0, "step": 69546 }, { "epoch": 1.35, "learning_rate": 2.7510044057993516e-05, "loss": 0.0, "step": 69548 }, { "epoch": 1.35, "learning_rate": 2.750939709773629e-05, "loss": 0.0054, "step": 69550 }, { "epoch": 1.35, "learning_rate": 2.7508750137479055e-05, "loss": 0.0, "step": 69552 }, { "epoch": 1.35, "learning_rate": 2.7508103177221824e-05, "loss": 0.0, "step": 69554 }, { "epoch": 1.35, "learning_rate": 2.7507456216964596e-05, "loss": 0.0, "step": 69556 }, { "epoch": 1.35, "learning_rate": 2.7506809256707362e-05, "loss": 0.0097, "step": 69558 }, { "epoch": 1.35, "learning_rate": 2.7506162296450134e-05, "loss": 0.0, "step": 69560 }, { "epoch": 1.35, "learning_rate": 2.75055153361929e-05, "loss": 0.0, "step": 69562 }, { "epoch": 1.35, "learning_rate": 2.7504868375935666e-05, "loss": 0.0001, "step": 69564 }, { "epoch": 1.35, "learning_rate": 2.7504221415678438e-05, "loss": 0.0, "step": 69566 }, { "epoch": 1.35, "learning_rate": 2.7503574455421204e-05, "loss": 0.0002, "step": 69568 }, { "epoch": 1.35, "learning_rate": 2.7502927495163973e-05, "loss": 0.0, "step": 69570 }, { "epoch": 1.35, "learning_rate": 2.7502280534906742e-05, "loss": 0.0, "step": 69572 }, { "epoch": 1.35, "learning_rate": 2.750163357464951e-05, "loss": 0.0001, "step": 69574 }, { "epoch": 1.35, "learning_rate": 2.7501310094520892e-05, "loss": 0.0161, "step": 69576 }, { "epoch": 1.35, "learning_rate": 2.7500663134263665e-05, "loss": 0.0028, "step": 69578 }, { "epoch": 1.35, "learning_rate": 2.7500016174006434e-05, "loss": 0.0, "step": 69580 }, { "epoch": 1.35, "learning_rate": 2.74993692137492e-05, "loss": 0.0, "step": 69582 }, { "epoch": 1.35, "learning_rate": 2.7498722253491972e-05, "loss": 0.0002, "step": 69584 }, { "epoch": 1.35, "learning_rate": 2.7498075293234738e-05, "loss": 0.0, "step": 69586 }, { "epoch": 1.35, "learning_rate": 2.7497428332977503e-05, "loss": 0.0, "step": 69588 }, { "epoch": 1.35, "learning_rate": 2.7496781372720276e-05, "loss": 0.0001, "step": 69590 }, { "epoch": 1.35, "learning_rate": 2.749613441246304e-05, "loss": 0.0004, "step": 69592 }, { "epoch": 1.35, "learning_rate": 2.7495487452205814e-05, "loss": 0.0, "step": 69594 }, { "epoch": 1.35, "learning_rate": 2.749484049194858e-05, "loss": 0.0, "step": 69596 }, { "epoch": 1.35, "learning_rate": 2.749419353169135e-05, "loss": 0.0054, "step": 69598 }, { "epoch": 1.35, "learning_rate": 2.749354657143412e-05, "loss": 0.0, "step": 69600 }, { "epoch": 1.35, "learning_rate": 2.7492899611176887e-05, "loss": 0.0, "step": 69602 }, { "epoch": 1.35, "learning_rate": 2.7492252650919653e-05, "loss": 0.0, "step": 69604 }, { "epoch": 1.35, "learning_rate": 2.7491605690662425e-05, "loss": 0.0001, "step": 69606 }, { "epoch": 1.35, "learning_rate": 2.749095873040519e-05, "loss": 0.0, "step": 69608 }, { "epoch": 1.35, "learning_rate": 2.749031177014796e-05, "loss": 0.0, "step": 69610 }, { "epoch": 1.35, "learning_rate": 2.748966480989073e-05, "loss": 0.0008, "step": 69612 }, { "epoch": 1.35, "learning_rate": 2.7489017849633498e-05, "loss": 0.0, "step": 69614 }, { "epoch": 1.35, "learning_rate": 2.748837088937627e-05, "loss": 0.0, "step": 69616 }, { "epoch": 1.35, "learning_rate": 2.7487723929119036e-05, "loss": 0.0, "step": 69618 }, { "epoch": 1.35, "learning_rate": 2.7487076968861802e-05, "loss": 0.0, "step": 69620 }, { "epoch": 1.35, "learning_rate": 2.7486430008604574e-05, "loss": 0.0092, "step": 69622 }, { "epoch": 1.35, "learning_rate": 2.748578304834734e-05, "loss": 0.0, "step": 69624 }, { "epoch": 1.35, "learning_rate": 2.748513608809011e-05, "loss": 0.0, "step": 69626 }, { "epoch": 1.35, "learning_rate": 2.7484489127832878e-05, "loss": 0.0099, "step": 69628 }, { "epoch": 1.35, "learning_rate": 2.7483842167575647e-05, "loss": 0.0, "step": 69630 }, { "epoch": 1.35, "learning_rate": 2.748319520731842e-05, "loss": 0.0001, "step": 69632 }, { "epoch": 1.35, "learning_rate": 2.7482548247061186e-05, "loss": 0.0, "step": 69634 }, { "epoch": 1.35, "learning_rate": 2.748190128680395e-05, "loss": 0.0, "step": 69636 }, { "epoch": 1.35, "learning_rate": 2.7481254326546724e-05, "loss": 0.0, "step": 69638 }, { "epoch": 1.35, "learning_rate": 2.748060736628949e-05, "loss": 0.0, "step": 69640 }, { "epoch": 1.35, "learning_rate": 2.747996040603226e-05, "loss": 0.0, "step": 69642 }, { "epoch": 1.35, "learning_rate": 2.7479313445775028e-05, "loss": 0.0, "step": 69644 }, { "epoch": 1.35, "learning_rate": 2.7478666485517797e-05, "loss": 0.0016, "step": 69646 }, { "epoch": 1.35, "learning_rate": 2.7478019525260562e-05, "loss": 0.0, "step": 69648 }, { "epoch": 1.35, "learning_rate": 2.7477372565003335e-05, "loss": 0.0, "step": 69650 }, { "epoch": 1.35, "learning_rate": 2.74767256047461e-05, "loss": 0.0, "step": 69652 }, { "epoch": 1.35, "learning_rate": 2.7476078644488873e-05, "loss": 0.0, "step": 69654 }, { "epoch": 1.35, "learning_rate": 2.747543168423164e-05, "loss": 0.0004, "step": 69656 }, { "epoch": 1.35, "learning_rate": 2.7474784723974408e-05, "loss": 0.0, "step": 69658 }, { "epoch": 1.35, "learning_rate": 2.7474137763717177e-05, "loss": 0.0413, "step": 69660 }, { "epoch": 1.35, "learning_rate": 2.7473490803459946e-05, "loss": 0.0001, "step": 69662 }, { "epoch": 1.35, "learning_rate": 2.747284384320271e-05, "loss": 0.0, "step": 69664 }, { "epoch": 1.35, "learning_rate": 2.7472196882945484e-05, "loss": 0.0, "step": 69666 }, { "epoch": 1.35, "learning_rate": 2.747154992268825e-05, "loss": 0.0, "step": 69668 }, { "epoch": 1.35, "learning_rate": 2.7470902962431015e-05, "loss": 0.0, "step": 69670 }, { "epoch": 1.35, "learning_rate": 2.7470256002173788e-05, "loss": 0.0, "step": 69672 }, { "epoch": 1.35, "learning_rate": 2.7469609041916554e-05, "loss": 0.0, "step": 69674 }, { "epoch": 1.35, "learning_rate": 2.7468962081659326e-05, "loss": 0.0, "step": 69676 }, { "epoch": 1.35, "learning_rate": 2.7468315121402095e-05, "loss": 0.0, "step": 69678 }, { "epoch": 1.35, "learning_rate": 2.746766816114486e-05, "loss": 0.0, "step": 69680 }, { "epoch": 1.35, "learning_rate": 2.7467021200887633e-05, "loss": 0.0, "step": 69682 }, { "epoch": 1.35, "learning_rate": 2.74663742406304e-05, "loss": 0.0063, "step": 69684 }, { "epoch": 1.35, "learning_rate": 2.7465727280373165e-05, "loss": 0.0001, "step": 69686 }, { "epoch": 1.35, "learning_rate": 2.7465080320115937e-05, "loss": 0.0, "step": 69688 }, { "epoch": 1.35, "learning_rate": 2.7464433359858703e-05, "loss": 0.0, "step": 69690 }, { "epoch": 1.35, "learning_rate": 2.7463786399601472e-05, "loss": 0.0, "step": 69692 }, { "epoch": 1.35, "learning_rate": 2.7463139439344244e-05, "loss": 0.0, "step": 69694 }, { "epoch": 1.35, "learning_rate": 2.746249247908701e-05, "loss": 0.0, "step": 69696 }, { "epoch": 1.35, "learning_rate": 2.7461845518829783e-05, "loss": 0.0, "step": 69698 }, { "epoch": 1.35, "learning_rate": 2.746119855857255e-05, "loss": 0.0, "step": 69700 }, { "epoch": 1.35, "learning_rate": 2.7460551598315314e-05, "loss": 0.0, "step": 69702 }, { "epoch": 1.35, "learning_rate": 2.7459904638058087e-05, "loss": 0.0, "step": 69704 }, { "epoch": 1.35, "learning_rate": 2.7459257677800852e-05, "loss": 0.0004, "step": 69706 }, { "epoch": 1.35, "learning_rate": 2.745861071754362e-05, "loss": 0.0, "step": 69708 }, { "epoch": 1.35, "learning_rate": 2.7457963757286394e-05, "loss": 0.0048, "step": 69710 }, { "epoch": 1.35, "learning_rate": 2.745731679702916e-05, "loss": 0.0, "step": 69712 }, { "epoch": 1.35, "learning_rate": 2.7456669836771932e-05, "loss": 0.0, "step": 69714 }, { "epoch": 1.35, "learning_rate": 2.7456022876514698e-05, "loss": 0.0, "step": 69716 }, { "epoch": 1.35, "learning_rate": 2.7455375916257463e-05, "loss": 0.0, "step": 69718 }, { "epoch": 1.35, "learning_rate": 2.7454728956000236e-05, "loss": 0.0, "step": 69720 }, { "epoch": 1.35, "learning_rate": 2.7454081995743e-05, "loss": 0.0075, "step": 69722 }, { "epoch": 1.35, "learning_rate": 2.745343503548577e-05, "loss": 0.0022, "step": 69724 }, { "epoch": 1.35, "learning_rate": 2.7452788075228543e-05, "loss": 0.0, "step": 69726 }, { "epoch": 1.35, "learning_rate": 2.745214111497131e-05, "loss": 0.0, "step": 69728 }, { "epoch": 1.35, "learning_rate": 2.7451494154714074e-05, "loss": 0.0, "step": 69730 }, { "epoch": 1.35, "learning_rate": 2.7450847194456847e-05, "loss": 0.0, "step": 69732 }, { "epoch": 1.35, "learning_rate": 2.7450200234199613e-05, "loss": 0.0008, "step": 69734 }, { "epoch": 1.35, "learning_rate": 2.7449553273942385e-05, "loss": 0.0, "step": 69736 }, { "epoch": 1.35, "learning_rate": 2.744890631368515e-05, "loss": 0.0002, "step": 69738 }, { "epoch": 1.35, "learning_rate": 2.744825935342792e-05, "loss": 0.0, "step": 69740 }, { "epoch": 1.35, "learning_rate": 2.7447612393170692e-05, "loss": 0.0, "step": 69742 }, { "epoch": 1.35, "learning_rate": 2.7446965432913458e-05, "loss": 0.0, "step": 69744 }, { "epoch": 1.35, "learning_rate": 2.7446318472656224e-05, "loss": 0.0, "step": 69746 }, { "epoch": 1.35, "learning_rate": 2.7445671512398996e-05, "loss": 0.0001, "step": 69748 }, { "epoch": 1.35, "learning_rate": 2.7445024552141762e-05, "loss": 0.0, "step": 69750 }, { "epoch": 1.35, "learning_rate": 2.7444377591884528e-05, "loss": 0.0, "step": 69752 }, { "epoch": 1.35, "learning_rate": 2.74437306316273e-05, "loss": 0.0011, "step": 69754 }, { "epoch": 1.35, "learning_rate": 2.744308367137007e-05, "loss": 0.0, "step": 69756 }, { "epoch": 1.35, "learning_rate": 2.744243671111284e-05, "loss": 0.0001, "step": 69758 }, { "epoch": 1.35, "learning_rate": 2.7441789750855607e-05, "loss": 0.0, "step": 69760 }, { "epoch": 1.35, "learning_rate": 2.7441142790598373e-05, "loss": 0.0, "step": 69762 }, { "epoch": 1.35, "learning_rate": 2.7440495830341146e-05, "loss": 0.0, "step": 69764 }, { "epoch": 1.35, "learning_rate": 2.743984887008391e-05, "loss": 0.0, "step": 69766 }, { "epoch": 1.35, "learning_rate": 2.7439201909826677e-05, "loss": 0.0, "step": 69768 }, { "epoch": 1.35, "learning_rate": 2.743855494956945e-05, "loss": 0.0, "step": 69770 }, { "epoch": 1.35, "learning_rate": 2.743790798931222e-05, "loss": 0.0, "step": 69772 }, { "epoch": 1.35, "learning_rate": 2.7437261029054988e-05, "loss": 0.0, "step": 69774 }, { "epoch": 1.35, "learning_rate": 2.7436614068797757e-05, "loss": 0.0, "step": 69776 }, { "epoch": 1.35, "learning_rate": 2.7435967108540522e-05, "loss": 0.0, "step": 69778 }, { "epoch": 1.35, "learning_rate": 2.7435320148283295e-05, "loss": 0.0, "step": 69780 }, { "epoch": 1.35, "learning_rate": 2.743467318802606e-05, "loss": 0.0002, "step": 69782 }, { "epoch": 1.35, "learning_rate": 2.7434026227768826e-05, "loss": 0.0012, "step": 69784 }, { "epoch": 1.35, "learning_rate": 2.74333792675116e-05, "loss": 0.0, "step": 69786 }, { "epoch": 1.35, "learning_rate": 2.7432732307254368e-05, "loss": 0.0, "step": 69788 }, { "epoch": 1.35, "learning_rate": 2.7432085346997133e-05, "loss": 0.0, "step": 69790 }, { "epoch": 1.35, "learning_rate": 2.7431438386739906e-05, "loss": 0.0, "step": 69792 }, { "epoch": 1.35, "learning_rate": 2.743079142648267e-05, "loss": 0.0, "step": 69794 }, { "epoch": 1.35, "learning_rate": 2.7430144466225444e-05, "loss": 0.0002, "step": 69796 }, { "epoch": 1.35, "learning_rate": 2.742949750596821e-05, "loss": 0.0, "step": 69798 }, { "epoch": 1.35, "learning_rate": 2.7428850545710975e-05, "loss": 0.0, "step": 69800 }, { "epoch": 1.35, "learning_rate": 2.7428203585453748e-05, "loss": 0.0001, "step": 69802 }, { "epoch": 1.35, "learning_rate": 2.7427556625196517e-05, "loss": 0.0, "step": 69804 }, { "epoch": 1.35, "learning_rate": 2.7426909664939283e-05, "loss": 0.0, "step": 69806 }, { "epoch": 1.35, "learning_rate": 2.7426262704682055e-05, "loss": 0.0, "step": 69808 }, { "epoch": 1.35, "learning_rate": 2.742561574442482e-05, "loss": 0.0, "step": 69810 }, { "epoch": 1.35, "learning_rate": 2.7424968784167587e-05, "loss": 0.0, "step": 69812 }, { "epoch": 1.36, "learning_rate": 2.742432182391036e-05, "loss": 0.0, "step": 69814 }, { "epoch": 1.36, "learning_rate": 2.7423674863653125e-05, "loss": 0.0, "step": 69816 }, { "epoch": 1.36, "learning_rate": 2.7423027903395897e-05, "loss": 0.0, "step": 69818 }, { "epoch": 1.36, "learning_rate": 2.7422380943138666e-05, "loss": 0.0001, "step": 69820 }, { "epoch": 1.36, "learning_rate": 2.7421733982881432e-05, "loss": 0.0109, "step": 69822 }, { "epoch": 1.36, "learning_rate": 2.7421087022624204e-05, "loss": 0.0, "step": 69824 }, { "epoch": 1.36, "learning_rate": 2.742044006236697e-05, "loss": 0.0, "step": 69826 }, { "epoch": 1.36, "learning_rate": 2.7419793102109736e-05, "loss": 0.0, "step": 69828 }, { "epoch": 1.36, "learning_rate": 2.741914614185251e-05, "loss": 0.0, "step": 69830 }, { "epoch": 1.36, "learning_rate": 2.7418499181595274e-05, "loss": 0.0, "step": 69832 }, { "epoch": 1.36, "learning_rate": 2.7417852221338043e-05, "loss": 0.0, "step": 69834 }, { "epoch": 1.36, "learning_rate": 2.7417205261080812e-05, "loss": 0.0, "step": 69836 }, { "epoch": 1.36, "learning_rate": 2.741655830082358e-05, "loss": 0.0058, "step": 69838 }, { "epoch": 1.36, "learning_rate": 2.7415911340566354e-05, "loss": 0.0, "step": 69840 }, { "epoch": 1.36, "learning_rate": 2.741526438030912e-05, "loss": 0.0, "step": 69842 }, { "epoch": 1.36, "learning_rate": 2.7414617420051885e-05, "loss": 0.0, "step": 69844 }, { "epoch": 1.36, "learning_rate": 2.7413970459794658e-05, "loss": 0.0, "step": 69846 }, { "epoch": 1.36, "learning_rate": 2.7413323499537423e-05, "loss": 0.0, "step": 69848 }, { "epoch": 1.36, "learning_rate": 2.7412676539280192e-05, "loss": 0.0005, "step": 69850 }, { "epoch": 1.36, "learning_rate": 2.741202957902296e-05, "loss": 0.0, "step": 69852 }, { "epoch": 1.36, "learning_rate": 2.741138261876573e-05, "loss": 0.0, "step": 69854 }, { "epoch": 1.36, "learning_rate": 2.7410735658508503e-05, "loss": 0.0024, "step": 69856 }, { "epoch": 1.36, "learning_rate": 2.741008869825127e-05, "loss": 0.0, "step": 69858 }, { "epoch": 1.36, "learning_rate": 2.7409441737994034e-05, "loss": 0.0, "step": 69860 }, { "epoch": 1.36, "learning_rate": 2.7408794777736807e-05, "loss": 0.0028, "step": 69862 }, { "epoch": 1.36, "learning_rate": 2.7408147817479573e-05, "loss": 0.0, "step": 69864 }, { "epoch": 1.36, "learning_rate": 2.7407500857222342e-05, "loss": 0.0, "step": 69866 }, { "epoch": 1.36, "learning_rate": 2.740685389696511e-05, "loss": 0.0, "step": 69868 }, { "epoch": 1.36, "learning_rate": 2.740620693670788e-05, "loss": 0.0, "step": 69870 }, { "epoch": 1.36, "learning_rate": 2.7405559976450646e-05, "loss": 0.0, "step": 69872 }, { "epoch": 1.36, "learning_rate": 2.7404913016193418e-05, "loss": 0.0, "step": 69874 }, { "epoch": 1.36, "learning_rate": 2.7404266055936184e-05, "loss": 0.0161, "step": 69876 }, { "epoch": 1.36, "learning_rate": 2.7403619095678956e-05, "loss": 0.0, "step": 69878 }, { "epoch": 1.36, "learning_rate": 2.7402972135421722e-05, "loss": 0.0, "step": 69880 }, { "epoch": 1.36, "learning_rate": 2.740232517516449e-05, "loss": 0.0, "step": 69882 }, { "epoch": 1.36, "learning_rate": 2.740167821490726e-05, "loss": 0.0, "step": 69884 }, { "epoch": 1.36, "learning_rate": 2.740103125465003e-05, "loss": 0.0, "step": 69886 }, { "epoch": 1.36, "learning_rate": 2.7400384294392795e-05, "loss": 0.0, "step": 69888 }, { "epoch": 1.36, "learning_rate": 2.7399737334135567e-05, "loss": 0.0, "step": 69890 }, { "epoch": 1.36, "learning_rate": 2.7399090373878333e-05, "loss": 0.0, "step": 69892 }, { "epoch": 1.36, "learning_rate": 2.73984434136211e-05, "loss": 0.0, "step": 69894 }, { "epoch": 1.36, "learning_rate": 2.739779645336387e-05, "loss": 0.0, "step": 69896 }, { "epoch": 1.36, "learning_rate": 2.739714949310664e-05, "loss": 0.0, "step": 69898 }, { "epoch": 1.36, "learning_rate": 2.739650253284941e-05, "loss": 0.0, "step": 69900 }, { "epoch": 1.36, "learning_rate": 2.739585557259218e-05, "loss": 0.0, "step": 69902 }, { "epoch": 1.36, "learning_rate": 2.7395208612334944e-05, "loss": 0.0033, "step": 69904 }, { "epoch": 1.36, "learning_rate": 2.7394561652077717e-05, "loss": 0.0, "step": 69906 }, { "epoch": 1.36, "learning_rate": 2.7393914691820482e-05, "loss": 0.0, "step": 69908 }, { "epoch": 1.36, "learning_rate": 2.7393267731563248e-05, "loss": 0.0, "step": 69910 }, { "epoch": 1.36, "learning_rate": 2.739262077130602e-05, "loss": 0.0005, "step": 69912 }, { "epoch": 1.36, "learning_rate": 2.7391973811048786e-05, "loss": 0.0, "step": 69914 }, { "epoch": 1.36, "learning_rate": 2.739132685079156e-05, "loss": 0.0, "step": 69916 }, { "epoch": 1.36, "learning_rate": 2.7390679890534328e-05, "loss": 0.0, "step": 69918 }, { "epoch": 1.36, "learning_rate": 2.7390032930277093e-05, "loss": 0.0, "step": 69920 }, { "epoch": 1.36, "learning_rate": 2.7389385970019866e-05, "loss": 0.0, "step": 69922 }, { "epoch": 1.36, "learning_rate": 2.738873900976263e-05, "loss": 0.0, "step": 69924 }, { "epoch": 1.36, "learning_rate": 2.7388092049505397e-05, "loss": 0.0, "step": 69926 }, { "epoch": 1.36, "learning_rate": 2.738744508924817e-05, "loss": 0.0, "step": 69928 }, { "epoch": 1.36, "learning_rate": 2.7386798128990935e-05, "loss": 0.0, "step": 69930 }, { "epoch": 1.36, "learning_rate": 2.7386151168733705e-05, "loss": 0.0, "step": 69932 }, { "epoch": 1.36, "learning_rate": 2.7385504208476477e-05, "loss": 0.0, "step": 69934 }, { "epoch": 1.36, "learning_rate": 2.7384857248219243e-05, "loss": 0.0002, "step": 69936 }, { "epoch": 1.36, "learning_rate": 2.7384210287962015e-05, "loss": 0.0006, "step": 69938 }, { "epoch": 1.36, "learning_rate": 2.738356332770478e-05, "loss": 0.0, "step": 69940 }, { "epoch": 1.36, "learning_rate": 2.7382916367447547e-05, "loss": 0.0, "step": 69942 }, { "epoch": 1.36, "learning_rate": 2.738226940719032e-05, "loss": 0.0002, "step": 69944 }, { "epoch": 1.36, "learning_rate": 2.7381622446933085e-05, "loss": 0.0, "step": 69946 }, { "epoch": 1.36, "learning_rate": 2.7380975486675854e-05, "loss": 0.0, "step": 69948 }, { "epoch": 1.36, "learning_rate": 2.7380328526418626e-05, "loss": 0.0, "step": 69950 }, { "epoch": 1.36, "learning_rate": 2.7379681566161392e-05, "loss": 0.0125, "step": 69952 }, { "epoch": 1.36, "learning_rate": 2.7379034605904158e-05, "loss": 0.0012, "step": 69954 }, { "epoch": 1.36, "learning_rate": 2.737838764564693e-05, "loss": 0.0, "step": 69956 }, { "epoch": 1.36, "learning_rate": 2.7377740685389696e-05, "loss": 0.0, "step": 69958 }, { "epoch": 1.36, "learning_rate": 2.737709372513247e-05, "loss": 0.0, "step": 69960 }, { "epoch": 1.36, "learning_rate": 2.7376446764875234e-05, "loss": 0.0009, "step": 69962 }, { "epoch": 1.36, "learning_rate": 2.7375799804618003e-05, "loss": 0.0, "step": 69964 }, { "epoch": 1.36, "learning_rate": 2.7375152844360776e-05, "loss": 0.0003, "step": 69966 }, { "epoch": 1.36, "learning_rate": 2.737450588410354e-05, "loss": 0.0, "step": 69968 }, { "epoch": 1.36, "learning_rate": 2.7373858923846307e-05, "loss": 0.0, "step": 69970 }, { "epoch": 1.36, "learning_rate": 2.737321196358908e-05, "loss": 0.0, "step": 69972 }, { "epoch": 1.36, "learning_rate": 2.7372565003331845e-05, "loss": 0.0, "step": 69974 }, { "epoch": 1.36, "learning_rate": 2.737191804307461e-05, "loss": 0.0003, "step": 69976 }, { "epoch": 1.36, "learning_rate": 2.7371271082817383e-05, "loss": 0.0012, "step": 69978 }, { "epoch": 1.36, "learning_rate": 2.7370624122560152e-05, "loss": 0.0001, "step": 69980 }, { "epoch": 1.36, "learning_rate": 2.7369977162302925e-05, "loss": 0.0, "step": 69982 }, { "epoch": 1.36, "learning_rate": 2.736933020204569e-05, "loss": 0.0002, "step": 69984 }, { "epoch": 1.36, "learning_rate": 2.7368683241788456e-05, "loss": 0.0186, "step": 69986 }, { "epoch": 1.36, "learning_rate": 2.736803628153123e-05, "loss": 0.0003, "step": 69988 }, { "epoch": 1.36, "learning_rate": 2.7367389321273994e-05, "loss": 0.0, "step": 69990 }, { "epoch": 1.36, "learning_rate": 2.736674236101676e-05, "loss": 0.0, "step": 69992 }, { "epoch": 1.36, "learning_rate": 2.7366095400759533e-05, "loss": 0.0001, "step": 69994 }, { "epoch": 1.36, "learning_rate": 2.73654484405023e-05, "loss": 0.0, "step": 69996 }, { "epoch": 1.36, "learning_rate": 2.7364801480245074e-05, "loss": 0.0, "step": 69998 }, { "epoch": 1.36, "learning_rate": 2.736415451998784e-05, "loss": 0.0009, "step": 70000 }, { "epoch": 1.36, "learning_rate": 2.7363507559730606e-05, "loss": 0.0002, "step": 70002 }, { "epoch": 1.36, "learning_rate": 2.7362860599473378e-05, "loss": 0.0, "step": 70004 }, { "epoch": 1.36, "learning_rate": 2.7362213639216144e-05, "loss": 0.0007, "step": 70006 }, { "epoch": 1.36, "learning_rate": 2.736156667895891e-05, "loss": 0.0, "step": 70008 }, { "epoch": 1.36, "learning_rate": 2.7360919718701682e-05, "loss": 0.0018, "step": 70010 }, { "epoch": 1.36, "learning_rate": 2.736027275844445e-05, "loss": 0.0016, "step": 70012 }, { "epoch": 1.36, "learning_rate": 2.7359625798187217e-05, "loss": 0.0, "step": 70014 }, { "epoch": 1.36, "learning_rate": 2.735897883792999e-05, "loss": 0.0, "step": 70016 }, { "epoch": 1.36, "learning_rate": 2.7358331877672755e-05, "loss": 0.0, "step": 70018 }, { "epoch": 1.36, "learning_rate": 2.7357684917415527e-05, "loss": 0.0, "step": 70020 }, { "epoch": 1.36, "learning_rate": 2.7357037957158293e-05, "loss": 0.0, "step": 70022 }, { "epoch": 1.36, "learning_rate": 2.735639099690106e-05, "loss": 0.0, "step": 70024 }, { "epoch": 1.36, "learning_rate": 2.735574403664383e-05, "loss": 0.0002, "step": 70026 }, { "epoch": 1.36, "learning_rate": 2.73550970763866e-05, "loss": 0.0, "step": 70028 }, { "epoch": 1.36, "learning_rate": 2.7354450116129366e-05, "loss": 0.0, "step": 70030 }, { "epoch": 1.36, "learning_rate": 2.735380315587214e-05, "loss": 0.0, "step": 70032 }, { "epoch": 1.36, "learning_rate": 2.7353156195614904e-05, "loss": 0.0, "step": 70034 }, { "epoch": 1.36, "learning_rate": 2.735250923535767e-05, "loss": 0.0008, "step": 70036 }, { "epoch": 1.36, "learning_rate": 2.7351862275100442e-05, "loss": 0.0, "step": 70038 }, { "epoch": 1.36, "learning_rate": 2.7351215314843208e-05, "loss": 0.0, "step": 70040 }, { "epoch": 1.36, "learning_rate": 2.735056835458598e-05, "loss": 0.0002, "step": 70042 }, { "epoch": 1.36, "learning_rate": 2.734992139432875e-05, "loss": 0.0, "step": 70044 }, { "epoch": 1.36, "learning_rate": 2.7349274434071515e-05, "loss": 0.0, "step": 70046 }, { "epoch": 1.36, "learning_rate": 2.7348627473814288e-05, "loss": 0.0, "step": 70048 }, { "epoch": 1.36, "learning_rate": 2.7347980513557053e-05, "loss": 0.0009, "step": 70050 }, { "epoch": 1.36, "learning_rate": 2.734733355329982e-05, "loss": 0.0, "step": 70052 }, { "epoch": 1.36, "learning_rate": 2.734668659304259e-05, "loss": 0.0, "step": 70054 }, { "epoch": 1.36, "learning_rate": 2.7346039632785357e-05, "loss": 0.0, "step": 70056 }, { "epoch": 1.36, "learning_rate": 2.734539267252813e-05, "loss": 0.0, "step": 70058 }, { "epoch": 1.36, "learning_rate": 2.73447457122709e-05, "loss": 0.0, "step": 70060 }, { "epoch": 1.36, "learning_rate": 2.7344098752013665e-05, "loss": 0.0002, "step": 70062 }, { "epoch": 1.36, "learning_rate": 2.7343451791756437e-05, "loss": 0.0034, "step": 70064 }, { "epoch": 1.36, "learning_rate": 2.7342804831499203e-05, "loss": 0.0, "step": 70066 }, { "epoch": 1.36, "learning_rate": 2.734215787124197e-05, "loss": 0.0, "step": 70068 }, { "epoch": 1.36, "learning_rate": 2.734151091098474e-05, "loss": 0.0, "step": 70070 }, { "epoch": 1.36, "learning_rate": 2.7340863950727507e-05, "loss": 0.0, "step": 70072 }, { "epoch": 1.36, "learning_rate": 2.7340216990470276e-05, "loss": 0.0, "step": 70074 }, { "epoch": 1.36, "learning_rate": 2.7339570030213045e-05, "loss": 0.0043, "step": 70076 }, { "epoch": 1.36, "learning_rate": 2.7338923069955814e-05, "loss": 0.0, "step": 70078 }, { "epoch": 1.36, "learning_rate": 2.7338276109698586e-05, "loss": 0.0, "step": 70080 }, { "epoch": 1.36, "learning_rate": 2.7337629149441352e-05, "loss": 0.0, "step": 70082 }, { "epoch": 1.36, "learning_rate": 2.7336982189184118e-05, "loss": 0.0, "step": 70084 }, { "epoch": 1.36, "learning_rate": 2.733633522892689e-05, "loss": 0.0, "step": 70086 }, { "epoch": 1.36, "learning_rate": 2.7335688268669656e-05, "loss": 0.0, "step": 70088 }, { "epoch": 1.36, "learning_rate": 2.7335041308412425e-05, "loss": 0.0, "step": 70090 }, { "epoch": 1.36, "learning_rate": 2.7334394348155194e-05, "loss": 0.0003, "step": 70092 }, { "epoch": 1.36, "learning_rate": 2.7333747387897963e-05, "loss": 0.0011, "step": 70094 }, { "epoch": 1.36, "learning_rate": 2.733310042764073e-05, "loss": 0.0, "step": 70096 }, { "epoch": 1.36, "learning_rate": 2.73324534673835e-05, "loss": 0.0, "step": 70098 }, { "epoch": 1.36, "learning_rate": 2.7331806507126267e-05, "loss": 0.0, "step": 70100 }, { "epoch": 1.36, "learning_rate": 2.733115954686904e-05, "loss": 0.0, "step": 70102 }, { "epoch": 1.36, "learning_rate": 2.7330512586611805e-05, "loss": 0.0, "step": 70104 }, { "epoch": 1.36, "learning_rate": 2.7329865626354574e-05, "loss": 0.0, "step": 70106 }, { "epoch": 1.36, "learning_rate": 2.7329218666097343e-05, "loss": 0.0, "step": 70108 }, { "epoch": 1.36, "learning_rate": 2.7328571705840112e-05, "loss": 0.0001, "step": 70110 }, { "epoch": 1.36, "learning_rate": 2.7327924745582878e-05, "loss": 0.0, "step": 70112 }, { "epoch": 1.36, "learning_rate": 2.732727778532565e-05, "loss": 0.0002, "step": 70114 }, { "epoch": 1.36, "learning_rate": 2.7326630825068416e-05, "loss": 0.0, "step": 70116 }, { "epoch": 1.36, "learning_rate": 2.7325983864811182e-05, "loss": 0.0009, "step": 70118 }, { "epoch": 1.36, "learning_rate": 2.7325336904553954e-05, "loss": 0.0, "step": 70120 }, { "epoch": 1.36, "learning_rate": 2.7324689944296724e-05, "loss": 0.0002, "step": 70122 }, { "epoch": 1.36, "learning_rate": 2.7324042984039493e-05, "loss": 0.0086, "step": 70124 }, { "epoch": 1.36, "learning_rate": 2.732339602378226e-05, "loss": 0.0, "step": 70126 }, { "epoch": 1.36, "learning_rate": 2.7322749063525027e-05, "loss": 0.0, "step": 70128 }, { "epoch": 1.36, "learning_rate": 2.73221021032678e-05, "loss": 0.0217, "step": 70130 }, { "epoch": 1.36, "learning_rate": 2.7321455143010566e-05, "loss": 0.0, "step": 70132 }, { "epoch": 1.36, "learning_rate": 2.732080818275333e-05, "loss": 0.0, "step": 70134 }, { "epoch": 1.36, "learning_rate": 2.7320161222496104e-05, "loss": 0.0, "step": 70136 }, { "epoch": 1.36, "learning_rate": 2.7319514262238873e-05, "loss": 0.0, "step": 70138 }, { "epoch": 1.36, "learning_rate": 2.7318867301981642e-05, "loss": 0.0093, "step": 70140 }, { "epoch": 1.36, "learning_rate": 2.731822034172441e-05, "loss": 0.0, "step": 70142 }, { "epoch": 1.36, "learning_rate": 2.7317573381467177e-05, "loss": 0.0, "step": 70144 }, { "epoch": 1.36, "learning_rate": 2.731692642120995e-05, "loss": 0.0, "step": 70146 }, { "epoch": 1.36, "learning_rate": 2.7316279460952715e-05, "loss": 0.0, "step": 70148 }, { "epoch": 1.36, "learning_rate": 2.731563250069548e-05, "loss": 0.0, "step": 70150 }, { "epoch": 1.36, "learning_rate": 2.7314985540438253e-05, "loss": 0.0011, "step": 70152 }, { "epoch": 1.36, "learning_rate": 2.731433858018102e-05, "loss": 0.0, "step": 70154 }, { "epoch": 1.36, "learning_rate": 2.7313691619923788e-05, "loss": 0.0001, "step": 70156 }, { "epoch": 1.36, "learning_rate": 2.731304465966656e-05, "loss": 0.0, "step": 70158 }, { "epoch": 1.36, "learning_rate": 2.7312397699409326e-05, "loss": 0.0, "step": 70160 }, { "epoch": 1.36, "learning_rate": 2.73117507391521e-05, "loss": 0.0, "step": 70162 }, { "epoch": 1.36, "learning_rate": 2.7311103778894864e-05, "loss": 0.0, "step": 70164 }, { "epoch": 1.36, "learning_rate": 2.731045681863763e-05, "loss": 0.0, "step": 70166 }, { "epoch": 1.36, "learning_rate": 2.7309809858380402e-05, "loss": 0.0, "step": 70168 }, { "epoch": 1.36, "learning_rate": 2.7309162898123168e-05, "loss": 0.0002, "step": 70170 }, { "epoch": 1.36, "learning_rate": 2.7308515937865937e-05, "loss": 0.0, "step": 70172 }, { "epoch": 1.36, "learning_rate": 2.730786897760871e-05, "loss": 0.0, "step": 70174 }, { "epoch": 1.36, "learning_rate": 2.7307222017351475e-05, "loss": 0.0001, "step": 70176 }, { "epoch": 1.36, "learning_rate": 2.730657505709424e-05, "loss": 0.0, "step": 70178 }, { "epoch": 1.36, "learning_rate": 2.7305928096837013e-05, "loss": 0.0, "step": 70180 }, { "epoch": 1.36, "learning_rate": 2.730528113657978e-05, "loss": 0.0, "step": 70182 }, { "epoch": 1.36, "learning_rate": 2.730463417632255e-05, "loss": 0.011, "step": 70184 }, { "epoch": 1.36, "learning_rate": 2.7303987216065317e-05, "loss": 0.0, "step": 70186 }, { "epoch": 1.36, "learning_rate": 2.7303340255808086e-05, "loss": 0.0, "step": 70188 }, { "epoch": 1.36, "learning_rate": 2.730269329555086e-05, "loss": 0.0, "step": 70190 }, { "epoch": 1.36, "learning_rate": 2.7302046335293625e-05, "loss": 0.0001, "step": 70192 }, { "epoch": 1.36, "learning_rate": 2.730139937503639e-05, "loss": 0.0, "step": 70194 }, { "epoch": 1.36, "learning_rate": 2.7300752414779163e-05, "loss": 0.0001, "step": 70196 }, { "epoch": 1.36, "learning_rate": 2.730010545452193e-05, "loss": 0.0, "step": 70198 }, { "epoch": 1.36, "learning_rate": 2.72994584942647e-05, "loss": 0.0, "step": 70200 }, { "epoch": 1.36, "learning_rate": 2.7298811534007467e-05, "loss": 0.0, "step": 70202 }, { "epoch": 1.36, "learning_rate": 2.7298164573750236e-05, "loss": 0.0, "step": 70204 }, { "epoch": 1.36, "learning_rate": 2.7297517613493008e-05, "loss": 0.0, "step": 70206 }, { "epoch": 1.36, "learning_rate": 2.7296870653235774e-05, "loss": 0.0, "step": 70208 }, { "epoch": 1.36, "learning_rate": 2.729622369297854e-05, "loss": 0.0, "step": 70210 }, { "epoch": 1.36, "learning_rate": 2.7295576732721312e-05, "loss": 0.0, "step": 70212 }, { "epoch": 1.36, "learning_rate": 2.7294929772464078e-05, "loss": 0.0, "step": 70214 }, { "epoch": 1.36, "learning_rate": 2.7294282812206843e-05, "loss": 0.0, "step": 70216 }, { "epoch": 1.36, "learning_rate": 2.7293635851949616e-05, "loss": 0.0, "step": 70218 }, { "epoch": 1.36, "learning_rate": 2.7292988891692385e-05, "loss": 0.0, "step": 70220 }, { "epoch": 1.36, "learning_rate": 2.7292341931435157e-05, "loss": 0.0, "step": 70222 }, { "epoch": 1.36, "learning_rate": 2.7291694971177923e-05, "loss": 0.0, "step": 70224 }, { "epoch": 1.36, "learning_rate": 2.729104801092069e-05, "loss": 0.0018, "step": 70226 }, { "epoch": 1.36, "learning_rate": 2.729040105066346e-05, "loss": 0.0, "step": 70228 }, { "epoch": 1.36, "learning_rate": 2.7289754090406227e-05, "loss": 0.0202, "step": 70230 }, { "epoch": 1.36, "learning_rate": 2.7289107130148993e-05, "loss": 0.0, "step": 70232 }, { "epoch": 1.36, "learning_rate": 2.7288460169891765e-05, "loss": 0.0, "step": 70234 }, { "epoch": 1.36, "learning_rate": 2.7287813209634534e-05, "loss": 0.0, "step": 70236 }, { "epoch": 1.36, "learning_rate": 2.72871662493773e-05, "loss": 0.0, "step": 70238 }, { "epoch": 1.36, "learning_rate": 2.7286519289120072e-05, "loss": 0.0115, "step": 70240 }, { "epoch": 1.36, "learning_rate": 2.7285872328862838e-05, "loss": 0.0, "step": 70242 }, { "epoch": 1.36, "learning_rate": 2.728522536860561e-05, "loss": 0.0, "step": 70244 }, { "epoch": 1.36, "learning_rate": 2.7284578408348376e-05, "loss": 0.0, "step": 70246 }, { "epoch": 1.36, "learning_rate": 2.7283931448091142e-05, "loss": 0.0, "step": 70248 }, { "epoch": 1.36, "learning_rate": 2.7283284487833914e-05, "loss": 0.0, "step": 70250 }, { "epoch": 1.36, "learning_rate": 2.7282637527576684e-05, "loss": 0.0, "step": 70252 }, { "epoch": 1.36, "learning_rate": 2.728199056731945e-05, "loss": 0.0, "step": 70254 }, { "epoch": 1.36, "learning_rate": 2.728134360706222e-05, "loss": 0.0, "step": 70256 }, { "epoch": 1.36, "learning_rate": 2.7280696646804987e-05, "loss": 0.0006, "step": 70258 }, { "epoch": 1.36, "learning_rate": 2.7280049686547753e-05, "loss": 0.0002, "step": 70260 }, { "epoch": 1.36, "learning_rate": 2.7279402726290526e-05, "loss": 0.0, "step": 70262 }, { "epoch": 1.36, "learning_rate": 2.727875576603329e-05, "loss": 0.0001, "step": 70264 }, { "epoch": 1.36, "learning_rate": 2.7278108805776064e-05, "loss": 0.0, "step": 70266 }, { "epoch": 1.36, "learning_rate": 2.7277461845518833e-05, "loss": 0.0, "step": 70268 }, { "epoch": 1.36, "learning_rate": 2.72768148852616e-05, "loss": 0.0, "step": 70270 }, { "epoch": 1.36, "learning_rate": 2.727616792500437e-05, "loss": 0.0043, "step": 70272 }, { "epoch": 1.36, "learning_rate": 2.7275520964747137e-05, "loss": 0.0, "step": 70274 }, { "epoch": 1.36, "learning_rate": 2.7274874004489902e-05, "loss": 0.0, "step": 70276 }, { "epoch": 1.36, "learning_rate": 2.7274227044232675e-05, "loss": 0.0, "step": 70278 }, { "epoch": 1.36, "learning_rate": 2.727358008397544e-05, "loss": 0.0018, "step": 70280 }, { "epoch": 1.36, "learning_rate": 2.7272933123718213e-05, "loss": 0.0, "step": 70282 }, { "epoch": 1.36, "learning_rate": 2.7272286163460982e-05, "loss": 0.0, "step": 70284 }, { "epoch": 1.36, "learning_rate": 2.7271639203203748e-05, "loss": 0.0, "step": 70286 }, { "epoch": 1.36, "learning_rate": 2.727099224294652e-05, "loss": 0.0, "step": 70288 }, { "epoch": 1.36, "learning_rate": 2.7270345282689286e-05, "loss": 0.0, "step": 70290 }, { "epoch": 1.36, "learning_rate": 2.726969832243205e-05, "loss": 0.0, "step": 70292 }, { "epoch": 1.36, "learning_rate": 2.7269051362174824e-05, "loss": 0.005, "step": 70294 }, { "epoch": 1.36, "learning_rate": 2.726840440191759e-05, "loss": 0.0, "step": 70296 }, { "epoch": 1.36, "learning_rate": 2.726775744166036e-05, "loss": 0.0004, "step": 70298 }, { "epoch": 1.36, "learning_rate": 2.726711048140313e-05, "loss": 0.0, "step": 70300 }, { "epoch": 1.36, "learning_rate": 2.7266463521145897e-05, "loss": 0.0005, "step": 70302 }, { "epoch": 1.36, "learning_rate": 2.726581656088867e-05, "loss": 0.0242, "step": 70304 }, { "epoch": 1.36, "learning_rate": 2.7265169600631435e-05, "loss": 0.0, "step": 70306 }, { "epoch": 1.36, "learning_rate": 2.72645226403742e-05, "loss": 0.0, "step": 70308 }, { "epoch": 1.36, "learning_rate": 2.7263875680116973e-05, "loss": 0.0, "step": 70310 }, { "epoch": 1.36, "learning_rate": 2.726322871985974e-05, "loss": 0.0, "step": 70312 }, { "epoch": 1.36, "learning_rate": 2.7262581759602508e-05, "loss": 0.0073, "step": 70314 }, { "epoch": 1.36, "learning_rate": 2.7261934799345277e-05, "loss": 0.0, "step": 70316 }, { "epoch": 1.36, "learning_rate": 2.7261287839088046e-05, "loss": 0.0046, "step": 70318 }, { "epoch": 1.36, "learning_rate": 2.7260640878830812e-05, "loss": 0.0, "step": 70320 }, { "epoch": 1.36, "learning_rate": 2.7259993918573585e-05, "loss": 0.0, "step": 70322 }, { "epoch": 1.36, "learning_rate": 2.725934695831635e-05, "loss": 0.0, "step": 70324 }, { "epoch": 1.36, "learning_rate": 2.7258699998059123e-05, "loss": 0.0, "step": 70326 }, { "epoch": 1.36, "learning_rate": 2.725805303780189e-05, "loss": 0.004, "step": 70328 }, { "epoch": 1.37, "learning_rate": 2.7257406077544657e-05, "loss": 0.0, "step": 70330 }, { "epoch": 1.37, "learning_rate": 2.7256759117287427e-05, "loss": 0.0, "step": 70332 }, { "epoch": 1.37, "learning_rate": 2.7256112157030196e-05, "loss": 0.0, "step": 70334 }, { "epoch": 1.37, "learning_rate": 2.725546519677296e-05, "loss": 0.0, "step": 70336 }, { "epoch": 1.37, "learning_rate": 2.7254818236515734e-05, "loss": 0.0, "step": 70338 }, { "epoch": 1.37, "learning_rate": 2.72541712762585e-05, "loss": 0.0002, "step": 70340 }, { "epoch": 1.37, "learning_rate": 2.7253524316001272e-05, "loss": 0.0, "step": 70342 }, { "epoch": 1.37, "learning_rate": 2.7252877355744038e-05, "loss": 0.0, "step": 70344 }, { "epoch": 1.37, "learning_rate": 2.7252230395486807e-05, "loss": 0.0, "step": 70346 }, { "epoch": 1.37, "learning_rate": 2.7251583435229576e-05, "loss": 0.0014, "step": 70348 }, { "epoch": 1.37, "learning_rate": 2.7250936474972345e-05, "loss": 0.0, "step": 70350 }, { "epoch": 1.37, "learning_rate": 2.725028951471511e-05, "loss": 0.0, "step": 70352 }, { "epoch": 1.37, "learning_rate": 2.7249642554457883e-05, "loss": 0.0, "step": 70354 }, { "epoch": 1.37, "learning_rate": 2.724899559420065e-05, "loss": 0.0, "step": 70356 }, { "epoch": 1.37, "learning_rate": 2.7248348633943414e-05, "loss": 0.0, "step": 70358 }, { "epoch": 1.37, "learning_rate": 2.7247701673686187e-05, "loss": 0.0, "step": 70360 }, { "epoch": 1.37, "learning_rate": 2.7247054713428956e-05, "loss": 0.0016, "step": 70362 }, { "epoch": 1.37, "learning_rate": 2.7246407753171725e-05, "loss": 0.0033, "step": 70364 }, { "epoch": 1.37, "learning_rate": 2.7245760792914494e-05, "loss": 0.0, "step": 70366 }, { "epoch": 1.37, "learning_rate": 2.724511383265726e-05, "loss": 0.0, "step": 70368 }, { "epoch": 1.37, "learning_rate": 2.7244466872400032e-05, "loss": 0.0, "step": 70370 }, { "epoch": 1.37, "learning_rate": 2.7243819912142798e-05, "loss": 0.0101, "step": 70372 }, { "epoch": 1.37, "learning_rate": 2.7243172951885564e-05, "loss": 0.0, "step": 70374 }, { "epoch": 1.37, "learning_rate": 2.7242525991628336e-05, "loss": 0.0, "step": 70376 }, { "epoch": 1.37, "learning_rate": 2.7241879031371105e-05, "loss": 0.0, "step": 70378 }, { "epoch": 1.37, "learning_rate": 2.724123207111387e-05, "loss": 0.0, "step": 70380 }, { "epoch": 1.37, "learning_rate": 2.7240585110856643e-05, "loss": 0.0, "step": 70382 }, { "epoch": 1.37, "learning_rate": 2.723993815059941e-05, "loss": 0.0001, "step": 70384 }, { "epoch": 1.37, "learning_rate": 2.723929119034218e-05, "loss": 0.0, "step": 70386 }, { "epoch": 1.37, "learning_rate": 2.7238644230084947e-05, "loss": 0.0, "step": 70388 }, { "epoch": 1.37, "learning_rate": 2.7237997269827713e-05, "loss": 0.0, "step": 70390 }, { "epoch": 1.37, "learning_rate": 2.7237350309570486e-05, "loss": 0.0, "step": 70392 }, { "epoch": 1.37, "learning_rate": 2.723670334931325e-05, "loss": 0.0001, "step": 70394 }, { "epoch": 1.37, "learning_rate": 2.723605638905602e-05, "loss": 0.0, "step": 70396 }, { "epoch": 1.37, "learning_rate": 2.7235409428798793e-05, "loss": 0.0, "step": 70398 }, { "epoch": 1.37, "learning_rate": 2.723476246854156e-05, "loss": 0.0, "step": 70400 }, { "epoch": 1.37, "learning_rate": 2.7234115508284324e-05, "loss": 0.0019, "step": 70402 }, { "epoch": 1.37, "learning_rate": 2.7233468548027097e-05, "loss": 0.0, "step": 70404 }, { "epoch": 1.37, "learning_rate": 2.7232821587769862e-05, "loss": 0.0003, "step": 70406 }, { "epoch": 1.37, "learning_rate": 2.7232174627512635e-05, "loss": 0.0001, "step": 70408 }, { "epoch": 1.37, "learning_rate": 2.72315276672554e-05, "loss": 0.0, "step": 70410 }, { "epoch": 1.37, "learning_rate": 2.723088070699817e-05, "loss": 0.0041, "step": 70412 }, { "epoch": 1.37, "learning_rate": 2.7230233746740942e-05, "loss": 0.0, "step": 70414 }, { "epoch": 1.37, "learning_rate": 2.7229586786483708e-05, "loss": 0.0001, "step": 70416 }, { "epoch": 1.37, "learning_rate": 2.7228939826226473e-05, "loss": 0.0, "step": 70418 }, { "epoch": 1.37, "learning_rate": 2.7228292865969246e-05, "loss": 0.0002, "step": 70420 }, { "epoch": 1.37, "learning_rate": 2.722764590571201e-05, "loss": 0.0038, "step": 70422 }, { "epoch": 1.37, "learning_rate": 2.7226998945454784e-05, "loss": 0.0028, "step": 70424 }, { "epoch": 1.37, "learning_rate": 2.722635198519755e-05, "loss": 0.0, "step": 70426 }, { "epoch": 1.37, "learning_rate": 2.722570502494032e-05, "loss": 0.0, "step": 70428 }, { "epoch": 1.37, "learning_rate": 2.722505806468309e-05, "loss": 0.0031, "step": 70430 }, { "epoch": 1.37, "learning_rate": 2.7224411104425857e-05, "loss": 0.0, "step": 70432 }, { "epoch": 1.37, "learning_rate": 2.7223764144168623e-05, "loss": 0.0, "step": 70434 }, { "epoch": 1.37, "learning_rate": 2.7223117183911395e-05, "loss": 0.0002, "step": 70436 }, { "epoch": 1.37, "learning_rate": 2.722247022365416e-05, "loss": 0.0, "step": 70438 }, { "epoch": 1.37, "learning_rate": 2.722182326339693e-05, "loss": 0.0, "step": 70440 }, { "epoch": 1.37, "learning_rate": 2.72211763031397e-05, "loss": 0.0, "step": 70442 }, { "epoch": 1.37, "learning_rate": 2.7220529342882468e-05, "loss": 0.001, "step": 70444 }, { "epoch": 1.37, "learning_rate": 2.721988238262524e-05, "loss": 0.0022, "step": 70446 }, { "epoch": 1.37, "learning_rate": 2.7219235422368006e-05, "loss": 0.0017, "step": 70448 }, { "epoch": 1.37, "learning_rate": 2.7218588462110772e-05, "loss": 0.0, "step": 70450 }, { "epoch": 1.37, "learning_rate": 2.7217941501853545e-05, "loss": 0.0038, "step": 70452 }, { "epoch": 1.37, "learning_rate": 2.721729454159631e-05, "loss": 0.0003, "step": 70454 }, { "epoch": 1.37, "learning_rate": 2.7216647581339076e-05, "loss": 0.0009, "step": 70456 }, { "epoch": 1.37, "learning_rate": 2.721600062108185e-05, "loss": 0.0078, "step": 70458 }, { "epoch": 1.37, "learning_rate": 2.7215353660824617e-05, "loss": 0.0151, "step": 70460 }, { "epoch": 1.37, "learning_rate": 2.7214706700567383e-05, "loss": 0.0, "step": 70462 }, { "epoch": 1.37, "learning_rate": 2.7214059740310156e-05, "loss": 0.0, "step": 70464 }, { "epoch": 1.37, "learning_rate": 2.721341278005292e-05, "loss": 0.0, "step": 70466 }, { "epoch": 1.37, "learning_rate": 2.7212765819795694e-05, "loss": 0.0, "step": 70468 }, { "epoch": 1.37, "learning_rate": 2.721211885953846e-05, "loss": 0.0008, "step": 70470 }, { "epoch": 1.37, "learning_rate": 2.7211471899281225e-05, "loss": 0.0001, "step": 70472 }, { "epoch": 1.37, "learning_rate": 2.7210824939023998e-05, "loss": 0.0, "step": 70474 }, { "epoch": 1.37, "learning_rate": 2.7210177978766767e-05, "loss": 0.0, "step": 70476 }, { "epoch": 1.37, "learning_rate": 2.7209531018509532e-05, "loss": 0.0, "step": 70478 }, { "epoch": 1.37, "learning_rate": 2.7208884058252305e-05, "loss": 0.0, "step": 70480 }, { "epoch": 1.37, "learning_rate": 2.720823709799507e-05, "loss": 0.0, "step": 70482 }, { "epoch": 1.37, "learning_rate": 2.7207590137737843e-05, "loss": 0.0002, "step": 70484 }, { "epoch": 1.37, "learning_rate": 2.720694317748061e-05, "loss": 0.0, "step": 70486 }, { "epoch": 1.37, "learning_rate": 2.7206296217223374e-05, "loss": 0.0, "step": 70488 }, { "epoch": 1.37, "learning_rate": 2.7205649256966147e-05, "loss": 0.0, "step": 70490 }, { "epoch": 1.37, "learning_rate": 2.7205002296708916e-05, "loss": 0.0053, "step": 70492 }, { "epoch": 1.37, "learning_rate": 2.7204355336451682e-05, "loss": 0.0431, "step": 70494 }, { "epoch": 1.37, "learning_rate": 2.7203708376194454e-05, "loss": 0.0, "step": 70496 }, { "epoch": 1.37, "learning_rate": 2.720306141593722e-05, "loss": 0.0, "step": 70498 }, { "epoch": 1.37, "learning_rate": 2.7202414455679986e-05, "loss": 0.0001, "step": 70500 }, { "epoch": 1.37, "learning_rate": 2.7201767495422758e-05, "loss": 0.0, "step": 70502 }, { "epoch": 1.37, "learning_rate": 2.7201120535165524e-05, "loss": 0.0, "step": 70504 }, { "epoch": 1.37, "learning_rate": 2.7200473574908296e-05, "loss": 0.0, "step": 70506 }, { "epoch": 1.37, "learning_rate": 2.7199826614651065e-05, "loss": 0.0, "step": 70508 }, { "epoch": 1.37, "learning_rate": 2.719917965439383e-05, "loss": 0.0, "step": 70510 }, { "epoch": 1.37, "learning_rate": 2.7198532694136603e-05, "loss": 0.0, "step": 70512 }, { "epoch": 1.37, "learning_rate": 2.719788573387937e-05, "loss": 0.0976, "step": 70514 }, { "epoch": 1.37, "learning_rate": 2.7197238773622135e-05, "loss": 0.0088, "step": 70516 }, { "epoch": 1.37, "learning_rate": 2.7196591813364907e-05, "loss": 0.0001, "step": 70518 }, { "epoch": 1.37, "learning_rate": 2.7195944853107673e-05, "loss": 0.0001, "step": 70520 }, { "epoch": 1.37, "learning_rate": 2.7195297892850442e-05, "loss": 0.0, "step": 70522 }, { "epoch": 1.37, "learning_rate": 2.7194650932593215e-05, "loss": 0.0, "step": 70524 }, { "epoch": 1.37, "learning_rate": 2.719400397233598e-05, "loss": 0.0, "step": 70526 }, { "epoch": 1.37, "learning_rate": 2.7193357012078753e-05, "loss": 0.0, "step": 70528 }, { "epoch": 1.37, "learning_rate": 2.719271005182152e-05, "loss": 0.0, "step": 70530 }, { "epoch": 1.37, "learning_rate": 2.7192063091564284e-05, "loss": 0.0, "step": 70532 }, { "epoch": 1.37, "learning_rate": 2.7191416131307057e-05, "loss": 0.0, "step": 70534 }, { "epoch": 1.37, "learning_rate": 2.7190769171049822e-05, "loss": 0.0, "step": 70536 }, { "epoch": 1.37, "learning_rate": 2.719012221079259e-05, "loss": 0.0, "step": 70538 }, { "epoch": 1.37, "learning_rate": 2.7189475250535364e-05, "loss": 0.0, "step": 70540 }, { "epoch": 1.37, "learning_rate": 2.718882829027813e-05, "loss": 0.0, "step": 70542 }, { "epoch": 1.37, "learning_rate": 2.7188181330020895e-05, "loss": 0.0005, "step": 70544 }, { "epoch": 1.37, "learning_rate": 2.7187534369763668e-05, "loss": 0.0, "step": 70546 }, { "epoch": 1.37, "learning_rate": 2.7186887409506433e-05, "loss": 0.0, "step": 70548 }, { "epoch": 1.37, "learning_rate": 2.7186240449249206e-05, "loss": 0.0, "step": 70550 }, { "epoch": 1.37, "learning_rate": 2.718559348899197e-05, "loss": 0.0, "step": 70552 }, { "epoch": 1.37, "learning_rate": 2.718494652873474e-05, "loss": 0.0103, "step": 70554 }, { "epoch": 1.37, "learning_rate": 2.718429956847751e-05, "loss": 0.0, "step": 70556 }, { "epoch": 1.37, "learning_rate": 2.718365260822028e-05, "loss": 0.0, "step": 70558 }, { "epoch": 1.37, "learning_rate": 2.7183005647963045e-05, "loss": 0.0, "step": 70560 }, { "epoch": 1.37, "learning_rate": 2.7182358687705817e-05, "loss": 0.0026, "step": 70562 }, { "epoch": 1.37, "learning_rate": 2.7181711727448583e-05, "loss": 0.0001, "step": 70564 }, { "epoch": 1.37, "learning_rate": 2.7181064767191355e-05, "loss": 0.0, "step": 70566 }, { "epoch": 1.37, "learning_rate": 2.718041780693412e-05, "loss": 0.0, "step": 70568 }, { "epoch": 1.37, "learning_rate": 2.717977084667689e-05, "loss": 0.005, "step": 70570 }, { "epoch": 1.37, "learning_rate": 2.717912388641966e-05, "loss": 0.0, "step": 70572 }, { "epoch": 1.37, "learning_rate": 2.7178476926162428e-05, "loss": 0.0001, "step": 70574 }, { "epoch": 1.37, "learning_rate": 2.7177829965905194e-05, "loss": 0.0017, "step": 70576 }, { "epoch": 1.37, "learning_rate": 2.7177183005647966e-05, "loss": 0.0, "step": 70578 }, { "epoch": 1.37, "learning_rate": 2.7176536045390732e-05, "loss": 0.0, "step": 70580 }, { "epoch": 1.37, "learning_rate": 2.7175889085133498e-05, "loss": 0.0003, "step": 70582 }, { "epoch": 1.37, "learning_rate": 2.717524212487627e-05, "loss": 0.0016, "step": 70584 }, { "epoch": 1.37, "learning_rate": 2.717459516461904e-05, "loss": 0.0, "step": 70586 }, { "epoch": 1.37, "learning_rate": 2.717394820436181e-05, "loss": 0.0, "step": 70588 }, { "epoch": 1.37, "learning_rate": 2.7173301244104577e-05, "loss": 0.0, "step": 70590 }, { "epoch": 1.37, "learning_rate": 2.7172654283847343e-05, "loss": 0.0001, "step": 70592 }, { "epoch": 1.37, "learning_rate": 2.7172007323590116e-05, "loss": 0.0, "step": 70594 }, { "epoch": 1.37, "learning_rate": 2.717136036333288e-05, "loss": 0.0, "step": 70596 }, { "epoch": 1.37, "learning_rate": 2.7170713403075647e-05, "loss": 0.0, "step": 70598 }, { "epoch": 1.37, "learning_rate": 2.717006644281842e-05, "loss": 0.0, "step": 70600 }, { "epoch": 1.37, "learning_rate": 2.716941948256119e-05, "loss": 0.0, "step": 70602 }, { "epoch": 1.37, "learning_rate": 2.7168772522303954e-05, "loss": 0.0, "step": 70604 }, { "epoch": 1.37, "learning_rate": 2.7168125562046727e-05, "loss": 0.0012, "step": 70606 }, { "epoch": 1.37, "learning_rate": 2.7167478601789492e-05, "loss": 0.0, "step": 70608 }, { "epoch": 1.37, "learning_rate": 2.7166831641532265e-05, "loss": 0.0, "step": 70610 }, { "epoch": 1.37, "learning_rate": 2.716618468127503e-05, "loss": 0.0, "step": 70612 }, { "epoch": 1.37, "learning_rate": 2.7165537721017796e-05, "loss": 0.0, "step": 70614 }, { "epoch": 1.37, "learning_rate": 2.716489076076057e-05, "loss": 0.0, "step": 70616 }, { "epoch": 1.37, "learning_rate": 2.7164243800503338e-05, "loss": 0.0001, "step": 70618 }, { "epoch": 1.37, "learning_rate": 2.7163596840246104e-05, "loss": 0.0, "step": 70620 }, { "epoch": 1.37, "learning_rate": 2.7162949879988876e-05, "loss": 0.0, "step": 70622 }, { "epoch": 1.37, "learning_rate": 2.7162302919731642e-05, "loss": 0.0004, "step": 70624 }, { "epoch": 1.37, "learning_rate": 2.7161655959474414e-05, "loss": 0.0001, "step": 70626 }, { "epoch": 1.37, "learning_rate": 2.716100899921718e-05, "loss": 0.0, "step": 70628 }, { "epoch": 1.37, "learning_rate": 2.7160362038959946e-05, "loss": 0.0, "step": 70630 }, { "epoch": 1.37, "learning_rate": 2.7159715078702718e-05, "loss": 0.0, "step": 70632 }, { "epoch": 1.37, "learning_rate": 2.7159068118445484e-05, "loss": 0.0, "step": 70634 }, { "epoch": 1.37, "learning_rate": 2.7158421158188253e-05, "loss": 0.0, "step": 70636 }, { "epoch": 1.37, "learning_rate": 2.7157774197931025e-05, "loss": 0.0007, "step": 70638 }, { "epoch": 1.37, "learning_rate": 2.715712723767379e-05, "loss": 0.0, "step": 70640 }, { "epoch": 1.37, "learning_rate": 2.7156480277416557e-05, "loss": 0.0, "step": 70642 }, { "epoch": 1.37, "learning_rate": 2.715583331715933e-05, "loss": 0.002, "step": 70644 }, { "epoch": 1.37, "learning_rate": 2.7155186356902095e-05, "loss": 0.0, "step": 70646 }, { "epoch": 1.37, "learning_rate": 2.7154539396644867e-05, "loss": 0.0, "step": 70648 }, { "epoch": 1.37, "learning_rate": 2.7153892436387633e-05, "loss": 0.0002, "step": 70650 }, { "epoch": 1.37, "learning_rate": 2.7153245476130402e-05, "loss": 0.0, "step": 70652 }, { "epoch": 1.37, "learning_rate": 2.7152598515873175e-05, "loss": 0.0005, "step": 70654 }, { "epoch": 1.37, "learning_rate": 2.715195155561594e-05, "loss": 0.0, "step": 70656 }, { "epoch": 1.37, "learning_rate": 2.7151304595358706e-05, "loss": 0.0003, "step": 70658 }, { "epoch": 1.37, "learning_rate": 2.715065763510148e-05, "loss": 0.0001, "step": 70660 }, { "epoch": 1.37, "learning_rate": 2.7150010674844244e-05, "loss": 0.0, "step": 70662 }, { "epoch": 1.37, "learning_rate": 2.7149363714587013e-05, "loss": 0.0228, "step": 70664 }, { "epoch": 1.37, "learning_rate": 2.7148716754329782e-05, "loss": 0.0, "step": 70666 }, { "epoch": 1.37, "learning_rate": 2.714806979407255e-05, "loss": 0.0009, "step": 70668 }, { "epoch": 1.37, "learning_rate": 2.7147422833815324e-05, "loss": 0.0, "step": 70670 }, { "epoch": 1.37, "learning_rate": 2.714677587355809e-05, "loss": 0.0, "step": 70672 }, { "epoch": 1.37, "learning_rate": 2.7146128913300855e-05, "loss": 0.0001, "step": 70674 }, { "epoch": 1.37, "learning_rate": 2.7145481953043628e-05, "loss": 0.0, "step": 70676 }, { "epoch": 1.37, "learning_rate": 2.7144834992786393e-05, "loss": 0.0, "step": 70678 }, { "epoch": 1.37, "learning_rate": 2.7144188032529163e-05, "loss": 0.0, "step": 70680 }, { "epoch": 1.37, "learning_rate": 2.714354107227193e-05, "loss": 0.0003, "step": 70682 }, { "epoch": 1.37, "learning_rate": 2.71428941120147e-05, "loss": 0.0, "step": 70684 }, { "epoch": 1.37, "learning_rate": 2.7142247151757466e-05, "loss": 0.0, "step": 70686 }, { "epoch": 1.37, "learning_rate": 2.714160019150024e-05, "loss": 0.0092, "step": 70688 }, { "epoch": 1.37, "learning_rate": 2.7140953231243005e-05, "loss": 0.0, "step": 70690 }, { "epoch": 1.37, "learning_rate": 2.7140306270985777e-05, "loss": 0.0, "step": 70692 }, { "epoch": 1.37, "learning_rate": 2.7139659310728543e-05, "loss": 0.0, "step": 70694 }, { "epoch": 1.37, "learning_rate": 2.713901235047131e-05, "loss": 0.0, "step": 70696 }, { "epoch": 1.37, "learning_rate": 2.713836539021408e-05, "loss": 0.0, "step": 70698 }, { "epoch": 1.37, "learning_rate": 2.713771842995685e-05, "loss": 0.0, "step": 70700 }, { "epoch": 1.37, "learning_rate": 2.7137071469699616e-05, "loss": 0.0, "step": 70702 }, { "epoch": 1.37, "learning_rate": 2.7136424509442388e-05, "loss": 0.0083, "step": 70704 }, { "epoch": 1.37, "learning_rate": 2.7135777549185154e-05, "loss": 0.0, "step": 70706 }, { "epoch": 1.37, "learning_rate": 2.7135130588927926e-05, "loss": 0.0, "step": 70708 }, { "epoch": 1.37, "learning_rate": 2.7134483628670692e-05, "loss": 0.0, "step": 70710 }, { "epoch": 1.37, "learning_rate": 2.7133836668413458e-05, "loss": 0.0043, "step": 70712 }, { "epoch": 1.37, "learning_rate": 2.713318970815623e-05, "loss": 0.0, "step": 70714 }, { "epoch": 1.37, "learning_rate": 2.7132542747899e-05, "loss": 0.0, "step": 70716 }, { "epoch": 1.37, "learning_rate": 2.7131895787641765e-05, "loss": 0.0005, "step": 70718 }, { "epoch": 1.37, "learning_rate": 2.7131248827384537e-05, "loss": 0.0, "step": 70720 }, { "epoch": 1.37, "learning_rate": 2.7130601867127303e-05, "loss": 0.0, "step": 70722 }, { "epoch": 1.37, "learning_rate": 2.712995490687007e-05, "loss": 0.0004, "step": 70724 }, { "epoch": 1.37, "learning_rate": 2.712930794661284e-05, "loss": 0.0, "step": 70726 }, { "epoch": 1.37, "learning_rate": 2.7128660986355607e-05, "loss": 0.0, "step": 70728 }, { "epoch": 1.37, "learning_rate": 2.712801402609838e-05, "loss": 0.0, "step": 70730 }, { "epoch": 1.37, "learning_rate": 2.712736706584115e-05, "loss": 0.0, "step": 70732 }, { "epoch": 1.37, "learning_rate": 2.7126720105583914e-05, "loss": 0.0, "step": 70734 }, { "epoch": 1.37, "learning_rate": 2.7126073145326687e-05, "loss": 0.0, "step": 70736 }, { "epoch": 1.37, "learning_rate": 2.7125426185069452e-05, "loss": 0.0, "step": 70738 }, { "epoch": 1.37, "learning_rate": 2.7124779224812218e-05, "loss": 0.0005, "step": 70740 }, { "epoch": 1.37, "learning_rate": 2.712413226455499e-05, "loss": 0.0, "step": 70742 }, { "epoch": 1.37, "learning_rate": 2.7123485304297756e-05, "loss": 0.0095, "step": 70744 }, { "epoch": 1.37, "learning_rate": 2.7122838344040525e-05, "loss": 0.0, "step": 70746 }, { "epoch": 1.37, "learning_rate": 2.7122191383783298e-05, "loss": 0.0, "step": 70748 }, { "epoch": 1.37, "learning_rate": 2.7121544423526064e-05, "loss": 0.0, "step": 70750 }, { "epoch": 1.37, "learning_rate": 2.7120897463268836e-05, "loss": 0.0, "step": 70752 }, { "epoch": 1.37, "learning_rate": 2.7120250503011602e-05, "loss": 0.0001, "step": 70754 }, { "epoch": 1.37, "learning_rate": 2.7119603542754367e-05, "loss": 0.0, "step": 70756 }, { "epoch": 1.37, "learning_rate": 2.711895658249714e-05, "loss": 0.0, "step": 70758 }, { "epoch": 1.37, "learning_rate": 2.7118309622239906e-05, "loss": 0.0, "step": 70760 }, { "epoch": 1.37, "learning_rate": 2.7117662661982675e-05, "loss": 0.0, "step": 70762 }, { "epoch": 1.37, "learning_rate": 2.7117015701725447e-05, "loss": 0.0, "step": 70764 }, { "epoch": 1.37, "learning_rate": 2.7116368741468213e-05, "loss": 0.0, "step": 70766 }, { "epoch": 1.37, "learning_rate": 2.7115721781210985e-05, "loss": 0.0, "step": 70768 }, { "epoch": 1.37, "learning_rate": 2.711507482095375e-05, "loss": 0.0, "step": 70770 }, { "epoch": 1.37, "learning_rate": 2.7114427860696517e-05, "loss": 0.0, "step": 70772 }, { "epoch": 1.37, "learning_rate": 2.711378090043929e-05, "loss": 0.0, "step": 70774 }, { "epoch": 1.37, "learning_rate": 2.7113133940182055e-05, "loss": 0.0, "step": 70776 }, { "epoch": 1.37, "learning_rate": 2.7112486979924824e-05, "loss": 0.0, "step": 70778 }, { "epoch": 1.37, "learning_rate": 2.7111840019667596e-05, "loss": 0.0, "step": 70780 }, { "epoch": 1.37, "learning_rate": 2.7111193059410362e-05, "loss": 0.0, "step": 70782 }, { "epoch": 1.37, "learning_rate": 2.7110546099153128e-05, "loss": 0.0, "step": 70784 }, { "epoch": 1.37, "learning_rate": 2.71098991388959e-05, "loss": 0.0, "step": 70786 }, { "epoch": 1.37, "learning_rate": 2.7109252178638666e-05, "loss": 0.0, "step": 70788 }, { "epoch": 1.37, "learning_rate": 2.710860521838144e-05, "loss": 0.0, "step": 70790 }, { "epoch": 1.37, "learning_rate": 2.7107958258124204e-05, "loss": 0.0, "step": 70792 }, { "epoch": 1.37, "learning_rate": 2.7107311297866973e-05, "loss": 0.0, "step": 70794 }, { "epoch": 1.37, "learning_rate": 2.7106664337609742e-05, "loss": 0.0, "step": 70796 }, { "epoch": 1.37, "learning_rate": 2.710601737735251e-05, "loss": 0.007, "step": 70798 }, { "epoch": 1.37, "learning_rate": 2.7105370417095277e-05, "loss": 0.0, "step": 70800 }, { "epoch": 1.37, "learning_rate": 2.710472345683805e-05, "loss": 0.0, "step": 70802 }, { "epoch": 1.37, "learning_rate": 2.7104076496580815e-05, "loss": 0.0, "step": 70804 }, { "epoch": 1.37, "learning_rate": 2.710342953632358e-05, "loss": 0.0, "step": 70806 }, { "epoch": 1.37, "learning_rate": 2.7102782576066353e-05, "loss": 0.0, "step": 70808 }, { "epoch": 1.37, "learning_rate": 2.7102135615809123e-05, "loss": 0.0, "step": 70810 }, { "epoch": 1.37, "learning_rate": 2.710148865555189e-05, "loss": 0.0, "step": 70812 }, { "epoch": 1.37, "learning_rate": 2.710084169529466e-05, "loss": 0.0, "step": 70814 }, { "epoch": 1.37, "learning_rate": 2.7100194735037426e-05, "loss": 0.0, "step": 70816 }, { "epoch": 1.37, "learning_rate": 2.70995477747802e-05, "loss": 0.0, "step": 70818 }, { "epoch": 1.37, "learning_rate": 2.7098900814522965e-05, "loss": 0.005, "step": 70820 }, { "epoch": 1.37, "learning_rate": 2.709825385426573e-05, "loss": 0.0001, "step": 70822 }, { "epoch": 1.37, "learning_rate": 2.7097606894008503e-05, "loss": 0.0018, "step": 70824 }, { "epoch": 1.37, "learning_rate": 2.7096959933751272e-05, "loss": 0.0, "step": 70826 }, { "epoch": 1.37, "learning_rate": 2.7096312973494037e-05, "loss": 0.0, "step": 70828 }, { "epoch": 1.37, "learning_rate": 2.709566601323681e-05, "loss": 0.0002, "step": 70830 }, { "epoch": 1.37, "learning_rate": 2.7095019052979576e-05, "loss": 0.0, "step": 70832 }, { "epoch": 1.37, "learning_rate": 2.7094372092722348e-05, "loss": 0.0, "step": 70834 }, { "epoch": 1.37, "learning_rate": 2.7093725132465114e-05, "loss": 0.0007, "step": 70836 }, { "epoch": 1.37, "learning_rate": 2.709307817220788e-05, "loss": 0.0, "step": 70838 }, { "epoch": 1.37, "learning_rate": 2.7092431211950652e-05, "loss": 0.0, "step": 70840 }, { "epoch": 1.37, "learning_rate": 2.709178425169342e-05, "loss": 0.0, "step": 70842 }, { "epoch": 1.37, "learning_rate": 2.7091137291436187e-05, "loss": 0.0, "step": 70844 }, { "epoch": 1.38, "learning_rate": 2.709049033117896e-05, "loss": 0.0, "step": 70846 }, { "epoch": 1.38, "learning_rate": 2.7089843370921725e-05, "loss": 0.0024, "step": 70848 }, { "epoch": 1.38, "learning_rate": 2.7089196410664497e-05, "loss": 0.0001, "step": 70850 }, { "epoch": 1.38, "learning_rate": 2.7088549450407263e-05, "loss": 0.0, "step": 70852 }, { "epoch": 1.38, "learning_rate": 2.708790249015003e-05, "loss": 0.0, "step": 70854 }, { "epoch": 1.38, "learning_rate": 2.70872555298928e-05, "loss": 0.0001, "step": 70856 }, { "epoch": 1.38, "learning_rate": 2.708660856963557e-05, "loss": 0.0, "step": 70858 }, { "epoch": 1.38, "learning_rate": 2.7085961609378336e-05, "loss": 0.0002, "step": 70860 }, { "epoch": 1.38, "learning_rate": 2.708531464912111e-05, "loss": 0.0, "step": 70862 }, { "epoch": 1.38, "learning_rate": 2.7084667688863874e-05, "loss": 0.0, "step": 70864 }, { "epoch": 1.38, "learning_rate": 2.708402072860664e-05, "loss": 0.0, "step": 70866 }, { "epoch": 1.38, "learning_rate": 2.7083373768349412e-05, "loss": 0.0, "step": 70868 }, { "epoch": 1.38, "learning_rate": 2.7082726808092178e-05, "loss": 0.0174, "step": 70870 }, { "epoch": 1.38, "learning_rate": 2.708207984783495e-05, "loss": 0.0, "step": 70872 }, { "epoch": 1.38, "learning_rate": 2.7081432887577716e-05, "loss": 0.0034, "step": 70874 }, { "epoch": 1.38, "learning_rate": 2.7080785927320485e-05, "loss": 0.0012, "step": 70876 }, { "epoch": 1.38, "learning_rate": 2.7080138967063258e-05, "loss": 0.0, "step": 70878 }, { "epoch": 1.38, "learning_rate": 2.7079492006806024e-05, "loss": 0.0, "step": 70880 }, { "epoch": 1.38, "learning_rate": 2.707884504654879e-05, "loss": 0.0, "step": 70882 }, { "epoch": 1.38, "learning_rate": 2.7078198086291562e-05, "loss": 0.0, "step": 70884 }, { "epoch": 1.38, "learning_rate": 2.7077551126034327e-05, "loss": 0.0, "step": 70886 }, { "epoch": 1.38, "learning_rate": 2.7076904165777096e-05, "loss": 0.0, "step": 70888 }, { "epoch": 1.38, "learning_rate": 2.7076257205519866e-05, "loss": 0.0, "step": 70890 }, { "epoch": 1.38, "learning_rate": 2.7075610245262635e-05, "loss": 0.0, "step": 70892 }, { "epoch": 1.38, "learning_rate": 2.7074963285005407e-05, "loss": 0.0, "step": 70894 }, { "epoch": 1.38, "learning_rate": 2.7074316324748173e-05, "loss": 0.0, "step": 70896 }, { "epoch": 1.38, "learning_rate": 2.707366936449094e-05, "loss": 0.0, "step": 70898 }, { "epoch": 1.38, "learning_rate": 2.707302240423371e-05, "loss": 0.0, "step": 70900 }, { "epoch": 1.38, "learning_rate": 2.7072375443976477e-05, "loss": 0.0003, "step": 70902 }, { "epoch": 1.38, "learning_rate": 2.7071728483719246e-05, "loss": 0.0, "step": 70904 }, { "epoch": 1.38, "learning_rate": 2.7071081523462015e-05, "loss": 0.0, "step": 70906 }, { "epoch": 1.38, "learning_rate": 2.7070434563204784e-05, "loss": 0.0, "step": 70908 }, { "epoch": 1.38, "learning_rate": 2.7069787602947556e-05, "loss": 0.0, "step": 70910 }, { "epoch": 1.38, "learning_rate": 2.7069140642690322e-05, "loss": 0.0, "step": 70912 }, { "epoch": 1.38, "learning_rate": 2.7068493682433088e-05, "loss": 0.0, "step": 70914 }, { "epoch": 1.38, "learning_rate": 2.706784672217586e-05, "loss": 0.0, "step": 70916 }, { "epoch": 1.38, "learning_rate": 2.7067199761918626e-05, "loss": 0.0, "step": 70918 }, { "epoch": 1.38, "learning_rate": 2.7066552801661395e-05, "loss": 0.0001, "step": 70920 }, { "epoch": 1.38, "learning_rate": 2.7065905841404164e-05, "loss": 0.0, "step": 70922 }, { "epoch": 1.38, "learning_rate": 2.7065258881146933e-05, "loss": 0.0, "step": 70924 }, { "epoch": 1.38, "learning_rate": 2.70646119208897e-05, "loss": 0.0013, "step": 70926 }, { "epoch": 1.38, "learning_rate": 2.706396496063247e-05, "loss": 0.0, "step": 70928 }, { "epoch": 1.38, "learning_rate": 2.7063318000375237e-05, "loss": 0.0, "step": 70930 }, { "epoch": 1.38, "learning_rate": 2.706267104011801e-05, "loss": 0.0, "step": 70932 }, { "epoch": 1.38, "learning_rate": 2.7062024079860775e-05, "loss": 0.0002, "step": 70934 }, { "epoch": 1.38, "learning_rate": 2.706137711960354e-05, "loss": 0.0, "step": 70936 }, { "epoch": 1.38, "learning_rate": 2.7060730159346313e-05, "loss": 0.0, "step": 70938 }, { "epoch": 1.38, "learning_rate": 2.7060083199089083e-05, "loss": 0.0, "step": 70940 }, { "epoch": 1.38, "learning_rate": 2.7059436238831848e-05, "loss": 0.0, "step": 70942 }, { "epoch": 1.38, "learning_rate": 2.705878927857462e-05, "loss": 0.0, "step": 70944 }, { "epoch": 1.38, "learning_rate": 2.7058142318317386e-05, "loss": 0.0, "step": 70946 }, { "epoch": 1.38, "learning_rate": 2.7057495358060152e-05, "loss": 0.0, "step": 70948 }, { "epoch": 1.38, "learning_rate": 2.7056848397802925e-05, "loss": 0.0, "step": 70950 }, { "epoch": 1.38, "learning_rate": 2.705620143754569e-05, "loss": 0.0005, "step": 70952 }, { "epoch": 1.38, "learning_rate": 2.7055554477288463e-05, "loss": 0.0074, "step": 70954 }, { "epoch": 1.38, "learning_rate": 2.7054907517031232e-05, "loss": 0.0112, "step": 70956 }, { "epoch": 1.38, "learning_rate": 2.7054260556773997e-05, "loss": 0.0, "step": 70958 }, { "epoch": 1.38, "learning_rate": 2.705361359651677e-05, "loss": 0.0, "step": 70960 }, { "epoch": 1.38, "learning_rate": 2.7052966636259536e-05, "loss": 0.0001, "step": 70962 }, { "epoch": 1.38, "learning_rate": 2.70523196760023e-05, "loss": 0.0, "step": 70964 }, { "epoch": 1.38, "learning_rate": 2.7051672715745074e-05, "loss": 0.0, "step": 70966 }, { "epoch": 1.38, "learning_rate": 2.705102575548784e-05, "loss": 0.0, "step": 70968 }, { "epoch": 1.38, "learning_rate": 2.705037879523061e-05, "loss": 0.0, "step": 70970 }, { "epoch": 1.38, "learning_rate": 2.704973183497338e-05, "loss": 0.0009, "step": 70972 }, { "epoch": 1.38, "learning_rate": 2.7049084874716147e-05, "loss": 0.0, "step": 70974 }, { "epoch": 1.38, "learning_rate": 2.704843791445892e-05, "loss": 0.0, "step": 70976 }, { "epoch": 1.38, "learning_rate": 2.7047790954201685e-05, "loss": 0.005, "step": 70978 }, { "epoch": 1.38, "learning_rate": 2.704714399394445e-05, "loss": 0.0, "step": 70980 }, { "epoch": 1.38, "learning_rate": 2.7046497033687223e-05, "loss": 0.0001, "step": 70982 }, { "epoch": 1.38, "learning_rate": 2.704585007342999e-05, "loss": 0.0, "step": 70984 }, { "epoch": 1.38, "learning_rate": 2.7045203113172758e-05, "loss": 0.0, "step": 70986 }, { "epoch": 1.38, "learning_rate": 2.704455615291553e-05, "loss": 0.0, "step": 70988 }, { "epoch": 1.38, "learning_rate": 2.7043909192658296e-05, "loss": 0.0, "step": 70990 }, { "epoch": 1.38, "learning_rate": 2.7043585712529677e-05, "loss": 0.1108, "step": 70992 }, { "epoch": 1.38, "learning_rate": 2.704293875227245e-05, "loss": 0.0001, "step": 70994 }, { "epoch": 1.38, "learning_rate": 2.704229179201522e-05, "loss": 0.0, "step": 70996 }, { "epoch": 1.38, "learning_rate": 2.7041644831757984e-05, "loss": 0.0, "step": 70998 }, { "epoch": 1.38, "learning_rate": 2.7040997871500757e-05, "loss": 0.0001, "step": 71000 }, { "epoch": 1.38, "learning_rate": 2.7040350911243523e-05, "loss": 0.0001, "step": 71002 }, { "epoch": 1.38, "learning_rate": 2.7039703950986295e-05, "loss": 0.0, "step": 71004 }, { "epoch": 1.38, "learning_rate": 2.703905699072906e-05, "loss": 0.004, "step": 71006 }, { "epoch": 1.38, "learning_rate": 2.7038410030471827e-05, "loss": 0.0001, "step": 71008 }, { "epoch": 1.38, "learning_rate": 2.70377630702146e-05, "loss": 0.0, "step": 71010 }, { "epoch": 1.38, "learning_rate": 2.7037116109957368e-05, "loss": 0.0, "step": 71012 }, { "epoch": 1.38, "learning_rate": 2.7036469149700134e-05, "loss": 0.0, "step": 71014 }, { "epoch": 1.38, "learning_rate": 2.7035822189442906e-05, "loss": 0.0, "step": 71016 }, { "epoch": 1.38, "learning_rate": 2.7035175229185672e-05, "loss": 0.0, "step": 71018 }, { "epoch": 1.38, "learning_rate": 2.7034528268928438e-05, "loss": 0.0, "step": 71020 }, { "epoch": 1.38, "learning_rate": 2.703388130867121e-05, "loss": 0.0, "step": 71022 }, { "epoch": 1.38, "learning_rate": 2.7033234348413976e-05, "loss": 0.0003, "step": 71024 }, { "epoch": 1.38, "learning_rate": 2.7032587388156748e-05, "loss": 0.0083, "step": 71026 }, { "epoch": 1.38, "learning_rate": 2.7031940427899517e-05, "loss": 0.0, "step": 71028 }, { "epoch": 1.38, "learning_rate": 2.7031293467642283e-05, "loss": 0.0, "step": 71030 }, { "epoch": 1.38, "learning_rate": 2.7030646507385056e-05, "loss": 0.0061, "step": 71032 }, { "epoch": 1.38, "learning_rate": 2.702999954712782e-05, "loss": 0.0, "step": 71034 }, { "epoch": 1.38, "learning_rate": 2.7029352586870587e-05, "loss": 0.0004, "step": 71036 }, { "epoch": 1.38, "learning_rate": 2.702870562661336e-05, "loss": 0.0, "step": 71038 }, { "epoch": 1.38, "learning_rate": 2.7028058666356125e-05, "loss": 0.0001, "step": 71040 }, { "epoch": 1.38, "learning_rate": 2.7027411706098894e-05, "loss": 0.0, "step": 71042 }, { "epoch": 1.38, "learning_rate": 2.7026764745841667e-05, "loss": 0.0, "step": 71044 }, { "epoch": 1.38, "learning_rate": 2.7026117785584432e-05, "loss": 0.0, "step": 71046 }, { "epoch": 1.38, "learning_rate": 2.7025470825327205e-05, "loss": 0.0, "step": 71048 }, { "epoch": 1.38, "learning_rate": 2.702482386506997e-05, "loss": 0.0, "step": 71050 }, { "epoch": 1.38, "learning_rate": 2.7024176904812736e-05, "loss": 0.0251, "step": 71052 }, { "epoch": 1.38, "learning_rate": 2.702352994455551e-05, "loss": 0.0, "step": 71054 }, { "epoch": 1.38, "learning_rate": 2.7022882984298274e-05, "loss": 0.0123, "step": 71056 }, { "epoch": 1.38, "learning_rate": 2.7022236024041043e-05, "loss": 0.0008, "step": 71058 }, { "epoch": 1.38, "learning_rate": 2.7021589063783816e-05, "loss": 0.0, "step": 71060 }, { "epoch": 1.38, "learning_rate": 2.702094210352658e-05, "loss": 0.0112, "step": 71062 }, { "epoch": 1.38, "learning_rate": 2.7020295143269354e-05, "loss": 0.0, "step": 71064 }, { "epoch": 1.38, "learning_rate": 2.701964818301212e-05, "loss": 0.0002, "step": 71066 }, { "epoch": 1.38, "learning_rate": 2.7019001222754885e-05, "loss": 0.0005, "step": 71068 }, { "epoch": 1.38, "learning_rate": 2.7018354262497658e-05, "loss": 0.0, "step": 71070 }, { "epoch": 1.38, "learning_rate": 2.7017707302240424e-05, "loss": 0.0, "step": 71072 }, { "epoch": 1.38, "learning_rate": 2.7017060341983193e-05, "loss": 0.0001, "step": 71074 }, { "epoch": 1.38, "learning_rate": 2.7016413381725962e-05, "loss": 0.0039, "step": 71076 }, { "epoch": 1.38, "learning_rate": 2.701576642146873e-05, "loss": 0.0, "step": 71078 }, { "epoch": 1.38, "learning_rate": 2.7015119461211497e-05, "loss": 0.0, "step": 71080 }, { "epoch": 1.38, "learning_rate": 2.701447250095427e-05, "loss": 0.0034, "step": 71082 }, { "epoch": 1.38, "learning_rate": 2.7013825540697035e-05, "loss": 0.0001, "step": 71084 }, { "epoch": 1.38, "learning_rate": 2.7013178580439807e-05, "loss": 0.0, "step": 71086 }, { "epoch": 1.38, "learning_rate": 2.7012531620182573e-05, "loss": 0.0016, "step": 71088 }, { "epoch": 1.38, "learning_rate": 2.7011884659925342e-05, "loss": 0.0, "step": 71090 }, { "epoch": 1.38, "learning_rate": 2.701123769966811e-05, "loss": 0.0, "step": 71092 }, { "epoch": 1.38, "learning_rate": 2.701059073941088e-05, "loss": 0.0, "step": 71094 }, { "epoch": 1.38, "learning_rate": 2.7009943779153646e-05, "loss": 0.0, "step": 71096 }, { "epoch": 1.38, "learning_rate": 2.700929681889642e-05, "loss": 0.0001, "step": 71098 }, { "epoch": 1.38, "learning_rate": 2.7008649858639184e-05, "loss": 0.0, "step": 71100 }, { "epoch": 1.38, "learning_rate": 2.700800289838195e-05, "loss": 0.0, "step": 71102 }, { "epoch": 1.38, "learning_rate": 2.7007355938124722e-05, "loss": 0.0, "step": 71104 }, { "epoch": 1.38, "learning_rate": 2.700670897786749e-05, "loss": 0.0, "step": 71106 }, { "epoch": 1.38, "learning_rate": 2.700606201761026e-05, "loss": 0.0001, "step": 71108 }, { "epoch": 1.38, "learning_rate": 2.700541505735303e-05, "loss": 0.0, "step": 71110 }, { "epoch": 1.38, "learning_rate": 2.7004768097095795e-05, "loss": 0.0, "step": 71112 }, { "epoch": 1.38, "learning_rate": 2.7004121136838568e-05, "loss": 0.0, "step": 71114 }, { "epoch": 1.38, "learning_rate": 2.7003474176581333e-05, "loss": 0.0, "step": 71116 }, { "epoch": 1.38, "learning_rate": 2.70028272163241e-05, "loss": 0.0, "step": 71118 }, { "epoch": 1.38, "learning_rate": 2.700218025606687e-05, "loss": 0.0009, "step": 71120 }, { "epoch": 1.38, "learning_rate": 2.700153329580964e-05, "loss": 0.0, "step": 71122 }, { "epoch": 1.38, "learning_rate": 2.7000886335552406e-05, "loss": 0.0, "step": 71124 }, { "epoch": 1.38, "learning_rate": 2.700023937529518e-05, "loss": 0.0001, "step": 71126 }, { "epoch": 1.38, "learning_rate": 2.6999592415037944e-05, "loss": 0.0008, "step": 71128 }, { "epoch": 1.38, "learning_rate": 2.6998945454780717e-05, "loss": 0.0, "step": 71130 }, { "epoch": 1.38, "learning_rate": 2.6998298494523483e-05, "loss": 0.0, "step": 71132 }, { "epoch": 1.38, "learning_rate": 2.699765153426625e-05, "loss": 0.0001, "step": 71134 }, { "epoch": 1.38, "learning_rate": 2.699700457400902e-05, "loss": 0.0, "step": 71136 }, { "epoch": 1.38, "learning_rate": 2.6996357613751786e-05, "loss": 0.0, "step": 71138 }, { "epoch": 1.38, "learning_rate": 2.6995710653494556e-05, "loss": 0.0, "step": 71140 }, { "epoch": 1.38, "learning_rate": 2.6995063693237328e-05, "loss": 0.0045, "step": 71142 }, { "epoch": 1.38, "learning_rate": 2.6994416732980094e-05, "loss": 0.0002, "step": 71144 }, { "epoch": 1.38, "learning_rate": 2.6993769772722866e-05, "loss": 0.0, "step": 71146 }, { "epoch": 1.38, "learning_rate": 2.6993122812465632e-05, "loss": 0.0, "step": 71148 }, { "epoch": 1.38, "learning_rate": 2.6992475852208398e-05, "loss": 0.0061, "step": 71150 }, { "epoch": 1.38, "learning_rate": 2.699182889195117e-05, "loss": 0.0, "step": 71152 }, { "epoch": 1.38, "learning_rate": 2.6991181931693936e-05, "loss": 0.0004, "step": 71154 }, { "epoch": 1.38, "learning_rate": 2.6990534971436705e-05, "loss": 0.0, "step": 71156 }, { "epoch": 1.38, "learning_rate": 2.6989888011179477e-05, "loss": 0.0001, "step": 71158 }, { "epoch": 1.38, "learning_rate": 2.6989241050922243e-05, "loss": 0.0, "step": 71160 }, { "epoch": 1.38, "learning_rate": 2.698859409066501e-05, "loss": 0.0, "step": 71162 }, { "epoch": 1.38, "learning_rate": 2.698794713040778e-05, "loss": 0.0, "step": 71164 }, { "epoch": 1.38, "learning_rate": 2.6987300170150547e-05, "loss": 0.0, "step": 71166 }, { "epoch": 1.38, "learning_rate": 2.698665320989332e-05, "loss": 0.0, "step": 71168 }, { "epoch": 1.38, "learning_rate": 2.6986006249636085e-05, "loss": 0.0, "step": 71170 }, { "epoch": 1.38, "learning_rate": 2.6985359289378854e-05, "loss": 0.0, "step": 71172 }, { "epoch": 1.38, "learning_rate": 2.6984712329121627e-05, "loss": 0.0, "step": 71174 }, { "epoch": 1.38, "learning_rate": 2.6984065368864392e-05, "loss": 0.0005, "step": 71176 }, { "epoch": 1.38, "learning_rate": 2.6983418408607158e-05, "loss": 0.0, "step": 71178 }, { "epoch": 1.38, "learning_rate": 2.698277144834993e-05, "loss": 0.0001, "step": 71180 }, { "epoch": 1.38, "learning_rate": 2.6982124488092696e-05, "loss": 0.0, "step": 71182 }, { "epoch": 1.38, "learning_rate": 2.6981477527835465e-05, "loss": 0.0, "step": 71184 }, { "epoch": 1.38, "learning_rate": 2.6980830567578234e-05, "loss": 0.0016, "step": 71186 }, { "epoch": 1.38, "learning_rate": 2.6980183607321003e-05, "loss": 0.0, "step": 71188 }, { "epoch": 1.38, "learning_rate": 2.6979536647063776e-05, "loss": 0.0, "step": 71190 }, { "epoch": 1.38, "learning_rate": 2.697888968680654e-05, "loss": 0.0, "step": 71192 }, { "epoch": 1.38, "learning_rate": 2.6978242726549307e-05, "loss": 0.0, "step": 71194 }, { "epoch": 1.38, "learning_rate": 2.697759576629208e-05, "loss": 0.0, "step": 71196 }, { "epoch": 1.38, "learning_rate": 2.6976948806034845e-05, "loss": 0.0057, "step": 71198 }, { "epoch": 1.38, "learning_rate": 2.6976301845777615e-05, "loss": 0.0008, "step": 71200 }, { "epoch": 1.38, "learning_rate": 2.6975654885520384e-05, "loss": 0.0, "step": 71202 }, { "epoch": 1.38, "learning_rate": 2.6975007925263153e-05, "loss": 0.0, "step": 71204 }, { "epoch": 1.38, "learning_rate": 2.6974360965005925e-05, "loss": 0.0, "step": 71206 }, { "epoch": 1.38, "learning_rate": 2.697371400474869e-05, "loss": 0.0, "step": 71208 }, { "epoch": 1.38, "learning_rate": 2.6973067044491457e-05, "loss": 0.0, "step": 71210 }, { "epoch": 1.38, "learning_rate": 2.697242008423423e-05, "loss": 0.0, "step": 71212 }, { "epoch": 1.38, "learning_rate": 2.6971773123976995e-05, "loss": 0.0, "step": 71214 }, { "epoch": 1.38, "learning_rate": 2.697112616371976e-05, "loss": 0.0, "step": 71216 }, { "epoch": 1.38, "learning_rate": 2.6970479203462533e-05, "loss": 0.0, "step": 71218 }, { "epoch": 1.38, "learning_rate": 2.6969832243205302e-05, "loss": 0.0, "step": 71220 }, { "epoch": 1.38, "learning_rate": 2.6969185282948068e-05, "loss": 0.0001, "step": 71222 }, { "epoch": 1.38, "learning_rate": 2.696853832269084e-05, "loss": 0.0, "step": 71224 }, { "epoch": 1.38, "learning_rate": 2.6967891362433606e-05, "loss": 0.0, "step": 71226 }, { "epoch": 1.38, "learning_rate": 2.696724440217638e-05, "loss": 0.0, "step": 71228 }, { "epoch": 1.38, "learning_rate": 2.6966597441919144e-05, "loss": 0.0, "step": 71230 }, { "epoch": 1.38, "learning_rate": 2.696595048166191e-05, "loss": 0.0001, "step": 71232 }, { "epoch": 1.38, "learning_rate": 2.6965303521404682e-05, "loss": 0.0017, "step": 71234 }, { "epoch": 1.38, "learning_rate": 2.696465656114745e-05, "loss": 0.0, "step": 71236 }, { "epoch": 1.38, "learning_rate": 2.6964009600890217e-05, "loss": 0.0, "step": 71238 }, { "epoch": 1.38, "learning_rate": 2.696336264063299e-05, "loss": 0.0, "step": 71240 }, { "epoch": 1.38, "learning_rate": 2.6962715680375755e-05, "loss": 0.0, "step": 71242 }, { "epoch": 1.38, "learning_rate": 2.696206872011852e-05, "loss": 0.0001, "step": 71244 }, { "epoch": 1.38, "learning_rate": 2.6961421759861293e-05, "loss": 0.0, "step": 71246 }, { "epoch": 1.38, "learning_rate": 2.696077479960406e-05, "loss": 0.0, "step": 71248 }, { "epoch": 1.38, "learning_rate": 2.696012783934683e-05, "loss": 0.0001, "step": 71250 }, { "epoch": 1.38, "learning_rate": 2.69594808790896e-05, "loss": 0.0, "step": 71252 }, { "epoch": 1.38, "learning_rate": 2.6958833918832366e-05, "loss": 0.0, "step": 71254 }, { "epoch": 1.38, "learning_rate": 2.695818695857514e-05, "loss": 0.0, "step": 71256 }, { "epoch": 1.38, "learning_rate": 2.6957539998317904e-05, "loss": 0.0, "step": 71258 }, { "epoch": 1.38, "learning_rate": 2.695689303806067e-05, "loss": 0.0001, "step": 71260 }, { "epoch": 1.38, "learning_rate": 2.6956246077803443e-05, "loss": 0.0, "step": 71262 }, { "epoch": 1.38, "learning_rate": 2.695559911754621e-05, "loss": 0.0, "step": 71264 }, { "epoch": 1.38, "learning_rate": 2.6954952157288977e-05, "loss": 0.0, "step": 71266 }, { "epoch": 1.38, "learning_rate": 2.695430519703175e-05, "loss": 0.0062, "step": 71268 }, { "epoch": 1.38, "learning_rate": 2.6953658236774516e-05, "loss": 0.0001, "step": 71270 }, { "epoch": 1.38, "learning_rate": 2.6953011276517288e-05, "loss": 0.0005, "step": 71272 }, { "epoch": 1.38, "learning_rate": 2.6952364316260054e-05, "loss": 0.0, "step": 71274 }, { "epoch": 1.38, "learning_rate": 2.695171735600282e-05, "loss": 0.0, "step": 71276 }, { "epoch": 1.38, "learning_rate": 2.6951070395745592e-05, "loss": 0.0002, "step": 71278 }, { "epoch": 1.38, "learning_rate": 2.6950423435488358e-05, "loss": 0.0, "step": 71280 }, { "epoch": 1.38, "learning_rate": 2.6949776475231127e-05, "loss": 0.0003, "step": 71282 }, { "epoch": 1.38, "learning_rate": 2.69491295149739e-05, "loss": 0.0012, "step": 71284 }, { "epoch": 1.38, "learning_rate": 2.6948482554716665e-05, "loss": 0.0001, "step": 71286 }, { "epoch": 1.38, "learning_rate": 2.6947835594459437e-05, "loss": 0.0, "step": 71288 }, { "epoch": 1.38, "learning_rate": 2.6947188634202203e-05, "loss": 0.0039, "step": 71290 }, { "epoch": 1.38, "learning_rate": 2.694654167394497e-05, "loss": 0.0066, "step": 71292 }, { "epoch": 1.38, "learning_rate": 2.694589471368774e-05, "loss": 0.0, "step": 71294 }, { "epoch": 1.38, "learning_rate": 2.6945247753430507e-05, "loss": 0.0, "step": 71296 }, { "epoch": 1.38, "learning_rate": 2.6944600793173276e-05, "loss": 0.0, "step": 71298 }, { "epoch": 1.38, "learning_rate": 2.694395383291605e-05, "loss": 0.0, "step": 71300 }, { "epoch": 1.38, "learning_rate": 2.6943306872658814e-05, "loss": 0.0, "step": 71302 }, { "epoch": 1.38, "learning_rate": 2.694265991240158e-05, "loss": 0.0, "step": 71304 }, { "epoch": 1.38, "learning_rate": 2.6942012952144352e-05, "loss": 0.0022, "step": 71306 }, { "epoch": 1.38, "learning_rate": 2.6941365991887118e-05, "loss": 0.0004, "step": 71308 }, { "epoch": 1.38, "learning_rate": 2.694071903162989e-05, "loss": 0.0, "step": 71310 }, { "epoch": 1.38, "learning_rate": 2.6940072071372656e-05, "loss": 0.0, "step": 71312 }, { "epoch": 1.38, "learning_rate": 2.6939425111115425e-05, "loss": 0.0, "step": 71314 }, { "epoch": 1.38, "learning_rate": 2.6938778150858194e-05, "loss": 0.0, "step": 71316 }, { "epoch": 1.38, "learning_rate": 2.6938131190600963e-05, "loss": 0.0, "step": 71318 }, { "epoch": 1.38, "learning_rate": 2.693748423034373e-05, "loss": 0.0, "step": 71320 }, { "epoch": 1.38, "learning_rate": 2.69368372700865e-05, "loss": 0.0007, "step": 71322 }, { "epoch": 1.38, "learning_rate": 2.6936190309829267e-05, "loss": 0.0002, "step": 71324 }, { "epoch": 1.38, "learning_rate": 2.6935543349572033e-05, "loss": 0.0, "step": 71326 }, { "epoch": 1.38, "learning_rate": 2.6934896389314805e-05, "loss": 0.0, "step": 71328 }, { "epoch": 1.38, "learning_rate": 2.6934249429057575e-05, "loss": 0.0003, "step": 71330 }, { "epoch": 1.38, "learning_rate": 2.6933602468800344e-05, "loss": 0.0, "step": 71332 }, { "epoch": 1.38, "learning_rate": 2.6932955508543113e-05, "loss": 0.0005, "step": 71334 }, { "epoch": 1.38, "learning_rate": 2.693230854828588e-05, "loss": 0.0001, "step": 71336 }, { "epoch": 1.38, "learning_rate": 2.693166158802865e-05, "loss": 0.0, "step": 71338 }, { "epoch": 1.38, "learning_rate": 2.6931014627771417e-05, "loss": 0.0001, "step": 71340 }, { "epoch": 1.38, "learning_rate": 2.6930367667514182e-05, "loss": 0.0001, "step": 71342 }, { "epoch": 1.38, "learning_rate": 2.6929720707256955e-05, "loss": 0.0, "step": 71344 }, { "epoch": 1.38, "learning_rate": 2.6929073746999724e-05, "loss": 0.0006, "step": 71346 }, { "epoch": 1.38, "learning_rate": 2.6928426786742493e-05, "loss": 0.001, "step": 71348 }, { "epoch": 1.38, "learning_rate": 2.6927779826485262e-05, "loss": 0.0, "step": 71350 }, { "epoch": 1.38, "learning_rate": 2.6927132866228028e-05, "loss": 0.0004, "step": 71352 }, { "epoch": 1.38, "learning_rate": 2.69264859059708e-05, "loss": 0.0005, "step": 71354 }, { "epoch": 1.38, "learning_rate": 2.6925838945713566e-05, "loss": 0.008, "step": 71356 }, { "epoch": 1.38, "learning_rate": 2.692519198545633e-05, "loss": 0.0, "step": 71358 }, { "epoch": 1.39, "learning_rate": 2.6924545025199104e-05, "loss": 0.0, "step": 71360 }, { "epoch": 1.39, "learning_rate": 2.6923898064941873e-05, "loss": 0.0, "step": 71362 }, { "epoch": 1.39, "learning_rate": 2.692325110468464e-05, "loss": 0.0, "step": 71364 }, { "epoch": 1.39, "learning_rate": 2.692260414442741e-05, "loss": 0.0032, "step": 71366 }, { "epoch": 1.39, "learning_rate": 2.6921957184170177e-05, "loss": 0.0001, "step": 71368 }, { "epoch": 1.39, "learning_rate": 2.692131022391295e-05, "loss": 0.0, "step": 71370 }, { "epoch": 1.39, "learning_rate": 2.6920663263655715e-05, "loss": 0.0, "step": 71372 }, { "epoch": 1.39, "learning_rate": 2.692001630339848e-05, "loss": 0.0, "step": 71374 }, { "epoch": 1.39, "learning_rate": 2.6919369343141253e-05, "loss": 0.0, "step": 71376 }, { "epoch": 1.39, "learning_rate": 2.691872238288402e-05, "loss": 0.0, "step": 71378 }, { "epoch": 1.39, "learning_rate": 2.6918075422626788e-05, "loss": 0.0033, "step": 71380 }, { "epoch": 1.39, "learning_rate": 2.691742846236956e-05, "loss": 0.0, "step": 71382 }, { "epoch": 1.39, "learning_rate": 2.6916781502112326e-05, "loss": 0.0007, "step": 71384 }, { "epoch": 1.39, "learning_rate": 2.6916134541855092e-05, "loss": 0.0, "step": 71386 }, { "epoch": 1.39, "learning_rate": 2.6915487581597864e-05, "loss": 0.0057, "step": 71388 }, { "epoch": 1.39, "learning_rate": 2.691484062134063e-05, "loss": 0.0, "step": 71390 }, { "epoch": 1.39, "learning_rate": 2.6914193661083403e-05, "loss": 0.0, "step": 71392 }, { "epoch": 1.39, "learning_rate": 2.691354670082617e-05, "loss": 0.0244, "step": 71394 }, { "epoch": 1.39, "learning_rate": 2.6912899740568937e-05, "loss": 0.0013, "step": 71396 }, { "epoch": 1.39, "learning_rate": 2.691225278031171e-05, "loss": 0.0001, "step": 71398 }, { "epoch": 1.39, "learning_rate": 2.6911605820054476e-05, "loss": 0.0, "step": 71400 }, { "epoch": 1.39, "learning_rate": 2.691095885979724e-05, "loss": 0.0003, "step": 71402 }, { "epoch": 1.39, "learning_rate": 2.6910311899540014e-05, "loss": 0.0027, "step": 71404 }, { "epoch": 1.39, "learning_rate": 2.690966493928278e-05, "loss": 0.0, "step": 71406 }, { "epoch": 1.39, "learning_rate": 2.690901797902555e-05, "loss": 0.0, "step": 71408 }, { "epoch": 1.39, "learning_rate": 2.6908371018768318e-05, "loss": 0.0009, "step": 71410 }, { "epoch": 1.39, "learning_rate": 2.6907724058511087e-05, "loss": 0.0, "step": 71412 }, { "epoch": 1.39, "learning_rate": 2.690707709825386e-05, "loss": 0.0, "step": 71414 }, { "epoch": 1.39, "learning_rate": 2.6906430137996625e-05, "loss": 0.0005, "step": 71416 }, { "epoch": 1.39, "learning_rate": 2.690578317773939e-05, "loss": 0.0, "step": 71418 }, { "epoch": 1.39, "learning_rate": 2.6905136217482163e-05, "loss": 0.0, "step": 71420 }, { "epoch": 1.39, "learning_rate": 2.690448925722493e-05, "loss": 0.0, "step": 71422 }, { "epoch": 1.39, "learning_rate": 2.6903842296967698e-05, "loss": 0.0, "step": 71424 }, { "epoch": 1.39, "learning_rate": 2.6903195336710467e-05, "loss": 0.0001, "step": 71426 }, { "epoch": 1.39, "learning_rate": 2.6902548376453236e-05, "loss": 0.0, "step": 71428 }, { "epoch": 1.39, "learning_rate": 2.690190141619601e-05, "loss": 0.0, "step": 71430 }, { "epoch": 1.39, "learning_rate": 2.6901254455938774e-05, "loss": 0.0, "step": 71432 }, { "epoch": 1.39, "learning_rate": 2.690060749568154e-05, "loss": 0.0001, "step": 71434 }, { "epoch": 1.39, "learning_rate": 2.6899960535424312e-05, "loss": 0.0, "step": 71436 }, { "epoch": 1.39, "learning_rate": 2.6899313575167078e-05, "loss": 0.0, "step": 71438 }, { "epoch": 1.39, "learning_rate": 2.6898666614909847e-05, "loss": 0.003, "step": 71440 }, { "epoch": 1.39, "learning_rate": 2.6898019654652616e-05, "loss": 0.0, "step": 71442 }, { "epoch": 1.39, "learning_rate": 2.6897372694395385e-05, "loss": 0.0002, "step": 71444 }, { "epoch": 1.39, "learning_rate": 2.689672573413815e-05, "loss": 0.0001, "step": 71446 }, { "epoch": 1.39, "learning_rate": 2.6896078773880923e-05, "loss": 0.0, "step": 71448 }, { "epoch": 1.39, "learning_rate": 2.689543181362369e-05, "loss": 0.0, "step": 71450 }, { "epoch": 1.39, "learning_rate": 2.689478485336646e-05, "loss": 0.0, "step": 71452 }, { "epoch": 1.39, "learning_rate": 2.6894137893109227e-05, "loss": 0.0, "step": 71454 }, { "epoch": 1.39, "learning_rate": 2.6893490932851993e-05, "loss": 0.0001, "step": 71456 }, { "epoch": 1.39, "learning_rate": 2.6892843972594765e-05, "loss": 0.0, "step": 71458 }, { "epoch": 1.39, "learning_rate": 2.6892197012337535e-05, "loss": 0.0, "step": 71460 }, { "epoch": 1.39, "learning_rate": 2.68915500520803e-05, "loss": 0.0, "step": 71462 }, { "epoch": 1.39, "learning_rate": 2.6890903091823073e-05, "loss": 0.0, "step": 71464 }, { "epoch": 1.39, "learning_rate": 2.689025613156584e-05, "loss": 0.0, "step": 71466 }, { "epoch": 1.39, "learning_rate": 2.6889609171308604e-05, "loss": 0.0, "step": 71468 }, { "epoch": 1.39, "learning_rate": 2.6888962211051377e-05, "loss": 0.0004, "step": 71470 }, { "epoch": 1.39, "learning_rate": 2.6888315250794142e-05, "loss": 0.0, "step": 71472 }, { "epoch": 1.39, "learning_rate": 2.6887668290536915e-05, "loss": 0.0, "step": 71474 }, { "epoch": 1.39, "learning_rate": 2.6887021330279684e-05, "loss": 0.0, "step": 71476 }, { "epoch": 1.39, "learning_rate": 2.688637437002245e-05, "loss": 0.0023, "step": 71478 }, { "epoch": 1.39, "learning_rate": 2.6885727409765222e-05, "loss": 0.0, "step": 71480 }, { "epoch": 1.39, "learning_rate": 2.6885080449507988e-05, "loss": 0.0, "step": 71482 }, { "epoch": 1.39, "learning_rate": 2.6884433489250753e-05, "loss": 0.0, "step": 71484 }, { "epoch": 1.39, "learning_rate": 2.6883786528993526e-05, "loss": 0.0115, "step": 71486 }, { "epoch": 1.39, "learning_rate": 2.688313956873629e-05, "loss": 0.0, "step": 71488 }, { "epoch": 1.39, "learning_rate": 2.6882492608479064e-05, "loss": 0.0, "step": 71490 }, { "epoch": 1.39, "learning_rate": 2.6881845648221833e-05, "loss": 0.0, "step": 71492 }, { "epoch": 1.39, "learning_rate": 2.68811986879646e-05, "loss": 0.0, "step": 71494 }, { "epoch": 1.39, "learning_rate": 2.688055172770737e-05, "loss": 0.0033, "step": 71496 }, { "epoch": 1.39, "learning_rate": 2.6879904767450137e-05, "loss": 0.0, "step": 71498 }, { "epoch": 1.39, "learning_rate": 2.6879257807192903e-05, "loss": 0.0001, "step": 71500 }, { "epoch": 1.39, "learning_rate": 2.6878610846935675e-05, "loss": 0.0, "step": 71502 }, { "epoch": 1.39, "learning_rate": 2.687796388667844e-05, "loss": 0.0074, "step": 71504 }, { "epoch": 1.39, "learning_rate": 2.687731692642121e-05, "loss": 0.0053, "step": 71506 }, { "epoch": 1.39, "learning_rate": 2.6876669966163982e-05, "loss": 0.003, "step": 71508 }, { "epoch": 1.39, "learning_rate": 2.6876023005906748e-05, "loss": 0.0, "step": 71510 }, { "epoch": 1.39, "learning_rate": 2.687537604564952e-05, "loss": 0.0, "step": 71512 }, { "epoch": 1.39, "learning_rate": 2.6874729085392286e-05, "loss": 0.0, "step": 71514 }, { "epoch": 1.39, "learning_rate": 2.6874082125135052e-05, "loss": 0.0, "step": 71516 }, { "epoch": 1.39, "learning_rate": 2.6873435164877824e-05, "loss": 0.0, "step": 71518 }, { "epoch": 1.39, "learning_rate": 2.687278820462059e-05, "loss": 0.0, "step": 71520 }, { "epoch": 1.39, "learning_rate": 2.687214124436336e-05, "loss": 0.0, "step": 71522 }, { "epoch": 1.39, "learning_rate": 2.687149428410613e-05, "loss": 0.0, "step": 71524 }, { "epoch": 1.39, "learning_rate": 2.6870847323848897e-05, "loss": 0.0, "step": 71526 }, { "epoch": 1.39, "learning_rate": 2.6870200363591663e-05, "loss": 0.0004, "step": 71528 }, { "epoch": 1.39, "learning_rate": 2.6869553403334436e-05, "loss": 0.0009, "step": 71530 }, { "epoch": 1.39, "learning_rate": 2.68689064430772e-05, "loss": 0.0, "step": 71532 }, { "epoch": 1.39, "learning_rate": 2.6868259482819974e-05, "loss": 0.0138, "step": 71534 }, { "epoch": 1.39, "learning_rate": 2.686761252256274e-05, "loss": 0.0, "step": 71536 }, { "epoch": 1.39, "learning_rate": 2.686696556230551e-05, "loss": 0.0, "step": 71538 }, { "epoch": 1.39, "learning_rate": 2.686631860204828e-05, "loss": 0.0, "step": 71540 }, { "epoch": 1.39, "learning_rate": 2.6865671641791047e-05, "loss": 0.0373, "step": 71542 }, { "epoch": 1.39, "learning_rate": 2.6865024681533812e-05, "loss": 0.0, "step": 71544 }, { "epoch": 1.39, "learning_rate": 2.6864377721276585e-05, "loss": 0.0, "step": 71546 }, { "epoch": 1.39, "learning_rate": 2.686373076101935e-05, "loss": 0.0, "step": 71548 }, { "epoch": 1.39, "learning_rate": 2.6863083800762116e-05, "loss": 0.0, "step": 71550 }, { "epoch": 1.39, "learning_rate": 2.686243684050489e-05, "loss": 0.0, "step": 71552 }, { "epoch": 1.39, "learning_rate": 2.6861789880247658e-05, "loss": 0.0, "step": 71554 }, { "epoch": 1.39, "learning_rate": 2.6861142919990427e-05, "loss": 0.0, "step": 71556 }, { "epoch": 1.39, "learning_rate": 2.6860495959733196e-05, "loss": 0.0, "step": 71558 }, { "epoch": 1.39, "learning_rate": 2.685984899947596e-05, "loss": 0.0, "step": 71560 }, { "epoch": 1.39, "learning_rate": 2.6859202039218734e-05, "loss": 0.0, "step": 71562 }, { "epoch": 1.39, "learning_rate": 2.68585550789615e-05, "loss": 0.0, "step": 71564 }, { "epoch": 1.39, "learning_rate": 2.6857908118704266e-05, "loss": 0.0, "step": 71566 }, { "epoch": 1.39, "learning_rate": 2.6857261158447038e-05, "loss": 0.0, "step": 71568 }, { "epoch": 1.39, "learning_rate": 2.6856614198189807e-05, "loss": 0.0, "step": 71570 }, { "epoch": 1.39, "learning_rate": 2.6855967237932576e-05, "loss": 0.0003, "step": 71572 }, { "epoch": 1.39, "learning_rate": 2.6855320277675345e-05, "loss": 0.0244, "step": 71574 }, { "epoch": 1.39, "learning_rate": 2.685467331741811e-05, "loss": 0.0, "step": 71576 }, { "epoch": 1.39, "learning_rate": 2.6854026357160883e-05, "loss": 0.0, "step": 71578 }, { "epoch": 1.39, "learning_rate": 2.685337939690365e-05, "loss": 0.0, "step": 71580 }, { "epoch": 1.39, "learning_rate": 2.6852732436646415e-05, "loss": 0.0, "step": 71582 }, { "epoch": 1.39, "learning_rate": 2.6852085476389187e-05, "loss": 0.0001, "step": 71584 }, { "epoch": 1.39, "learning_rate": 2.6851438516131956e-05, "loss": 0.0, "step": 71586 }, { "epoch": 1.39, "learning_rate": 2.6850791555874722e-05, "loss": 0.0, "step": 71588 }, { "epoch": 1.39, "learning_rate": 2.6850144595617495e-05, "loss": 0.0, "step": 71590 }, { "epoch": 1.39, "learning_rate": 2.684949763536026e-05, "loss": 0.0, "step": 71592 }, { "epoch": 1.39, "learning_rate": 2.6848850675103033e-05, "loss": 0.0, "step": 71594 }, { "epoch": 1.39, "learning_rate": 2.68482037148458e-05, "loss": 0.0, "step": 71596 }, { "epoch": 1.39, "learning_rate": 2.6847556754588564e-05, "loss": 0.001, "step": 71598 }, { "epoch": 1.39, "learning_rate": 2.6846909794331337e-05, "loss": 0.0, "step": 71600 }, { "epoch": 1.39, "learning_rate": 2.6846262834074106e-05, "loss": 0.0, "step": 71602 }, { "epoch": 1.39, "learning_rate": 2.684561587381687e-05, "loss": 0.0, "step": 71604 }, { "epoch": 1.39, "learning_rate": 2.6844968913559644e-05, "loss": 0.0, "step": 71606 }, { "epoch": 1.39, "learning_rate": 2.684432195330241e-05, "loss": 0.0, "step": 71608 }, { "epoch": 1.39, "learning_rate": 2.6843674993045175e-05, "loss": 0.0001, "step": 71610 }, { "epoch": 1.39, "learning_rate": 2.6843028032787948e-05, "loss": 0.0, "step": 71612 }, { "epoch": 1.39, "learning_rate": 2.6842381072530713e-05, "loss": 0.0, "step": 71614 }, { "epoch": 1.39, "learning_rate": 2.6841734112273486e-05, "loss": 0.0, "step": 71616 }, { "epoch": 1.39, "learning_rate": 2.684108715201625e-05, "loss": 0.0057, "step": 71618 }, { "epoch": 1.39, "learning_rate": 2.684044019175902e-05, "loss": 0.0, "step": 71620 }, { "epoch": 1.39, "learning_rate": 2.6839793231501793e-05, "loss": 0.0004, "step": 71622 }, { "epoch": 1.39, "learning_rate": 2.683914627124456e-05, "loss": 0.0, "step": 71624 }, { "epoch": 1.39, "learning_rate": 2.6838499310987324e-05, "loss": 0.0001, "step": 71626 }, { "epoch": 1.39, "learning_rate": 2.6837852350730097e-05, "loss": 0.0051, "step": 71628 }, { "epoch": 1.39, "learning_rate": 2.6837205390472863e-05, "loss": 0.0011, "step": 71630 }, { "epoch": 1.39, "learning_rate": 2.6836558430215635e-05, "loss": 0.0, "step": 71632 }, { "epoch": 1.39, "learning_rate": 2.68359114699584e-05, "loss": 0.0, "step": 71634 }, { "epoch": 1.39, "learning_rate": 2.683526450970117e-05, "loss": 0.0, "step": 71636 }, { "epoch": 1.39, "learning_rate": 2.6834617549443942e-05, "loss": 0.0, "step": 71638 }, { "epoch": 1.39, "learning_rate": 2.6833970589186708e-05, "loss": 0.0134, "step": 71640 }, { "epoch": 1.39, "learning_rate": 2.6833323628929474e-05, "loss": 0.0016, "step": 71642 }, { "epoch": 1.39, "learning_rate": 2.6832676668672246e-05, "loss": 0.0002, "step": 71644 }, { "epoch": 1.39, "learning_rate": 2.6832029708415012e-05, "loss": 0.0, "step": 71646 }, { "epoch": 1.39, "learning_rate": 2.683138274815778e-05, "loss": 0.0, "step": 71648 }, { "epoch": 1.39, "learning_rate": 2.683073578790055e-05, "loss": 0.0, "step": 71650 }, { "epoch": 1.39, "learning_rate": 2.683008882764332e-05, "loss": 0.0, "step": 71652 }, { "epoch": 1.39, "learning_rate": 2.682944186738609e-05, "loss": 0.0003, "step": 71654 }, { "epoch": 1.39, "learning_rate": 2.6828794907128857e-05, "loss": 0.0, "step": 71656 }, { "epoch": 1.39, "learning_rate": 2.6828147946871623e-05, "loss": 0.0008, "step": 71658 }, { "epoch": 1.39, "learning_rate": 2.6827500986614396e-05, "loss": 0.0, "step": 71660 }, { "epoch": 1.39, "learning_rate": 2.682685402635716e-05, "loss": 0.0, "step": 71662 }, { "epoch": 1.39, "learning_rate": 2.682620706609993e-05, "loss": 0.0026, "step": 71664 }, { "epoch": 1.39, "learning_rate": 2.68255601058427e-05, "loss": 0.0009, "step": 71666 }, { "epoch": 1.39, "learning_rate": 2.682491314558547e-05, "loss": 0.0, "step": 71668 }, { "epoch": 1.39, "learning_rate": 2.6824266185328234e-05, "loss": 0.0, "step": 71670 }, { "epoch": 1.39, "learning_rate": 2.6823619225071007e-05, "loss": 0.0038, "step": 71672 }, { "epoch": 1.39, "learning_rate": 2.6822972264813772e-05, "loss": 0.0, "step": 71674 }, { "epoch": 1.39, "learning_rate": 2.6822325304556545e-05, "loss": 0.0046, "step": 71676 }, { "epoch": 1.39, "learning_rate": 2.682167834429931e-05, "loss": 0.0014, "step": 71678 }, { "epoch": 1.39, "learning_rate": 2.682103138404208e-05, "loss": 0.0, "step": 71680 }, { "epoch": 1.39, "learning_rate": 2.682038442378485e-05, "loss": 0.0, "step": 71682 }, { "epoch": 1.39, "learning_rate": 2.6819737463527618e-05, "loss": 0.0, "step": 71684 }, { "epoch": 1.39, "learning_rate": 2.6819090503270383e-05, "loss": 0.0001, "step": 71686 }, { "epoch": 1.39, "learning_rate": 2.6818443543013156e-05, "loss": 0.0, "step": 71688 }, { "epoch": 1.39, "learning_rate": 2.681779658275592e-05, "loss": 0.0, "step": 71690 }, { "epoch": 1.39, "learning_rate": 2.6817149622498687e-05, "loss": 0.0, "step": 71692 }, { "epoch": 1.39, "learning_rate": 2.681650266224146e-05, "loss": 0.0, "step": 71694 }, { "epoch": 1.39, "learning_rate": 2.6815855701984226e-05, "loss": 0.0003, "step": 71696 }, { "epoch": 1.39, "learning_rate": 2.6815208741726998e-05, "loss": 0.0065, "step": 71698 }, { "epoch": 1.39, "learning_rate": 2.6814561781469767e-05, "loss": 0.0006, "step": 71700 }, { "epoch": 1.39, "learning_rate": 2.6813914821212533e-05, "loss": 0.0, "step": 71702 }, { "epoch": 1.39, "learning_rate": 2.6813267860955305e-05, "loss": 0.0185, "step": 71704 }, { "epoch": 1.39, "learning_rate": 2.681262090069807e-05, "loss": 0.0, "step": 71706 }, { "epoch": 1.39, "learning_rate": 2.6811973940440837e-05, "loss": 0.0001, "step": 71708 }, { "epoch": 1.39, "learning_rate": 2.681132698018361e-05, "loss": 0.0006, "step": 71710 }, { "epoch": 1.39, "learning_rate": 2.6810680019926375e-05, "loss": 0.0, "step": 71712 }, { "epoch": 1.39, "learning_rate": 2.6810033059669147e-05, "loss": 0.0, "step": 71714 }, { "epoch": 1.39, "learning_rate": 2.6809386099411916e-05, "loss": 0.0, "step": 71716 }, { "epoch": 1.39, "learning_rate": 2.6808739139154682e-05, "loss": 0.0, "step": 71718 }, { "epoch": 1.39, "learning_rate": 2.6808092178897455e-05, "loss": 0.0, "step": 71720 }, { "epoch": 1.39, "learning_rate": 2.680744521864022e-05, "loss": 0.0, "step": 71722 }, { "epoch": 1.39, "learning_rate": 2.6806798258382986e-05, "loss": 0.0, "step": 71724 }, { "epoch": 1.39, "learning_rate": 2.680615129812576e-05, "loss": 0.0, "step": 71726 }, { "epoch": 1.39, "learning_rate": 2.6805504337868524e-05, "loss": 0.0, "step": 71728 }, { "epoch": 1.39, "learning_rate": 2.6804857377611293e-05, "loss": 0.0, "step": 71730 }, { "epoch": 1.39, "learning_rate": 2.6804210417354066e-05, "loss": 0.0, "step": 71732 }, { "epoch": 1.39, "learning_rate": 2.680356345709683e-05, "loss": 0.0, "step": 71734 }, { "epoch": 1.39, "learning_rate": 2.6802916496839604e-05, "loss": 0.0, "step": 71736 }, { "epoch": 1.39, "learning_rate": 2.680226953658237e-05, "loss": 0.0, "step": 71738 }, { "epoch": 1.39, "learning_rate": 2.6801622576325135e-05, "loss": 0.0, "step": 71740 }, { "epoch": 1.39, "learning_rate": 2.6800975616067908e-05, "loss": 0.0, "step": 71742 }, { "epoch": 1.39, "learning_rate": 2.6800328655810673e-05, "loss": 0.0, "step": 71744 }, { "epoch": 1.39, "learning_rate": 2.6799681695553442e-05, "loss": 0.0, "step": 71746 }, { "epoch": 1.39, "learning_rate": 2.6799034735296215e-05, "loss": 0.0012, "step": 71748 }, { "epoch": 1.39, "learning_rate": 2.679838777503898e-05, "loss": 0.0, "step": 71750 }, { "epoch": 1.39, "learning_rate": 2.6797740814781746e-05, "loss": 0.008, "step": 71752 }, { "epoch": 1.39, "learning_rate": 2.679709385452452e-05, "loss": 0.0, "step": 71754 }, { "epoch": 1.39, "learning_rate": 2.6796446894267284e-05, "loss": 0.0, "step": 71756 }, { "epoch": 1.39, "learning_rate": 2.6795799934010057e-05, "loss": 0.0001, "step": 71758 }, { "epoch": 1.39, "learning_rate": 2.6795152973752823e-05, "loss": 0.0025, "step": 71760 }, { "epoch": 1.39, "learning_rate": 2.6794506013495592e-05, "loss": 0.0, "step": 71762 }, { "epoch": 1.39, "learning_rate": 2.6793859053238364e-05, "loss": 0.0004, "step": 71764 }, { "epoch": 1.39, "learning_rate": 2.679321209298113e-05, "loss": 0.0, "step": 71766 }, { "epoch": 1.39, "learning_rate": 2.6792565132723896e-05, "loss": 0.0003, "step": 71768 }, { "epoch": 1.39, "learning_rate": 2.6791918172466668e-05, "loss": 0.0, "step": 71770 }, { "epoch": 1.39, "learning_rate": 2.6791271212209434e-05, "loss": 0.0002, "step": 71772 }, { "epoch": 1.39, "learning_rate": 2.6790624251952206e-05, "loss": 0.0, "step": 71774 }, { "epoch": 1.39, "learning_rate": 2.6789977291694972e-05, "loss": 0.0, "step": 71776 }, { "epoch": 1.39, "learning_rate": 2.678933033143774e-05, "loss": 0.0, "step": 71778 }, { "epoch": 1.39, "learning_rate": 2.6788683371180514e-05, "loss": 0.0, "step": 71780 }, { "epoch": 1.39, "learning_rate": 2.678803641092328e-05, "loss": 0.0, "step": 71782 }, { "epoch": 1.39, "learning_rate": 2.6787389450666045e-05, "loss": 0.0004, "step": 71784 }, { "epoch": 1.39, "learning_rate": 2.6786742490408817e-05, "loss": 0.0, "step": 71786 }, { "epoch": 1.39, "learning_rate": 2.6786095530151583e-05, "loss": 0.0, "step": 71788 }, { "epoch": 1.39, "learning_rate": 2.678544856989435e-05, "loss": 0.0003, "step": 71790 }, { "epoch": 1.39, "learning_rate": 2.678480160963712e-05, "loss": 0.0, "step": 71792 }, { "epoch": 1.39, "learning_rate": 2.678415464937989e-05, "loss": 0.0, "step": 71794 }, { "epoch": 1.39, "learning_rate": 2.678350768912266e-05, "loss": 0.0, "step": 71796 }, { "epoch": 1.39, "learning_rate": 2.678286072886543e-05, "loss": 0.0, "step": 71798 }, { "epoch": 1.39, "learning_rate": 2.6782213768608194e-05, "loss": 0.0, "step": 71800 }, { "epoch": 1.39, "learning_rate": 2.6781566808350967e-05, "loss": 0.0067, "step": 71802 }, { "epoch": 1.39, "learning_rate": 2.6780919848093732e-05, "loss": 0.0004, "step": 71804 }, { "epoch": 1.39, "learning_rate": 2.6780272887836498e-05, "loss": 0.0, "step": 71806 }, { "epoch": 1.39, "learning_rate": 2.677962592757927e-05, "loss": 0.0001, "step": 71808 }, { "epoch": 1.39, "learning_rate": 2.677897896732204e-05, "loss": 0.0, "step": 71810 }, { "epoch": 1.39, "learning_rate": 2.6778332007064805e-05, "loss": 0.0, "step": 71812 }, { "epoch": 1.39, "learning_rate": 2.6777685046807578e-05, "loss": 0.0, "step": 71814 }, { "epoch": 1.39, "learning_rate": 2.6777038086550343e-05, "loss": 0.0, "step": 71816 }, { "epoch": 1.39, "learning_rate": 2.6776391126293116e-05, "loss": 0.0, "step": 71818 }, { "epoch": 1.39, "learning_rate": 2.677574416603588e-05, "loss": 0.0, "step": 71820 }, { "epoch": 1.39, "learning_rate": 2.6775097205778647e-05, "loss": 0.0002, "step": 71822 }, { "epoch": 1.39, "learning_rate": 2.677445024552142e-05, "loss": 0.0, "step": 71824 }, { "epoch": 1.39, "learning_rate": 2.677380328526419e-05, "loss": 0.0, "step": 71826 }, { "epoch": 1.39, "learning_rate": 2.6773156325006955e-05, "loss": 0.0, "step": 71828 }, { "epoch": 1.39, "learning_rate": 2.6772509364749727e-05, "loss": 0.0, "step": 71830 }, { "epoch": 1.39, "learning_rate": 2.6771862404492493e-05, "loss": 0.0, "step": 71832 }, { "epoch": 1.39, "learning_rate": 2.677121544423526e-05, "loss": 0.0001, "step": 71834 }, { "epoch": 1.39, "learning_rate": 2.677056848397803e-05, "loss": 0.0001, "step": 71836 }, { "epoch": 1.39, "learning_rate": 2.6769921523720797e-05, "loss": 0.0069, "step": 71838 }, { "epoch": 1.39, "learning_rate": 2.676927456346357e-05, "loss": 0.0, "step": 71840 }, { "epoch": 1.39, "learning_rate": 2.6768627603206338e-05, "loss": 0.0003, "step": 71842 }, { "epoch": 1.39, "learning_rate": 2.6767980642949104e-05, "loss": 0.0, "step": 71844 }, { "epoch": 1.39, "learning_rate": 2.6767333682691876e-05, "loss": 0.004, "step": 71846 }, { "epoch": 1.39, "learning_rate": 2.6766686722434642e-05, "loss": 0.0, "step": 71848 }, { "epoch": 1.39, "learning_rate": 2.6766039762177408e-05, "loss": 0.0, "step": 71850 }, { "epoch": 1.39, "learning_rate": 2.676539280192018e-05, "loss": 0.0, "step": 71852 }, { "epoch": 1.39, "learning_rate": 2.6764745841662946e-05, "loss": 0.0, "step": 71854 }, { "epoch": 1.39, "learning_rate": 2.676409888140572e-05, "loss": 0.0, "step": 71856 }, { "epoch": 1.39, "learning_rate": 2.6763451921148484e-05, "loss": 0.0, "step": 71858 }, { "epoch": 1.39, "learning_rate": 2.6762804960891253e-05, "loss": 0.0, "step": 71860 }, { "epoch": 1.39, "learning_rate": 2.6762158000634026e-05, "loss": 0.0, "step": 71862 }, { "epoch": 1.39, "learning_rate": 2.676151104037679e-05, "loss": 0.0, "step": 71864 }, { "epoch": 1.39, "learning_rate": 2.6760864080119557e-05, "loss": 0.0015, "step": 71866 }, { "epoch": 1.39, "learning_rate": 2.676021711986233e-05, "loss": 0.0, "step": 71868 }, { "epoch": 1.39, "learning_rate": 2.6759570159605095e-05, "loss": 0.0, "step": 71870 }, { "epoch": 1.39, "learning_rate": 2.6758923199347864e-05, "loss": 0.0, "step": 71872 }, { "epoch": 1.39, "learning_rate": 2.6758276239090633e-05, "loss": 0.0, "step": 71874 }, { "epoch": 1.4, "learning_rate": 2.6757629278833402e-05, "loss": 0.0, "step": 71876 }, { "epoch": 1.4, "learning_rate": 2.6756982318576175e-05, "loss": 0.0, "step": 71878 }, { "epoch": 1.4, "learning_rate": 2.675633535831894e-05, "loss": 0.0005, "step": 71880 }, { "epoch": 1.4, "learning_rate": 2.6755688398061706e-05, "loss": 0.0, "step": 71882 }, { "epoch": 1.4, "learning_rate": 2.675504143780448e-05, "loss": 0.0, "step": 71884 }, { "epoch": 1.4, "learning_rate": 2.6754394477547244e-05, "loss": 0.0, "step": 71886 }, { "epoch": 1.4, "learning_rate": 2.6753747517290014e-05, "loss": 0.0, "step": 71888 }, { "epoch": 1.4, "learning_rate": 2.6753100557032783e-05, "loss": 0.0, "step": 71890 }, { "epoch": 1.4, "learning_rate": 2.6752453596775552e-05, "loss": 0.0, "step": 71892 }, { "epoch": 1.4, "learning_rate": 2.6751806636518317e-05, "loss": 0.0, "step": 71894 }, { "epoch": 1.4, "learning_rate": 2.675115967626109e-05, "loss": 0.0, "step": 71896 }, { "epoch": 1.4, "learning_rate": 2.6750512716003856e-05, "loss": 0.0025, "step": 71898 }, { "epoch": 1.4, "learning_rate": 2.6749865755746628e-05, "loss": 0.0, "step": 71900 }, { "epoch": 1.4, "learning_rate": 2.6749218795489394e-05, "loss": 0.023, "step": 71902 }, { "epoch": 1.4, "learning_rate": 2.6748571835232163e-05, "loss": 0.0, "step": 71904 }, { "epoch": 1.4, "learning_rate": 2.6747924874974932e-05, "loss": 0.0, "step": 71906 }, { "epoch": 1.4, "learning_rate": 2.67472779147177e-05, "loss": 0.0, "step": 71908 }, { "epoch": 1.4, "learning_rate": 2.6746630954460467e-05, "loss": 0.0, "step": 71910 }, { "epoch": 1.4, "learning_rate": 2.674598399420324e-05, "loss": 0.0, "step": 71912 }, { "epoch": 1.4, "learning_rate": 2.6745337033946005e-05, "loss": 0.0, "step": 71914 }, { "epoch": 1.4, "learning_rate": 2.6744690073688777e-05, "loss": 0.0, "step": 71916 }, { "epoch": 1.4, "learning_rate": 2.6744043113431543e-05, "loss": 0.0063, "step": 71918 }, { "epoch": 1.4, "learning_rate": 2.6743396153174312e-05, "loss": 0.0, "step": 71920 }, { "epoch": 1.4, "learning_rate": 2.674274919291708e-05, "loss": 0.0, "step": 71922 }, { "epoch": 1.4, "learning_rate": 2.674210223265985e-05, "loss": 0.0, "step": 71924 }, { "epoch": 1.4, "learning_rate": 2.6741455272402616e-05, "loss": 0.0005, "step": 71926 }, { "epoch": 1.4, "learning_rate": 2.674080831214539e-05, "loss": 0.0, "step": 71928 }, { "epoch": 1.4, "learning_rate": 2.6740161351888154e-05, "loss": 0.0048, "step": 71930 }, { "epoch": 1.4, "learning_rate": 2.673951439163092e-05, "loss": 0.0, "step": 71932 }, { "epoch": 1.4, "learning_rate": 2.6738867431373692e-05, "loss": 0.0, "step": 71934 }, { "epoch": 1.4, "learning_rate": 2.6738220471116458e-05, "loss": 0.0, "step": 71936 }, { "epoch": 1.4, "learning_rate": 2.673757351085923e-05, "loss": 0.0001, "step": 71938 }, { "epoch": 1.4, "learning_rate": 2.6736926550602e-05, "loss": 0.0, "step": 71940 }, { "epoch": 1.4, "learning_rate": 2.6736279590344765e-05, "loss": 0.0, "step": 71942 }, { "epoch": 1.4, "learning_rate": 2.6735632630087538e-05, "loss": 0.0075, "step": 71944 }, { "epoch": 1.4, "learning_rate": 2.6734985669830303e-05, "loss": 0.0, "step": 71946 }, { "epoch": 1.4, "learning_rate": 2.673433870957307e-05, "loss": 0.0, "step": 71948 }, { "epoch": 1.4, "learning_rate": 2.673369174931584e-05, "loss": 0.0, "step": 71950 }, { "epoch": 1.4, "learning_rate": 2.6733044789058607e-05, "loss": 0.0, "step": 71952 }, { "epoch": 1.4, "learning_rate": 2.6732397828801376e-05, "loss": 0.0, "step": 71954 }, { "epoch": 1.4, "learning_rate": 2.673175086854415e-05, "loss": 0.0, "step": 71956 }, { "epoch": 1.4, "learning_rate": 2.6731103908286915e-05, "loss": 0.0, "step": 71958 }, { "epoch": 1.4, "learning_rate": 2.6730456948029687e-05, "loss": 0.0, "step": 71960 }, { "epoch": 1.4, "learning_rate": 2.6729809987772453e-05, "loss": 0.0001, "step": 71962 }, { "epoch": 1.4, "learning_rate": 2.672916302751522e-05, "loss": 0.0, "step": 71964 }, { "epoch": 1.4, "learning_rate": 2.672851606725799e-05, "loss": 0.0, "step": 71966 }, { "epoch": 1.4, "learning_rate": 2.6727869107000757e-05, "loss": 0.0007, "step": 71968 }, { "epoch": 1.4, "learning_rate": 2.6727222146743526e-05, "loss": 0.0002, "step": 71970 }, { "epoch": 1.4, "learning_rate": 2.6726575186486298e-05, "loss": 0.0002, "step": 71972 }, { "epoch": 1.4, "learning_rate": 2.6725928226229064e-05, "loss": 0.0, "step": 71974 }, { "epoch": 1.4, "learning_rate": 2.672528126597183e-05, "loss": 0.0, "step": 71976 }, { "epoch": 1.4, "learning_rate": 2.6724634305714602e-05, "loss": 0.0, "step": 71978 }, { "epoch": 1.4, "learning_rate": 2.6723987345457368e-05, "loss": 0.0, "step": 71980 }, { "epoch": 1.4, "learning_rate": 2.672334038520014e-05, "loss": 0.0, "step": 71982 }, { "epoch": 1.4, "learning_rate": 2.6722693424942906e-05, "loss": 0.0, "step": 71984 }, { "epoch": 1.4, "learning_rate": 2.6722046464685675e-05, "loss": 0.0, "step": 71986 }, { "epoch": 1.4, "learning_rate": 2.6721399504428447e-05, "loss": 0.0, "step": 71988 }, { "epoch": 1.4, "learning_rate": 2.6720752544171213e-05, "loss": 0.0004, "step": 71990 }, { "epoch": 1.4, "learning_rate": 2.672010558391398e-05, "loss": 0.0001, "step": 71992 }, { "epoch": 1.4, "learning_rate": 2.671945862365675e-05, "loss": 0.0, "step": 71994 }, { "epoch": 1.4, "learning_rate": 2.6718811663399517e-05, "loss": 0.0005, "step": 71996 }, { "epoch": 1.4, "learning_rate": 2.671816470314229e-05, "loss": 0.0, "step": 71998 }, { "epoch": 1.4, "learning_rate": 2.6717517742885055e-05, "loss": 0.0, "step": 72000 }, { "epoch": 1.4, "learning_rate": 2.6716870782627824e-05, "loss": 0.0, "step": 72002 }, { "epoch": 1.4, "learning_rate": 2.6716223822370597e-05, "loss": 0.0, "step": 72004 }, { "epoch": 1.4, "learning_rate": 2.6715576862113362e-05, "loss": 0.0, "step": 72006 }, { "epoch": 1.4, "learning_rate": 2.6714929901856128e-05, "loss": 0.0, "step": 72008 }, { "epoch": 1.4, "learning_rate": 2.67142829415989e-05, "loss": 0.0, "step": 72010 }, { "epoch": 1.4, "learning_rate": 2.6713635981341666e-05, "loss": 0.0, "step": 72012 }, { "epoch": 1.4, "learning_rate": 2.6712989021084432e-05, "loss": 0.0, "step": 72014 }, { "epoch": 1.4, "learning_rate": 2.6712342060827204e-05, "loss": 0.0, "step": 72016 }, { "epoch": 1.4, "learning_rate": 2.6711695100569974e-05, "loss": 0.0, "step": 72018 }, { "epoch": 1.4, "learning_rate": 2.6711048140312746e-05, "loss": 0.0, "step": 72020 }, { "epoch": 1.4, "learning_rate": 2.6710401180055512e-05, "loss": 0.0, "step": 72022 }, { "epoch": 1.4, "learning_rate": 2.6709754219798277e-05, "loss": 0.0, "step": 72024 }, { "epoch": 1.4, "learning_rate": 2.670910725954105e-05, "loss": 0.0001, "step": 72026 }, { "epoch": 1.4, "learning_rate": 2.6708460299283816e-05, "loss": 0.0, "step": 72028 }, { "epoch": 1.4, "learning_rate": 2.670781333902658e-05, "loss": 0.0, "step": 72030 }, { "epoch": 1.4, "learning_rate": 2.6707166378769354e-05, "loss": 0.0008, "step": 72032 }, { "epoch": 1.4, "learning_rate": 2.6706519418512123e-05, "loss": 0.0081, "step": 72034 }, { "epoch": 1.4, "learning_rate": 2.670587245825489e-05, "loss": 0.0, "step": 72036 }, { "epoch": 1.4, "learning_rate": 2.670522549799766e-05, "loss": 0.0, "step": 72038 }, { "epoch": 1.4, "learning_rate": 2.6704578537740427e-05, "loss": 0.0, "step": 72040 }, { "epoch": 1.4, "learning_rate": 2.67039315774832e-05, "loss": 0.0, "step": 72042 }, { "epoch": 1.4, "learning_rate": 2.6703284617225965e-05, "loss": 0.0, "step": 72044 }, { "epoch": 1.4, "learning_rate": 2.670263765696873e-05, "loss": 0.0, "step": 72046 }, { "epoch": 1.4, "learning_rate": 2.6701990696711503e-05, "loss": 0.0001, "step": 72048 }, { "epoch": 1.4, "learning_rate": 2.6701343736454272e-05, "loss": 0.0, "step": 72050 }, { "epoch": 1.4, "learning_rate": 2.6700696776197038e-05, "loss": 0.0, "step": 72052 }, { "epoch": 1.4, "learning_rate": 2.670004981593981e-05, "loss": 0.0002, "step": 72054 }, { "epoch": 1.4, "learning_rate": 2.6699402855682576e-05, "loss": 0.0, "step": 72056 }, { "epoch": 1.4, "learning_rate": 2.669875589542535e-05, "loss": 0.0, "step": 72058 }, { "epoch": 1.4, "learning_rate": 2.6698108935168114e-05, "loss": 0.0002, "step": 72060 }, { "epoch": 1.4, "learning_rate": 2.669746197491088e-05, "loss": 0.0, "step": 72062 }, { "epoch": 1.4, "learning_rate": 2.6696815014653652e-05, "loss": 0.0, "step": 72064 }, { "epoch": 1.4, "learning_rate": 2.669616805439642e-05, "loss": 0.0009, "step": 72066 }, { "epoch": 1.4, "learning_rate": 2.6695521094139187e-05, "loss": 0.0, "step": 72068 }, { "epoch": 1.4, "learning_rate": 2.669487413388196e-05, "loss": 0.0, "step": 72070 }, { "epoch": 1.4, "learning_rate": 2.6694227173624725e-05, "loss": 0.0, "step": 72072 }, { "epoch": 1.4, "learning_rate": 2.669358021336749e-05, "loss": 0.0001, "step": 72074 }, { "epoch": 1.4, "learning_rate": 2.6692933253110263e-05, "loss": 0.0, "step": 72076 }, { "epoch": 1.4, "learning_rate": 2.669228629285303e-05, "loss": 0.0, "step": 72078 }, { "epoch": 1.4, "learning_rate": 2.66916393325958e-05, "loss": 0.0, "step": 72080 }, { "epoch": 1.4, "learning_rate": 2.669099237233857e-05, "loss": 0.0, "step": 72082 }, { "epoch": 1.4, "learning_rate": 2.6690345412081336e-05, "loss": 0.0001, "step": 72084 }, { "epoch": 1.4, "learning_rate": 2.668969845182411e-05, "loss": 0.0001, "step": 72086 }, { "epoch": 1.4, "learning_rate": 2.6689051491566875e-05, "loss": 0.0, "step": 72088 }, { "epoch": 1.4, "learning_rate": 2.668840453130964e-05, "loss": 0.0, "step": 72090 }, { "epoch": 1.4, "learning_rate": 2.6687757571052413e-05, "loss": 0.0, "step": 72092 }, { "epoch": 1.4, "learning_rate": 2.668711061079518e-05, "loss": 0.0, "step": 72094 }, { "epoch": 1.4, "learning_rate": 2.6686463650537948e-05, "loss": 0.0001, "step": 72096 }, { "epoch": 1.4, "learning_rate": 2.6685816690280717e-05, "loss": 0.0001, "step": 72098 }, { "epoch": 1.4, "learning_rate": 2.6685169730023486e-05, "loss": 0.0, "step": 72100 }, { "epoch": 1.4, "learning_rate": 2.6684522769766258e-05, "loss": 0.0, "step": 72102 }, { "epoch": 1.4, "learning_rate": 2.6683875809509024e-05, "loss": 0.0, "step": 72104 }, { "epoch": 1.4, "learning_rate": 2.668322884925179e-05, "loss": 0.0, "step": 72106 }, { "epoch": 1.4, "learning_rate": 2.6682581888994562e-05, "loss": 0.0041, "step": 72108 }, { "epoch": 1.4, "learning_rate": 2.6681934928737328e-05, "loss": 0.0, "step": 72110 }, { "epoch": 1.4, "learning_rate": 2.6681287968480097e-05, "loss": 0.0, "step": 72112 }, { "epoch": 1.4, "learning_rate": 2.6680641008222866e-05, "loss": 0.0, "step": 72114 }, { "epoch": 1.4, "learning_rate": 2.6679994047965635e-05, "loss": 0.0, "step": 72116 }, { "epoch": 1.4, "learning_rate": 2.66793470877084e-05, "loss": 0.0001, "step": 72118 }, { "epoch": 1.4, "learning_rate": 2.6678700127451173e-05, "loss": 0.0, "step": 72120 }, { "epoch": 1.4, "learning_rate": 2.667805316719394e-05, "loss": 0.0161, "step": 72122 }, { "epoch": 1.4, "learning_rate": 2.667740620693671e-05, "loss": 0.0, "step": 72124 }, { "epoch": 1.4, "learning_rate": 2.6676759246679477e-05, "loss": 0.0, "step": 72126 }, { "epoch": 1.4, "learning_rate": 2.6676112286422246e-05, "loss": 0.0, "step": 72128 }, { "epoch": 1.4, "learning_rate": 2.6675465326165015e-05, "loss": 0.0, "step": 72130 }, { "epoch": 1.4, "learning_rate": 2.6674818365907784e-05, "loss": 0.0079, "step": 72132 }, { "epoch": 1.4, "learning_rate": 2.667417140565055e-05, "loss": 0.0, "step": 72134 }, { "epoch": 1.4, "learning_rate": 2.6673524445393322e-05, "loss": 0.0, "step": 72136 }, { "epoch": 1.4, "learning_rate": 2.6672877485136088e-05, "loss": 0.0007, "step": 72138 }, { "epoch": 1.4, "learning_rate": 2.667223052487886e-05, "loss": 0.0, "step": 72140 }, { "epoch": 1.4, "learning_rate": 2.6671583564621626e-05, "loss": 0.0, "step": 72142 }, { "epoch": 1.4, "learning_rate": 2.6670936604364395e-05, "loss": 0.0, "step": 72144 }, { "epoch": 1.4, "learning_rate": 2.6670289644107164e-05, "loss": 0.0, "step": 72146 }, { "epoch": 1.4, "learning_rate": 2.6669642683849934e-05, "loss": 0.0002, "step": 72148 }, { "epoch": 1.4, "learning_rate": 2.66689957235927e-05, "loss": 0.0, "step": 72150 }, { "epoch": 1.4, "learning_rate": 2.6668348763335472e-05, "loss": 0.0, "step": 72152 }, { "epoch": 1.4, "learning_rate": 2.6667701803078237e-05, "loss": 0.0, "step": 72154 }, { "epoch": 1.4, "learning_rate": 2.6667054842821003e-05, "loss": 0.0094, "step": 72156 }, { "epoch": 1.4, "learning_rate": 2.6666407882563776e-05, "loss": 0.0, "step": 72158 }, { "epoch": 1.4, "learning_rate": 2.6665760922306545e-05, "loss": 0.0, "step": 72160 }, { "epoch": 1.4, "learning_rate": 2.6665113962049314e-05, "loss": 0.0002, "step": 72162 }, { "epoch": 1.4, "learning_rate": 2.6664467001792083e-05, "loss": 0.0, "step": 72164 }, { "epoch": 1.4, "learning_rate": 2.666382004153485e-05, "loss": 0.0, "step": 72166 }, { "epoch": 1.4, "learning_rate": 2.666317308127762e-05, "loss": 0.0, "step": 72168 }, { "epoch": 1.4, "learning_rate": 2.6662526121020387e-05, "loss": 0.0, "step": 72170 }, { "epoch": 1.4, "learning_rate": 2.6661879160763152e-05, "loss": 0.0, "step": 72172 }, { "epoch": 1.4, "learning_rate": 2.6661232200505925e-05, "loss": 0.0, "step": 72174 }, { "epoch": 1.4, "learning_rate": 2.666058524024869e-05, "loss": 0.0004, "step": 72176 }, { "epoch": 1.4, "learning_rate": 2.665993827999146e-05, "loss": 0.0, "step": 72178 }, { "epoch": 1.4, "learning_rate": 2.6659291319734232e-05, "loss": 0.0, "step": 72180 }, { "epoch": 1.4, "learning_rate": 2.6658644359476998e-05, "loss": 0.0, "step": 72182 }, { "epoch": 1.4, "learning_rate": 2.665799739921977e-05, "loss": 0.0, "step": 72184 }, { "epoch": 1.4, "learning_rate": 2.6657350438962536e-05, "loss": 0.0, "step": 72186 }, { "epoch": 1.4, "learning_rate": 2.66567034787053e-05, "loss": 0.0, "step": 72188 }, { "epoch": 1.4, "learning_rate": 2.6656056518448074e-05, "loss": 0.0, "step": 72190 }, { "epoch": 1.4, "learning_rate": 2.665540955819084e-05, "loss": 0.0, "step": 72192 }, { "epoch": 1.4, "learning_rate": 2.665476259793361e-05, "loss": 0.0, "step": 72194 }, { "epoch": 1.4, "learning_rate": 2.665411563767638e-05, "loss": 0.0, "step": 72196 }, { "epoch": 1.4, "learning_rate": 2.6653468677419147e-05, "loss": 0.0, "step": 72198 }, { "epoch": 1.4, "learning_rate": 2.665282171716192e-05, "loss": 0.0, "step": 72200 }, { "epoch": 1.4, "learning_rate": 2.6652174756904685e-05, "loss": 0.0, "step": 72202 }, { "epoch": 1.4, "learning_rate": 2.665152779664745e-05, "loss": 0.0, "step": 72204 }, { "epoch": 1.4, "learning_rate": 2.6650880836390223e-05, "loss": 0.0, "step": 72206 }, { "epoch": 1.4, "learning_rate": 2.665023387613299e-05, "loss": 0.0033, "step": 72208 }, { "epoch": 1.4, "learning_rate": 2.6649586915875758e-05, "loss": 0.0, "step": 72210 }, { "epoch": 1.4, "learning_rate": 2.664893995561853e-05, "loss": 0.0, "step": 72212 }, { "epoch": 1.4, "learning_rate": 2.6648292995361296e-05, "loss": 0.0, "step": 72214 }, { "epoch": 1.4, "learning_rate": 2.6647646035104062e-05, "loss": 0.0, "step": 72216 }, { "epoch": 1.4, "learning_rate": 2.6646999074846835e-05, "loss": 0.0, "step": 72218 }, { "epoch": 1.4, "learning_rate": 2.66463521145896e-05, "loss": 0.0, "step": 72220 }, { "epoch": 1.4, "learning_rate": 2.6645705154332373e-05, "loss": 0.0001, "step": 72222 }, { "epoch": 1.4, "learning_rate": 2.664505819407514e-05, "loss": 0.0, "step": 72224 }, { "epoch": 1.4, "learning_rate": 2.6644411233817908e-05, "loss": 0.0, "step": 72226 }, { "epoch": 1.4, "learning_rate": 2.664376427356068e-05, "loss": 0.0, "step": 72228 }, { "epoch": 1.4, "learning_rate": 2.6643117313303446e-05, "loss": 0.0, "step": 72230 }, { "epoch": 1.4, "learning_rate": 2.664247035304621e-05, "loss": 0.0, "step": 72232 }, { "epoch": 1.4, "learning_rate": 2.6641823392788984e-05, "loss": 0.0, "step": 72234 }, { "epoch": 1.4, "learning_rate": 2.664117643253175e-05, "loss": 0.0002, "step": 72236 }, { "epoch": 1.4, "learning_rate": 2.6640529472274515e-05, "loss": 0.0, "step": 72238 }, { "epoch": 1.4, "learning_rate": 2.6639882512017288e-05, "loss": 0.0, "step": 72240 }, { "epoch": 1.4, "learning_rate": 2.6639235551760057e-05, "loss": 0.0, "step": 72242 }, { "epoch": 1.4, "learning_rate": 2.663858859150283e-05, "loss": 0.0002, "step": 72244 }, { "epoch": 1.4, "learning_rate": 2.6637941631245595e-05, "loss": 0.0067, "step": 72246 }, { "epoch": 1.4, "learning_rate": 2.663729467098836e-05, "loss": 0.0, "step": 72248 }, { "epoch": 1.4, "learning_rate": 2.6636647710731133e-05, "loss": 0.0, "step": 72250 }, { "epoch": 1.4, "learning_rate": 2.66360007504739e-05, "loss": 0.0, "step": 72252 }, { "epoch": 1.4, "learning_rate": 2.6635353790216665e-05, "loss": 0.0, "step": 72254 }, { "epoch": 1.4, "learning_rate": 2.6634706829959437e-05, "loss": 0.0023, "step": 72256 }, { "epoch": 1.4, "learning_rate": 2.6634059869702206e-05, "loss": 0.0, "step": 72258 }, { "epoch": 1.4, "learning_rate": 2.6633412909444972e-05, "loss": 0.0, "step": 72260 }, { "epoch": 1.4, "learning_rate": 2.6632765949187744e-05, "loss": 0.0, "step": 72262 }, { "epoch": 1.4, "learning_rate": 2.663211898893051e-05, "loss": 0.0084, "step": 72264 }, { "epoch": 1.4, "learning_rate": 2.6631472028673282e-05, "loss": 0.0001, "step": 72266 }, { "epoch": 1.4, "learning_rate": 2.6630825068416048e-05, "loss": 0.0, "step": 72268 }, { "epoch": 1.4, "learning_rate": 2.6630178108158814e-05, "loss": 0.0059, "step": 72270 }, { "epoch": 1.4, "learning_rate": 2.6629531147901586e-05, "loss": 0.0, "step": 72272 }, { "epoch": 1.4, "learning_rate": 2.6628884187644355e-05, "loss": 0.0, "step": 72274 }, { "epoch": 1.4, "learning_rate": 2.662823722738712e-05, "loss": 0.0, "step": 72276 }, { "epoch": 1.4, "learning_rate": 2.6627590267129894e-05, "loss": 0.0, "step": 72278 }, { "epoch": 1.4, "learning_rate": 2.662694330687266e-05, "loss": 0.0217, "step": 72280 }, { "epoch": 1.4, "learning_rate": 2.6626296346615432e-05, "loss": 0.0, "step": 72282 }, { "epoch": 1.4, "learning_rate": 2.6625649386358197e-05, "loss": 0.0, "step": 72284 }, { "epoch": 1.4, "learning_rate": 2.6625002426100963e-05, "loss": 0.0, "step": 72286 }, { "epoch": 1.4, "learning_rate": 2.6624355465843736e-05, "loss": 0.0, "step": 72288 }, { "epoch": 1.4, "learning_rate": 2.6623708505586505e-05, "loss": 0.0, "step": 72290 }, { "epoch": 1.4, "learning_rate": 2.662306154532927e-05, "loss": 0.0, "step": 72292 }, { "epoch": 1.4, "learning_rate": 2.6622414585072043e-05, "loss": 0.0, "step": 72294 }, { "epoch": 1.4, "learning_rate": 2.662176762481481e-05, "loss": 0.0, "step": 72296 }, { "epoch": 1.4, "learning_rate": 2.6621120664557574e-05, "loss": 0.0002, "step": 72298 }, { "epoch": 1.4, "learning_rate": 2.6620473704300347e-05, "loss": 0.0, "step": 72300 }, { "epoch": 1.4, "learning_rate": 2.6619826744043112e-05, "loss": 0.0, "step": 72302 }, { "epoch": 1.4, "learning_rate": 2.6619179783785885e-05, "loss": 0.0, "step": 72304 }, { "epoch": 1.4, "learning_rate": 2.6618532823528654e-05, "loss": 0.0008, "step": 72306 }, { "epoch": 1.4, "learning_rate": 2.661788586327142e-05, "loss": 0.0001, "step": 72308 }, { "epoch": 1.4, "learning_rate": 2.6617238903014192e-05, "loss": 0.0, "step": 72310 }, { "epoch": 1.4, "learning_rate": 2.6616591942756958e-05, "loss": 0.0002, "step": 72312 }, { "epoch": 1.4, "learning_rate": 2.6615944982499724e-05, "loss": 0.0, "step": 72314 }, { "epoch": 1.4, "learning_rate": 2.6615298022242496e-05, "loss": 0.0, "step": 72316 }, { "epoch": 1.4, "learning_rate": 2.661465106198526e-05, "loss": 0.001, "step": 72318 }, { "epoch": 1.4, "learning_rate": 2.661400410172803e-05, "loss": 0.0, "step": 72320 }, { "epoch": 1.4, "learning_rate": 2.6613357141470803e-05, "loss": 0.0, "step": 72322 }, { "epoch": 1.4, "learning_rate": 2.661271018121357e-05, "loss": 0.0, "step": 72324 }, { "epoch": 1.4, "learning_rate": 2.661206322095634e-05, "loss": 0.0, "step": 72326 }, { "epoch": 1.4, "learning_rate": 2.6611416260699107e-05, "loss": 0.0, "step": 72328 }, { "epoch": 1.4, "learning_rate": 2.6610769300441873e-05, "loss": 0.0, "step": 72330 }, { "epoch": 1.4, "learning_rate": 2.6610122340184645e-05, "loss": 0.0069, "step": 72332 }, { "epoch": 1.4, "learning_rate": 2.660947537992741e-05, "loss": 0.0, "step": 72334 }, { "epoch": 1.4, "learning_rate": 2.660882841967018e-05, "loss": 0.0, "step": 72336 }, { "epoch": 1.4, "learning_rate": 2.660818145941295e-05, "loss": 0.0076, "step": 72338 }, { "epoch": 1.4, "learning_rate": 2.6607534499155718e-05, "loss": 0.0016, "step": 72340 }, { "epoch": 1.4, "learning_rate": 2.660688753889849e-05, "loss": 0.0051, "step": 72342 }, { "epoch": 1.4, "learning_rate": 2.6606240578641256e-05, "loss": 0.0, "step": 72344 }, { "epoch": 1.4, "learning_rate": 2.6605593618384022e-05, "loss": 0.0001, "step": 72346 }, { "epoch": 1.4, "learning_rate": 2.6604946658126795e-05, "loss": 0.0, "step": 72348 }, { "epoch": 1.4, "learning_rate": 2.660429969786956e-05, "loss": 0.0031, "step": 72350 }, { "epoch": 1.4, "learning_rate": 2.660365273761233e-05, "loss": 0.0, "step": 72352 }, { "epoch": 1.4, "learning_rate": 2.66030057773551e-05, "loss": 0.0, "step": 72354 }, { "epoch": 1.4, "learning_rate": 2.6602358817097867e-05, "loss": 0.0, "step": 72356 }, { "epoch": 1.4, "learning_rate": 2.6601711856840633e-05, "loss": 0.0, "step": 72358 }, { "epoch": 1.4, "learning_rate": 2.6601064896583406e-05, "loss": 0.0, "step": 72360 }, { "epoch": 1.4, "learning_rate": 2.660041793632617e-05, "loss": 0.0443, "step": 72362 }, { "epoch": 1.4, "learning_rate": 2.6599770976068944e-05, "loss": 0.0, "step": 72364 }, { "epoch": 1.4, "learning_rate": 2.659912401581171e-05, "loss": 0.0, "step": 72366 }, { "epoch": 1.4, "learning_rate": 2.659847705555448e-05, "loss": 0.0079, "step": 72368 }, { "epoch": 1.4, "learning_rate": 2.6597830095297248e-05, "loss": 0.0, "step": 72370 }, { "epoch": 1.4, "learning_rate": 2.6597183135040017e-05, "loss": 0.0, "step": 72372 }, { "epoch": 1.4, "learning_rate": 2.6596536174782782e-05, "loss": 0.0, "step": 72374 }, { "epoch": 1.4, "learning_rate": 2.6595889214525555e-05, "loss": 0.0, "step": 72376 }, { "epoch": 1.4, "learning_rate": 2.659524225426832e-05, "loss": 0.0, "step": 72378 }, { "epoch": 1.4, "learning_rate": 2.6594595294011086e-05, "loss": 0.0, "step": 72380 }, { "epoch": 1.4, "learning_rate": 2.659394833375386e-05, "loss": 0.0, "step": 72382 }, { "epoch": 1.4, "learning_rate": 2.6593301373496628e-05, "loss": 0.0, "step": 72384 }, { "epoch": 1.4, "learning_rate": 2.6592654413239397e-05, "loss": 0.0, "step": 72386 }, { "epoch": 1.4, "learning_rate": 2.6592007452982166e-05, "loss": 0.0, "step": 72388 }, { "epoch": 1.41, "learning_rate": 2.6591360492724932e-05, "loss": 0.0, "step": 72390 }, { "epoch": 1.41, "learning_rate": 2.6590713532467704e-05, "loss": 0.0, "step": 72392 }, { "epoch": 1.41, "learning_rate": 2.659006657221047e-05, "loss": 0.0, "step": 72394 }, { "epoch": 1.41, "learning_rate": 2.6589419611953236e-05, "loss": 0.0001, "step": 72396 }, { "epoch": 1.41, "learning_rate": 2.6588772651696008e-05, "loss": 0.0092, "step": 72398 }, { "epoch": 1.41, "learning_rate": 2.6588125691438777e-05, "loss": 0.0, "step": 72400 }, { "epoch": 1.41, "learning_rate": 2.6587478731181543e-05, "loss": 0.0031, "step": 72402 }, { "epoch": 1.41, "learning_rate": 2.6586831770924315e-05, "loss": 0.0, "step": 72404 }, { "epoch": 1.41, "learning_rate": 2.658618481066708e-05, "loss": 0.0, "step": 72406 }, { "epoch": 1.41, "learning_rate": 2.6585537850409854e-05, "loss": 0.0, "step": 72408 }, { "epoch": 1.41, "learning_rate": 2.658489089015262e-05, "loss": 0.0, "step": 72410 }, { "epoch": 1.41, "learning_rate": 2.6584243929895385e-05, "loss": 0.0, "step": 72412 }, { "epoch": 1.41, "learning_rate": 2.6583596969638157e-05, "loss": 0.0, "step": 72414 }, { "epoch": 1.41, "learning_rate": 2.6582950009380923e-05, "loss": 0.0, "step": 72416 }, { "epoch": 1.41, "learning_rate": 2.6582303049123692e-05, "loss": 0.0, "step": 72418 }, { "epoch": 1.41, "learning_rate": 2.6581656088866465e-05, "loss": 0.0, "step": 72420 }, { "epoch": 1.41, "learning_rate": 2.658100912860923e-05, "loss": 0.0002, "step": 72422 }, { "epoch": 1.41, "learning_rate": 2.6580362168352003e-05, "loss": 0.0, "step": 72424 }, { "epoch": 1.41, "learning_rate": 2.657971520809477e-05, "loss": 0.0, "step": 72426 }, { "epoch": 1.41, "learning_rate": 2.6579068247837534e-05, "loss": 0.0, "step": 72428 }, { "epoch": 1.41, "learning_rate": 2.6578421287580307e-05, "loss": 0.0, "step": 72430 }, { "epoch": 1.41, "learning_rate": 2.6577774327323072e-05, "loss": 0.0001, "step": 72432 }, { "epoch": 1.41, "learning_rate": 2.657712736706584e-05, "loss": 0.001, "step": 72434 }, { "epoch": 1.41, "learning_rate": 2.6576480406808614e-05, "loss": 0.0, "step": 72436 }, { "epoch": 1.41, "learning_rate": 2.657583344655138e-05, "loss": 0.0, "step": 72438 }, { "epoch": 1.41, "learning_rate": 2.6575186486294145e-05, "loss": 0.0, "step": 72440 }, { "epoch": 1.41, "learning_rate": 2.6574539526036918e-05, "loss": 0.0, "step": 72442 }, { "epoch": 1.41, "learning_rate": 2.6573892565779683e-05, "loss": 0.0, "step": 72444 }, { "epoch": 1.41, "learning_rate": 2.6573245605522456e-05, "loss": 0.0089, "step": 72446 }, { "epoch": 1.41, "learning_rate": 2.657259864526522e-05, "loss": 0.0, "step": 72448 }, { "epoch": 1.41, "learning_rate": 2.657195168500799e-05, "loss": 0.0, "step": 72450 }, { "epoch": 1.41, "learning_rate": 2.6571304724750763e-05, "loss": 0.0108, "step": 72452 }, { "epoch": 1.41, "learning_rate": 2.657065776449353e-05, "loss": 0.0, "step": 72454 }, { "epoch": 1.41, "learning_rate": 2.6570010804236295e-05, "loss": 0.0, "step": 72456 }, { "epoch": 1.41, "learning_rate": 2.6569363843979067e-05, "loss": 0.0, "step": 72458 }, { "epoch": 1.41, "learning_rate": 2.6568716883721833e-05, "loss": 0.0, "step": 72460 }, { "epoch": 1.41, "learning_rate": 2.6568069923464602e-05, "loss": 0.0, "step": 72462 }, { "epoch": 1.41, "learning_rate": 2.656742296320737e-05, "loss": 0.0, "step": 72464 }, { "epoch": 1.41, "learning_rate": 2.656677600295014e-05, "loss": 0.0, "step": 72466 }, { "epoch": 1.41, "learning_rate": 2.6566129042692913e-05, "loss": 0.0, "step": 72468 }, { "epoch": 1.41, "learning_rate": 2.6565482082435678e-05, "loss": 0.0, "step": 72470 }, { "epoch": 1.41, "learning_rate": 2.6564835122178444e-05, "loss": 0.0, "step": 72472 }, { "epoch": 1.41, "learning_rate": 2.6564188161921216e-05, "loss": 0.0001, "step": 72474 }, { "epoch": 1.41, "learning_rate": 2.6563541201663982e-05, "loss": 0.0031, "step": 72476 }, { "epoch": 1.41, "learning_rate": 2.6562894241406748e-05, "loss": 0.0002, "step": 72478 }, { "epoch": 1.41, "learning_rate": 2.656224728114952e-05, "loss": 0.0, "step": 72480 }, { "epoch": 1.41, "learning_rate": 2.656160032089229e-05, "loss": 0.0, "step": 72482 }, { "epoch": 1.41, "learning_rate": 2.6560953360635062e-05, "loss": 0.0, "step": 72484 }, { "epoch": 1.41, "learning_rate": 2.6560306400377827e-05, "loss": 0.0049, "step": 72486 }, { "epoch": 1.41, "learning_rate": 2.6559659440120593e-05, "loss": 0.0006, "step": 72488 }, { "epoch": 1.41, "learning_rate": 2.6559012479863366e-05, "loss": 0.0, "step": 72490 }, { "epoch": 1.41, "learning_rate": 2.655836551960613e-05, "loss": 0.0, "step": 72492 }, { "epoch": 1.41, "learning_rate": 2.6557718559348897e-05, "loss": 0.0, "step": 72494 }, { "epoch": 1.41, "learning_rate": 2.655707159909167e-05, "loss": 0.0, "step": 72496 }, { "epoch": 1.41, "learning_rate": 2.655642463883444e-05, "loss": 0.0009, "step": 72498 }, { "epoch": 1.41, "learning_rate": 2.6555777678577204e-05, "loss": 0.0, "step": 72500 }, { "epoch": 1.41, "learning_rate": 2.6555130718319977e-05, "loss": 0.0028, "step": 72502 }, { "epoch": 1.41, "learning_rate": 2.6554483758062742e-05, "loss": 0.0001, "step": 72504 }, { "epoch": 1.41, "learning_rate": 2.6553836797805515e-05, "loss": 0.0243, "step": 72506 }, { "epoch": 1.41, "learning_rate": 2.655318983754828e-05, "loss": 0.0, "step": 72508 }, { "epoch": 1.41, "learning_rate": 2.6552542877291046e-05, "loss": 0.0, "step": 72510 }, { "epoch": 1.41, "learning_rate": 2.655189591703382e-05, "loss": 0.0, "step": 72512 }, { "epoch": 1.41, "learning_rate": 2.6551248956776588e-05, "loss": 0.0, "step": 72514 }, { "epoch": 1.41, "learning_rate": 2.6550601996519354e-05, "loss": 0.0, "step": 72516 }, { "epoch": 1.41, "learning_rate": 2.6549955036262126e-05, "loss": 0.0, "step": 72518 }, { "epoch": 1.41, "learning_rate": 2.6549308076004892e-05, "loss": 0.0, "step": 72520 }, { "epoch": 1.41, "learning_rate": 2.6548661115747657e-05, "loss": 0.0, "step": 72522 }, { "epoch": 1.41, "learning_rate": 2.654801415549043e-05, "loss": 0.0, "step": 72524 }, { "epoch": 1.41, "learning_rate": 2.6547367195233196e-05, "loss": 0.0, "step": 72526 }, { "epoch": 1.41, "learning_rate": 2.6546720234975968e-05, "loss": 0.0, "step": 72528 }, { "epoch": 1.41, "learning_rate": 2.6546073274718737e-05, "loss": 0.0004, "step": 72530 }, { "epoch": 1.41, "learning_rate": 2.6545426314461503e-05, "loss": 0.0, "step": 72532 }, { "epoch": 1.41, "learning_rate": 2.6544779354204275e-05, "loss": 0.0007, "step": 72534 }, { "epoch": 1.41, "learning_rate": 2.654413239394704e-05, "loss": 0.0017, "step": 72536 }, { "epoch": 1.41, "learning_rate": 2.6543485433689807e-05, "loss": 0.0, "step": 72538 }, { "epoch": 1.41, "learning_rate": 2.654283847343258e-05, "loss": 0.0, "step": 72540 }, { "epoch": 1.41, "learning_rate": 2.6542191513175345e-05, "loss": 0.0002, "step": 72542 }, { "epoch": 1.41, "learning_rate": 2.6541544552918114e-05, "loss": 0.0001, "step": 72544 }, { "epoch": 1.41, "learning_rate": 2.6540897592660886e-05, "loss": 0.0, "step": 72546 }, { "epoch": 1.41, "learning_rate": 2.6540250632403652e-05, "loss": 0.0, "step": 72548 }, { "epoch": 1.41, "learning_rate": 2.6539603672146425e-05, "loss": 0.0, "step": 72550 }, { "epoch": 1.41, "learning_rate": 2.653895671188919e-05, "loss": 0.0, "step": 72552 }, { "epoch": 1.41, "learning_rate": 2.6538309751631956e-05, "loss": 0.0, "step": 72554 }, { "epoch": 1.41, "learning_rate": 2.653766279137473e-05, "loss": 0.0, "step": 72556 }, { "epoch": 1.41, "learning_rate": 2.6537015831117494e-05, "loss": 0.0, "step": 72558 }, { "epoch": 1.41, "learning_rate": 2.6536368870860263e-05, "loss": 0.0, "step": 72560 }, { "epoch": 1.41, "learning_rate": 2.6535721910603036e-05, "loss": 0.0035, "step": 72562 }, { "epoch": 1.41, "learning_rate": 2.65350749503458e-05, "loss": 0.0, "step": 72564 }, { "epoch": 1.41, "learning_rate": 2.6534427990088574e-05, "loss": 0.0, "step": 72566 }, { "epoch": 1.41, "learning_rate": 2.653378102983134e-05, "loss": 0.0, "step": 72568 }, { "epoch": 1.41, "learning_rate": 2.6533134069574105e-05, "loss": 0.0, "step": 72570 }, { "epoch": 1.41, "learning_rate": 2.6532487109316878e-05, "loss": 0.0007, "step": 72572 }, { "epoch": 1.41, "learning_rate": 2.6531840149059643e-05, "loss": 0.0, "step": 72574 }, { "epoch": 1.41, "learning_rate": 2.6531193188802413e-05, "loss": 0.0003, "step": 72576 }, { "epoch": 1.41, "learning_rate": 2.6530546228545185e-05, "loss": 0.0, "step": 72578 }, { "epoch": 1.41, "learning_rate": 2.652989926828795e-05, "loss": 0.0, "step": 72580 }, { "epoch": 1.41, "learning_rate": 2.6529252308030716e-05, "loss": 0.0107, "step": 72582 }, { "epoch": 1.41, "learning_rate": 2.652860534777349e-05, "loss": 0.0001, "step": 72584 }, { "epoch": 1.41, "learning_rate": 2.6527958387516255e-05, "loss": 0.0002, "step": 72586 }, { "epoch": 1.41, "learning_rate": 2.6527311427259027e-05, "loss": 0.0001, "step": 72588 }, { "epoch": 1.41, "learning_rate": 2.6526664467001793e-05, "loss": 0.0, "step": 72590 }, { "epoch": 1.41, "learning_rate": 2.6526017506744562e-05, "loss": 0.0, "step": 72592 }, { "epoch": 1.41, "learning_rate": 2.652537054648733e-05, "loss": 0.0, "step": 72594 }, { "epoch": 1.41, "learning_rate": 2.65247235862301e-05, "loss": 0.0, "step": 72596 }, { "epoch": 1.41, "learning_rate": 2.6524076625972866e-05, "loss": 0.0, "step": 72598 }, { "epoch": 1.41, "learning_rate": 2.6523429665715638e-05, "loss": 0.0, "step": 72600 }, { "epoch": 1.41, "learning_rate": 2.6522782705458404e-05, "loss": 0.0001, "step": 72602 }, { "epoch": 1.41, "learning_rate": 2.652213574520117e-05, "loss": 0.0, "step": 72604 }, { "epoch": 1.41, "learning_rate": 2.6521488784943942e-05, "loss": 0.0, "step": 72606 }, { "epoch": 1.41, "learning_rate": 2.652084182468671e-05, "loss": 0.0, "step": 72608 }, { "epoch": 1.41, "learning_rate": 2.652019486442948e-05, "loss": 0.0, "step": 72610 }, { "epoch": 1.41, "learning_rate": 2.651954790417225e-05, "loss": 0.0, "step": 72612 }, { "epoch": 1.41, "learning_rate": 2.6518900943915015e-05, "loss": 0.0, "step": 72614 }, { "epoch": 1.41, "learning_rate": 2.6518253983657787e-05, "loss": 0.0, "step": 72616 }, { "epoch": 1.41, "learning_rate": 2.6517607023400553e-05, "loss": 0.0001, "step": 72618 }, { "epoch": 1.41, "learning_rate": 2.651696006314332e-05, "loss": 0.0, "step": 72620 }, { "epoch": 1.41, "learning_rate": 2.651631310288609e-05, "loss": 0.0, "step": 72622 }, { "epoch": 1.41, "learning_rate": 2.651566614262886e-05, "loss": 0.0058, "step": 72624 }, { "epoch": 1.41, "learning_rate": 2.6515019182371626e-05, "loss": 0.0001, "step": 72626 }, { "epoch": 1.41, "learning_rate": 2.65143722221144e-05, "loss": 0.0, "step": 72628 }, { "epoch": 1.41, "learning_rate": 2.6513725261857164e-05, "loss": 0.0045, "step": 72630 }, { "epoch": 1.41, "learning_rate": 2.6513078301599937e-05, "loss": 0.002, "step": 72632 }, { "epoch": 1.41, "learning_rate": 2.6512431341342702e-05, "loss": 0.0, "step": 72634 }, { "epoch": 1.41, "learning_rate": 2.6511784381085468e-05, "loss": 0.0, "step": 72636 }, { "epoch": 1.41, "learning_rate": 2.651113742082824e-05, "loss": 0.0002, "step": 72638 }, { "epoch": 1.41, "learning_rate": 2.651049046057101e-05, "loss": 0.0, "step": 72640 }, { "epoch": 1.41, "learning_rate": 2.6509843500313775e-05, "loss": 0.0001, "step": 72642 }, { "epoch": 1.41, "learning_rate": 2.6509196540056548e-05, "loss": 0.0, "step": 72644 }, { "epoch": 1.41, "learning_rate": 2.6508549579799314e-05, "loss": 0.0, "step": 72646 }, { "epoch": 1.41, "learning_rate": 2.6507902619542086e-05, "loss": 0.0004, "step": 72648 }, { "epoch": 1.41, "learning_rate": 2.6507255659284852e-05, "loss": 0.0002, "step": 72650 }, { "epoch": 1.41, "learning_rate": 2.6506608699027617e-05, "loss": 0.0016, "step": 72652 }, { "epoch": 1.41, "learning_rate": 2.650596173877039e-05, "loss": 0.0, "step": 72654 }, { "epoch": 1.41, "learning_rate": 2.6505314778513156e-05, "loss": 0.0003, "step": 72656 }, { "epoch": 1.41, "learning_rate": 2.6504667818255925e-05, "loss": 0.0, "step": 72658 }, { "epoch": 1.41, "learning_rate": 2.6504020857998697e-05, "loss": 0.0001, "step": 72660 }, { "epoch": 1.41, "learning_rate": 2.6503373897741463e-05, "loss": 0.0, "step": 72662 }, { "epoch": 1.41, "learning_rate": 2.650272693748423e-05, "loss": 0.0, "step": 72664 }, { "epoch": 1.41, "learning_rate": 2.6502079977227e-05, "loss": 0.0, "step": 72666 }, { "epoch": 1.41, "learning_rate": 2.6501433016969767e-05, "loss": 0.0127, "step": 72668 }, { "epoch": 1.41, "learning_rate": 2.650078605671254e-05, "loss": 0.0124, "step": 72670 }, { "epoch": 1.41, "learning_rate": 2.6500139096455305e-05, "loss": 0.0, "step": 72672 }, { "epoch": 1.41, "learning_rate": 2.6499492136198074e-05, "loss": 0.0, "step": 72674 }, { "epoch": 1.41, "learning_rate": 2.6498845175940846e-05, "loss": 0.0, "step": 72676 }, { "epoch": 1.41, "learning_rate": 2.6498198215683612e-05, "loss": 0.0, "step": 72678 }, { "epoch": 1.41, "learning_rate": 2.6497551255426378e-05, "loss": 0.0, "step": 72680 }, { "epoch": 1.41, "learning_rate": 2.649690429516915e-05, "loss": 0.0, "step": 72682 }, { "epoch": 1.41, "learning_rate": 2.6496257334911916e-05, "loss": 0.0, "step": 72684 }, { "epoch": 1.41, "learning_rate": 2.6495610374654685e-05, "loss": 0.0105, "step": 72686 }, { "epoch": 1.41, "learning_rate": 2.6494963414397454e-05, "loss": 0.0003, "step": 72688 }, { "epoch": 1.41, "learning_rate": 2.6494316454140223e-05, "loss": 0.0, "step": 72690 }, { "epoch": 1.41, "learning_rate": 2.6493669493882996e-05, "loss": 0.0002, "step": 72692 }, { "epoch": 1.41, "learning_rate": 2.649302253362576e-05, "loss": 0.0091, "step": 72694 }, { "epoch": 1.41, "learning_rate": 2.6492375573368527e-05, "loss": 0.0, "step": 72696 }, { "epoch": 1.41, "learning_rate": 2.64917286131113e-05, "loss": 0.0, "step": 72698 }, { "epoch": 1.41, "learning_rate": 2.6491081652854065e-05, "loss": 0.0, "step": 72700 }, { "epoch": 1.41, "learning_rate": 2.6490434692596834e-05, "loss": 0.0001, "step": 72702 }, { "epoch": 1.41, "learning_rate": 2.6489787732339603e-05, "loss": 0.0, "step": 72704 }, { "epoch": 1.41, "learning_rate": 2.6489140772082373e-05, "loss": 0.0, "step": 72706 }, { "epoch": 1.41, "learning_rate": 2.6488493811825145e-05, "loss": 0.0, "step": 72708 }, { "epoch": 1.41, "learning_rate": 2.648784685156791e-05, "loss": 0.0002, "step": 72710 }, { "epoch": 1.41, "learning_rate": 2.6487199891310676e-05, "loss": 0.0, "step": 72712 }, { "epoch": 1.41, "learning_rate": 2.648655293105345e-05, "loss": 0.0, "step": 72714 }, { "epoch": 1.41, "learning_rate": 2.6485905970796215e-05, "loss": 0.0, "step": 72716 }, { "epoch": 1.41, "learning_rate": 2.648525901053898e-05, "loss": 0.0, "step": 72718 }, { "epoch": 1.41, "learning_rate": 2.6484612050281753e-05, "loss": 0.0, "step": 72720 }, { "epoch": 1.41, "learning_rate": 2.6483965090024522e-05, "loss": 0.0, "step": 72722 }, { "epoch": 1.41, "learning_rate": 2.6483318129767288e-05, "loss": 0.0, "step": 72724 }, { "epoch": 1.41, "learning_rate": 2.648267116951006e-05, "loss": 0.0, "step": 72726 }, { "epoch": 1.41, "learning_rate": 2.6482024209252826e-05, "loss": 0.0, "step": 72728 }, { "epoch": 1.41, "learning_rate": 2.6481377248995598e-05, "loss": 0.0, "step": 72730 }, { "epoch": 1.41, "learning_rate": 2.6480730288738364e-05, "loss": 0.0, "step": 72732 }, { "epoch": 1.41, "learning_rate": 2.648008332848113e-05, "loss": 0.0, "step": 72734 }, { "epoch": 1.41, "learning_rate": 2.6479436368223902e-05, "loss": 0.0, "step": 72736 }, { "epoch": 1.41, "learning_rate": 2.647878940796667e-05, "loss": 0.0, "step": 72738 }, { "epoch": 1.41, "learning_rate": 2.6478142447709437e-05, "loss": 0.0, "step": 72740 }, { "epoch": 1.41, "learning_rate": 2.647749548745221e-05, "loss": 0.0, "step": 72742 }, { "epoch": 1.41, "learning_rate": 2.6476848527194975e-05, "loss": 0.0001, "step": 72744 }, { "epoch": 1.41, "learning_rate": 2.647620156693774e-05, "loss": 0.0, "step": 72746 }, { "epoch": 1.41, "learning_rate": 2.6475554606680513e-05, "loss": 0.0, "step": 72748 }, { "epoch": 1.41, "learning_rate": 2.647490764642328e-05, "loss": 0.0, "step": 72750 }, { "epoch": 1.41, "learning_rate": 2.647426068616605e-05, "loss": 0.0, "step": 72752 }, { "epoch": 1.41, "learning_rate": 2.647361372590882e-05, "loss": 0.0, "step": 72754 }, { "epoch": 1.41, "learning_rate": 2.6472966765651586e-05, "loss": 0.0, "step": 72756 }, { "epoch": 1.41, "learning_rate": 2.647231980539436e-05, "loss": 0.0, "step": 72758 }, { "epoch": 1.41, "learning_rate": 2.6471672845137124e-05, "loss": 0.0, "step": 72760 }, { "epoch": 1.41, "learning_rate": 2.647102588487989e-05, "loss": 0.0, "step": 72762 }, { "epoch": 1.41, "learning_rate": 2.6470378924622662e-05, "loss": 0.0091, "step": 72764 }, { "epoch": 1.41, "learning_rate": 2.6469731964365428e-05, "loss": 0.0, "step": 72766 }, { "epoch": 1.41, "learning_rate": 2.6469085004108197e-05, "loss": 0.0, "step": 72768 }, { "epoch": 1.41, "learning_rate": 2.646843804385097e-05, "loss": 0.0, "step": 72770 }, { "epoch": 1.41, "learning_rate": 2.6467791083593735e-05, "loss": 0.0, "step": 72772 }, { "epoch": 1.41, "learning_rate": 2.6467144123336508e-05, "loss": 0.0, "step": 72774 }, { "epoch": 1.41, "learning_rate": 2.6466497163079274e-05, "loss": 0.0, "step": 72776 }, { "epoch": 1.41, "learning_rate": 2.646585020282204e-05, "loss": 0.0006, "step": 72778 }, { "epoch": 1.41, "learning_rate": 2.6465203242564812e-05, "loss": 0.0, "step": 72780 }, { "epoch": 1.41, "learning_rate": 2.6464556282307577e-05, "loss": 0.0, "step": 72782 }, { "epoch": 1.41, "learning_rate": 2.6463909322050347e-05, "loss": 0.0, "step": 72784 }, { "epoch": 1.41, "learning_rate": 2.646326236179312e-05, "loss": 0.0019, "step": 72786 }, { "epoch": 1.41, "learning_rate": 2.6462615401535885e-05, "loss": 0.0, "step": 72788 }, { "epoch": 1.41, "learning_rate": 2.6461968441278657e-05, "loss": 0.0, "step": 72790 }, { "epoch": 1.41, "learning_rate": 2.6461321481021423e-05, "loss": 0.0, "step": 72792 }, { "epoch": 1.41, "learning_rate": 2.646067452076419e-05, "loss": 0.0, "step": 72794 }, { "epoch": 1.41, "learning_rate": 2.646002756050696e-05, "loss": 0.0, "step": 72796 }, { "epoch": 1.41, "learning_rate": 2.6459380600249727e-05, "loss": 0.0001, "step": 72798 }, { "epoch": 1.41, "learning_rate": 2.6458733639992496e-05, "loss": 0.0058, "step": 72800 }, { "epoch": 1.41, "learning_rate": 2.6458086679735268e-05, "loss": 0.0052, "step": 72802 }, { "epoch": 1.41, "learning_rate": 2.6457439719478034e-05, "loss": 0.0006, "step": 72804 }, { "epoch": 1.41, "learning_rate": 2.64567927592208e-05, "loss": 0.0008, "step": 72806 }, { "epoch": 1.41, "learning_rate": 2.6456145798963572e-05, "loss": 0.0, "step": 72808 }, { "epoch": 1.41, "learning_rate": 2.6455498838706338e-05, "loss": 0.0, "step": 72810 }, { "epoch": 1.41, "learning_rate": 2.645485187844911e-05, "loss": 0.0, "step": 72812 }, { "epoch": 1.41, "learning_rate": 2.6454204918191876e-05, "loss": 0.0, "step": 72814 }, { "epoch": 1.41, "learning_rate": 2.6453557957934645e-05, "loss": 0.0006, "step": 72816 }, { "epoch": 1.41, "learning_rate": 2.6452910997677418e-05, "loss": 0.0, "step": 72818 }, { "epoch": 1.41, "learning_rate": 2.6452264037420183e-05, "loss": 0.0, "step": 72820 }, { "epoch": 1.41, "learning_rate": 2.645161707716295e-05, "loss": 0.0, "step": 72822 }, { "epoch": 1.41, "learning_rate": 2.645097011690572e-05, "loss": 0.0, "step": 72824 }, { "epoch": 1.41, "learning_rate": 2.6450323156648487e-05, "loss": 0.0044, "step": 72826 }, { "epoch": 1.41, "learning_rate": 2.6449676196391253e-05, "loss": 0.0, "step": 72828 }, { "epoch": 1.41, "learning_rate": 2.6449029236134025e-05, "loss": 0.0, "step": 72830 }, { "epoch": 1.41, "learning_rate": 2.6448382275876794e-05, "loss": 0.0, "step": 72832 }, { "epoch": 1.41, "learning_rate": 2.6447735315619563e-05, "loss": 0.0, "step": 72834 }, { "epoch": 1.41, "learning_rate": 2.6447088355362333e-05, "loss": 0.0001, "step": 72836 }, { "epoch": 1.41, "learning_rate": 2.6446441395105098e-05, "loss": 0.0, "step": 72838 }, { "epoch": 1.41, "learning_rate": 2.644579443484787e-05, "loss": 0.0124, "step": 72840 }, { "epoch": 1.41, "learning_rate": 2.6445147474590636e-05, "loss": 0.0, "step": 72842 }, { "epoch": 1.41, "learning_rate": 2.6444500514333402e-05, "loss": 0.0002, "step": 72844 }, { "epoch": 1.41, "learning_rate": 2.6443853554076175e-05, "loss": 0.0, "step": 72846 }, { "epoch": 1.41, "learning_rate": 2.6443206593818944e-05, "loss": 0.0, "step": 72848 }, { "epoch": 1.41, "learning_rate": 2.6442559633561713e-05, "loss": 0.0, "step": 72850 }, { "epoch": 1.41, "learning_rate": 2.6441912673304482e-05, "loss": 0.0, "step": 72852 }, { "epoch": 1.41, "learning_rate": 2.6441265713047248e-05, "loss": 0.0007, "step": 72854 }, { "epoch": 1.41, "learning_rate": 2.644061875279002e-05, "loss": 0.0001, "step": 72856 }, { "epoch": 1.41, "learning_rate": 2.6439971792532786e-05, "loss": 0.0, "step": 72858 }, { "epoch": 1.41, "learning_rate": 2.643932483227555e-05, "loss": 0.0001, "step": 72860 }, { "epoch": 1.41, "learning_rate": 2.6438677872018324e-05, "loss": 0.0, "step": 72862 }, { "epoch": 1.41, "learning_rate": 2.6438030911761093e-05, "loss": 0.0, "step": 72864 }, { "epoch": 1.41, "learning_rate": 2.643738395150386e-05, "loss": 0.0, "step": 72866 }, { "epoch": 1.41, "learning_rate": 2.643673699124663e-05, "loss": 0.0, "step": 72868 }, { "epoch": 1.41, "learning_rate": 2.6436090030989397e-05, "loss": 0.1006, "step": 72870 }, { "epoch": 1.41, "learning_rate": 2.643544307073217e-05, "loss": 0.0, "step": 72872 }, { "epoch": 1.41, "learning_rate": 2.6434796110474935e-05, "loss": 0.0, "step": 72874 }, { "epoch": 1.41, "learning_rate": 2.64341491502177e-05, "loss": 0.0001, "step": 72876 }, { "epoch": 1.41, "learning_rate": 2.6433502189960473e-05, "loss": 0.0004, "step": 72878 }, { "epoch": 1.41, "learning_rate": 2.6432855229703242e-05, "loss": 0.0001, "step": 72880 }, { "epoch": 1.41, "learning_rate": 2.6432208269446008e-05, "loss": 0.0017, "step": 72882 }, { "epoch": 1.41, "learning_rate": 2.643156130918878e-05, "loss": 0.0, "step": 72884 }, { "epoch": 1.41, "learning_rate": 2.6430914348931546e-05, "loss": 0.0, "step": 72886 }, { "epoch": 1.41, "learning_rate": 2.6430267388674312e-05, "loss": 0.0, "step": 72888 }, { "epoch": 1.41, "learning_rate": 2.6429620428417084e-05, "loss": 0.0002, "step": 72890 }, { "epoch": 1.41, "learning_rate": 2.642897346815985e-05, "loss": 0.0, "step": 72892 }, { "epoch": 1.41, "learning_rate": 2.6428326507902622e-05, "loss": 0.0, "step": 72894 }, { "epoch": 1.41, "learning_rate": 2.6427679547645388e-05, "loss": 0.0, "step": 72896 }, { "epoch": 1.41, "learning_rate": 2.6427032587388157e-05, "loss": 0.0266, "step": 72898 }, { "epoch": 1.41, "learning_rate": 2.642638562713093e-05, "loss": 0.0, "step": 72900 }, { "epoch": 1.41, "learning_rate": 2.6425738666873695e-05, "loss": 0.0, "step": 72902 }, { "epoch": 1.41, "learning_rate": 2.642509170661646e-05, "loss": 0.0, "step": 72904 }, { "epoch": 1.42, "learning_rate": 2.6424444746359234e-05, "loss": 0.0059, "step": 72906 }, { "epoch": 1.42, "learning_rate": 2.6423797786102e-05, "loss": 0.0, "step": 72908 }, { "epoch": 1.42, "learning_rate": 2.642315082584477e-05, "loss": 0.0, "step": 72910 }, { "epoch": 1.42, "learning_rate": 2.6422503865587537e-05, "loss": 0.0007, "step": 72912 }, { "epoch": 1.42, "learning_rate": 2.6421856905330307e-05, "loss": 0.0017, "step": 72914 }, { "epoch": 1.42, "learning_rate": 2.642120994507308e-05, "loss": 0.0127, "step": 72916 }, { "epoch": 1.42, "learning_rate": 2.6420562984815845e-05, "loss": 0.0, "step": 72918 }, { "epoch": 1.42, "learning_rate": 2.641991602455861e-05, "loss": 0.0, "step": 72920 }, { "epoch": 1.42, "learning_rate": 2.6419269064301383e-05, "loss": 0.0, "step": 72922 }, { "epoch": 1.42, "learning_rate": 2.641862210404415e-05, "loss": 0.0, "step": 72924 }, { "epoch": 1.42, "learning_rate": 2.6417975143786918e-05, "loss": 0.0045, "step": 72926 }, { "epoch": 1.42, "learning_rate": 2.6417328183529687e-05, "loss": 0.0, "step": 72928 }, { "epoch": 1.42, "learning_rate": 2.6416681223272456e-05, "loss": 0.0, "step": 72930 }, { "epoch": 1.42, "learning_rate": 2.6416034263015228e-05, "loss": 0.0, "step": 72932 }, { "epoch": 1.42, "learning_rate": 2.6415387302757994e-05, "loss": 0.0, "step": 72934 }, { "epoch": 1.42, "learning_rate": 2.641474034250076e-05, "loss": 0.0, "step": 72936 }, { "epoch": 1.42, "learning_rate": 2.6414093382243532e-05, "loss": 0.0, "step": 72938 }, { "epoch": 1.42, "learning_rate": 2.6413446421986298e-05, "loss": 0.0, "step": 72940 }, { "epoch": 1.42, "learning_rate": 2.6412799461729067e-05, "loss": 0.014, "step": 72942 }, { "epoch": 1.42, "learning_rate": 2.6412152501471836e-05, "loss": 0.0, "step": 72944 }, { "epoch": 1.42, "learning_rate": 2.6411505541214605e-05, "loss": 0.002, "step": 72946 }, { "epoch": 1.42, "learning_rate": 2.641085858095737e-05, "loss": 0.0004, "step": 72948 }, { "epoch": 1.42, "learning_rate": 2.6410211620700143e-05, "loss": 0.0, "step": 72950 }, { "epoch": 1.42, "learning_rate": 2.640956466044291e-05, "loss": 0.0, "step": 72952 }, { "epoch": 1.42, "learning_rate": 2.640891770018568e-05, "loss": 0.0024, "step": 72954 }, { "epoch": 1.42, "learning_rate": 2.6408270739928447e-05, "loss": 0.0, "step": 72956 }, { "epoch": 1.42, "learning_rate": 2.6407623779671213e-05, "loss": 0.0, "step": 72958 }, { "epoch": 1.42, "learning_rate": 2.6406976819413985e-05, "loss": 0.0, "step": 72960 }, { "epoch": 1.42, "learning_rate": 2.6406329859156754e-05, "loss": 0.0, "step": 72962 }, { "epoch": 1.42, "learning_rate": 2.640568289889952e-05, "loss": 0.0052, "step": 72964 }, { "epoch": 1.42, "learning_rate": 2.6405035938642293e-05, "loss": 0.0, "step": 72966 }, { "epoch": 1.42, "learning_rate": 2.6404388978385058e-05, "loss": 0.0001, "step": 72968 }, { "epoch": 1.42, "learning_rate": 2.6403742018127824e-05, "loss": 0.0, "step": 72970 }, { "epoch": 1.42, "learning_rate": 2.6403095057870596e-05, "loss": 0.0117, "step": 72972 }, { "epoch": 1.42, "learning_rate": 2.6402448097613362e-05, "loss": 0.0, "step": 72974 }, { "epoch": 1.42, "learning_rate": 2.6401801137356135e-05, "loss": 0.0001, "step": 72976 }, { "epoch": 1.42, "learning_rate": 2.6401154177098904e-05, "loss": 0.0, "step": 72978 }, { "epoch": 1.42, "learning_rate": 2.640050721684167e-05, "loss": 0.0, "step": 72980 }, { "epoch": 1.42, "learning_rate": 2.6399860256584442e-05, "loss": 0.0185, "step": 72982 }, { "epoch": 1.42, "learning_rate": 2.6399213296327208e-05, "loss": 0.0, "step": 72984 }, { "epoch": 1.42, "learning_rate": 2.6398566336069973e-05, "loss": 0.0, "step": 72986 }, { "epoch": 1.42, "learning_rate": 2.6397919375812746e-05, "loss": 0.0, "step": 72988 }, { "epoch": 1.42, "learning_rate": 2.639727241555551e-05, "loss": 0.0, "step": 72990 }, { "epoch": 1.42, "learning_rate": 2.6396625455298284e-05, "loss": 0.0, "step": 72992 }, { "epoch": 1.42, "learning_rate": 2.6395978495041053e-05, "loss": 0.0001, "step": 72994 }, { "epoch": 1.42, "learning_rate": 2.639533153478382e-05, "loss": 0.0077, "step": 72996 }, { "epoch": 1.42, "learning_rate": 2.639468457452659e-05, "loss": 0.0, "step": 72998 }, { "epoch": 1.42, "learning_rate": 2.6394037614269357e-05, "loss": 0.0001, "step": 73000 }, { "epoch": 1.42, "learning_rate": 2.6393390654012123e-05, "loss": 0.0, "step": 73002 }, { "epoch": 1.42, "learning_rate": 2.6392743693754895e-05, "loss": 0.0, "step": 73004 }, { "epoch": 1.42, "learning_rate": 2.639209673349766e-05, "loss": 0.0, "step": 73006 }, { "epoch": 1.42, "learning_rate": 2.639144977324043e-05, "loss": 0.0, "step": 73008 }, { "epoch": 1.42, "learning_rate": 2.6390802812983202e-05, "loss": 0.0117, "step": 73010 }, { "epoch": 1.42, "learning_rate": 2.6390155852725968e-05, "loss": 0.01, "step": 73012 }, { "epoch": 1.42, "learning_rate": 2.638950889246874e-05, "loss": 0.0, "step": 73014 }, { "epoch": 1.42, "learning_rate": 2.6388861932211506e-05, "loss": 0.0, "step": 73016 }, { "epoch": 1.42, "learning_rate": 2.6388214971954272e-05, "loss": 0.0, "step": 73018 }, { "epoch": 1.42, "learning_rate": 2.6387568011697044e-05, "loss": 0.0, "step": 73020 }, { "epoch": 1.42, "learning_rate": 2.638692105143981e-05, "loss": 0.0, "step": 73022 }, { "epoch": 1.42, "learning_rate": 2.638627409118258e-05, "loss": 0.0, "step": 73024 }, { "epoch": 1.42, "learning_rate": 2.638562713092535e-05, "loss": 0.0, "step": 73026 }, { "epoch": 1.42, "learning_rate": 2.6384980170668117e-05, "loss": 0.0, "step": 73028 }, { "epoch": 1.42, "learning_rate": 2.6384333210410883e-05, "loss": 0.0, "step": 73030 }, { "epoch": 1.42, "learning_rate": 2.6383686250153655e-05, "loss": 0.0, "step": 73032 }, { "epoch": 1.42, "learning_rate": 2.638303928989642e-05, "loss": 0.0014, "step": 73034 }, { "epoch": 1.42, "learning_rate": 2.6382392329639194e-05, "loss": 0.0001, "step": 73036 }, { "epoch": 1.42, "learning_rate": 2.638174536938196e-05, "loss": 0.0, "step": 73038 }, { "epoch": 1.42, "learning_rate": 2.638109840912473e-05, "loss": 0.0, "step": 73040 }, { "epoch": 1.42, "learning_rate": 2.63804514488675e-05, "loss": 0.0, "step": 73042 }, { "epoch": 1.42, "learning_rate": 2.6379804488610267e-05, "loss": 0.0, "step": 73044 }, { "epoch": 1.42, "learning_rate": 2.6379157528353032e-05, "loss": 0.0078, "step": 73046 }, { "epoch": 1.42, "learning_rate": 2.6378510568095805e-05, "loss": 0.0, "step": 73048 }, { "epoch": 1.42, "learning_rate": 2.637786360783857e-05, "loss": 0.0, "step": 73050 }, { "epoch": 1.42, "learning_rate": 2.6377216647581336e-05, "loss": 0.0, "step": 73052 }, { "epoch": 1.42, "learning_rate": 2.637656968732411e-05, "loss": 0.0, "step": 73054 }, { "epoch": 1.42, "learning_rate": 2.6375922727066878e-05, "loss": 0.0009, "step": 73056 }, { "epoch": 1.42, "learning_rate": 2.637527576680965e-05, "loss": 0.0, "step": 73058 }, { "epoch": 1.42, "learning_rate": 2.6374628806552416e-05, "loss": 0.0, "step": 73060 }, { "epoch": 1.42, "learning_rate": 2.637398184629518e-05, "loss": 0.0, "step": 73062 }, { "epoch": 1.42, "learning_rate": 2.6373334886037954e-05, "loss": 0.0, "step": 73064 }, { "epoch": 1.42, "learning_rate": 2.637268792578072e-05, "loss": 0.0007, "step": 73066 }, { "epoch": 1.42, "learning_rate": 2.6372040965523485e-05, "loss": 0.0, "step": 73068 }, { "epoch": 1.42, "learning_rate": 2.6371394005266258e-05, "loss": 0.0, "step": 73070 }, { "epoch": 1.42, "learning_rate": 2.6370747045009027e-05, "loss": 0.0, "step": 73072 }, { "epoch": 1.42, "learning_rate": 2.6370100084751796e-05, "loss": 0.0, "step": 73074 }, { "epoch": 1.42, "learning_rate": 2.6369453124494565e-05, "loss": 0.0, "step": 73076 }, { "epoch": 1.42, "learning_rate": 2.636880616423733e-05, "loss": 0.0, "step": 73078 }, { "epoch": 1.42, "learning_rate": 2.6368159203980103e-05, "loss": 0.0, "step": 73080 }, { "epoch": 1.42, "learning_rate": 2.636751224372287e-05, "loss": 0.0051, "step": 73082 }, { "epoch": 1.42, "learning_rate": 2.6366865283465635e-05, "loss": 0.0, "step": 73084 }, { "epoch": 1.42, "learning_rate": 2.6366218323208407e-05, "loss": 0.0, "step": 73086 }, { "epoch": 1.42, "learning_rate": 2.6365571362951176e-05, "loss": 0.0, "step": 73088 }, { "epoch": 1.42, "learning_rate": 2.6364924402693942e-05, "loss": 0.0, "step": 73090 }, { "epoch": 1.42, "learning_rate": 2.6364277442436714e-05, "loss": 0.0, "step": 73092 }, { "epoch": 1.42, "learning_rate": 2.636363048217948e-05, "loss": 0.0027, "step": 73094 }, { "epoch": 1.42, "learning_rate": 2.6362983521922253e-05, "loss": 0.0017, "step": 73096 }, { "epoch": 1.42, "learning_rate": 2.6362336561665018e-05, "loss": 0.0, "step": 73098 }, { "epoch": 1.42, "learning_rate": 2.6361689601407784e-05, "loss": 0.0, "step": 73100 }, { "epoch": 1.42, "learning_rate": 2.6361042641150556e-05, "loss": 0.0, "step": 73102 }, { "epoch": 1.42, "learning_rate": 2.6360395680893325e-05, "loss": 0.0, "step": 73104 }, { "epoch": 1.42, "learning_rate": 2.635974872063609e-05, "loss": 0.0, "step": 73106 }, { "epoch": 1.42, "learning_rate": 2.6359101760378864e-05, "loss": 0.0, "step": 73108 }, { "epoch": 1.42, "learning_rate": 2.635845480012163e-05, "loss": 0.0, "step": 73110 }, { "epoch": 1.42, "learning_rate": 2.6357807839864395e-05, "loss": 0.0, "step": 73112 }, { "epoch": 1.42, "learning_rate": 2.6357160879607168e-05, "loss": 0.0, "step": 73114 }, { "epoch": 1.42, "learning_rate": 2.6356513919349933e-05, "loss": 0.0, "step": 73116 }, { "epoch": 1.42, "learning_rate": 2.6355866959092706e-05, "loss": 0.0, "step": 73118 }, { "epoch": 1.42, "learning_rate": 2.6355219998835475e-05, "loss": 0.0, "step": 73120 }, { "epoch": 1.42, "learning_rate": 2.635457303857824e-05, "loss": 0.0022, "step": 73122 }, { "epoch": 1.42, "learning_rate": 2.6353926078321013e-05, "loss": 0.0017, "step": 73124 }, { "epoch": 1.42, "learning_rate": 2.635327911806378e-05, "loss": 0.0001, "step": 73126 }, { "epoch": 1.42, "learning_rate": 2.6352632157806544e-05, "loss": 0.0001, "step": 73128 }, { "epoch": 1.42, "learning_rate": 2.6351985197549317e-05, "loss": 0.0, "step": 73130 }, { "epoch": 1.42, "learning_rate": 2.6351338237292082e-05, "loss": 0.0001, "step": 73132 }, { "epoch": 1.42, "learning_rate": 2.6350691277034855e-05, "loss": 0.0087, "step": 73134 }, { "epoch": 1.42, "learning_rate": 2.635004431677762e-05, "loss": 0.0, "step": 73136 }, { "epoch": 1.42, "learning_rate": 2.634939735652039e-05, "loss": 0.0, "step": 73138 }, { "epoch": 1.42, "learning_rate": 2.6348750396263162e-05, "loss": 0.0, "step": 73140 }, { "epoch": 1.42, "learning_rate": 2.6348103436005928e-05, "loss": 0.0, "step": 73142 }, { "epoch": 1.42, "learning_rate": 2.6347456475748694e-05, "loss": 0.0, "step": 73144 }, { "epoch": 1.42, "learning_rate": 2.6346809515491466e-05, "loss": 0.0, "step": 73146 }, { "epoch": 1.42, "learning_rate": 2.6346162555234232e-05, "loss": 0.0004, "step": 73148 }, { "epoch": 1.42, "learning_rate": 2.6345515594977e-05, "loss": 0.0001, "step": 73150 }, { "epoch": 1.42, "learning_rate": 2.634486863471977e-05, "loss": 0.0, "step": 73152 }, { "epoch": 1.42, "learning_rate": 2.634422167446254e-05, "loss": 0.0, "step": 73154 }, { "epoch": 1.42, "learning_rate": 2.634357471420531e-05, "loss": 0.0, "step": 73156 }, { "epoch": 1.42, "learning_rate": 2.6342927753948077e-05, "loss": 0.0, "step": 73158 }, { "epoch": 1.42, "learning_rate": 2.6342280793690843e-05, "loss": 0.0, "step": 73160 }, { "epoch": 1.42, "learning_rate": 2.6341633833433615e-05, "loss": 0.0, "step": 73162 }, { "epoch": 1.42, "learning_rate": 2.634098687317638e-05, "loss": 0.0, "step": 73164 }, { "epoch": 1.42, "learning_rate": 2.634033991291915e-05, "loss": 0.0, "step": 73166 }, { "epoch": 1.42, "learning_rate": 2.633969295266192e-05, "loss": 0.0001, "step": 73168 }, { "epoch": 1.42, "learning_rate": 2.633904599240469e-05, "loss": 0.0, "step": 73170 }, { "epoch": 1.42, "learning_rate": 2.6338399032147454e-05, "loss": 0.0, "step": 73172 }, { "epoch": 1.42, "learning_rate": 2.6337752071890226e-05, "loss": 0.0016, "step": 73174 }, { "epoch": 1.42, "learning_rate": 2.6337105111632992e-05, "loss": 0.0, "step": 73176 }, { "epoch": 1.42, "learning_rate": 2.6336458151375765e-05, "loss": 0.0, "step": 73178 }, { "epoch": 1.42, "learning_rate": 2.633581119111853e-05, "loss": 0.0001, "step": 73180 }, { "epoch": 1.42, "learning_rate": 2.63351642308613e-05, "loss": 0.0264, "step": 73182 }, { "epoch": 1.42, "learning_rate": 2.633451727060407e-05, "loss": 0.0, "step": 73184 }, { "epoch": 1.42, "learning_rate": 2.6333870310346838e-05, "loss": 0.0001, "step": 73186 }, { "epoch": 1.42, "learning_rate": 2.6333223350089603e-05, "loss": 0.0002, "step": 73188 }, { "epoch": 1.42, "learning_rate": 2.6332576389832376e-05, "loss": 0.0, "step": 73190 }, { "epoch": 1.42, "learning_rate": 2.633192942957514e-05, "loss": 0.0, "step": 73192 }, { "epoch": 1.42, "learning_rate": 2.6331282469317907e-05, "loss": 0.0, "step": 73194 }, { "epoch": 1.42, "learning_rate": 2.633063550906068e-05, "loss": 0.0, "step": 73196 }, { "epoch": 1.42, "learning_rate": 2.6329988548803445e-05, "loss": 0.0, "step": 73198 }, { "epoch": 1.42, "learning_rate": 2.6329341588546218e-05, "loss": 0.0, "step": 73200 }, { "epoch": 1.42, "learning_rate": 2.6328694628288987e-05, "loss": 0.0021, "step": 73202 }, { "epoch": 1.42, "learning_rate": 2.6328047668031753e-05, "loss": 0.0128, "step": 73204 }, { "epoch": 1.42, "learning_rate": 2.6327400707774525e-05, "loss": 0.0, "step": 73206 }, { "epoch": 1.42, "learning_rate": 2.632675374751729e-05, "loss": 0.0, "step": 73208 }, { "epoch": 1.42, "learning_rate": 2.6326106787260056e-05, "loss": 0.0, "step": 73210 }, { "epoch": 1.42, "learning_rate": 2.632545982700283e-05, "loss": 0.0, "step": 73212 }, { "epoch": 1.42, "learning_rate": 2.6324812866745595e-05, "loss": 0.0, "step": 73214 }, { "epoch": 1.42, "learning_rate": 2.6324165906488367e-05, "loss": 0.0, "step": 73216 }, { "epoch": 1.42, "learning_rate": 2.6323518946231136e-05, "loss": 0.0, "step": 73218 }, { "epoch": 1.42, "learning_rate": 2.6322871985973902e-05, "loss": 0.0, "step": 73220 }, { "epoch": 1.42, "learning_rate": 2.6322225025716674e-05, "loss": 0.0005, "step": 73222 }, { "epoch": 1.42, "learning_rate": 2.632157806545944e-05, "loss": 0.0, "step": 73224 }, { "epoch": 1.42, "learning_rate": 2.6320931105202206e-05, "loss": 0.0, "step": 73226 }, { "epoch": 1.42, "learning_rate": 2.6320284144944978e-05, "loss": 0.0024, "step": 73228 }, { "epoch": 1.42, "learning_rate": 2.6319637184687744e-05, "loss": 0.0, "step": 73230 }, { "epoch": 1.42, "learning_rate": 2.6318990224430513e-05, "loss": 0.0, "step": 73232 }, { "epoch": 1.42, "learning_rate": 2.6318343264173285e-05, "loss": 0.0, "step": 73234 }, { "epoch": 1.42, "learning_rate": 2.631769630391605e-05, "loss": 0.0, "step": 73236 }, { "epoch": 1.42, "learning_rate": 2.6317049343658824e-05, "loss": 0.0, "step": 73238 }, { "epoch": 1.42, "learning_rate": 2.631640238340159e-05, "loss": 0.0, "step": 73240 }, { "epoch": 1.42, "learning_rate": 2.6315755423144355e-05, "loss": 0.0, "step": 73242 }, { "epoch": 1.42, "learning_rate": 2.6315108462887128e-05, "loss": 0.0, "step": 73244 }, { "epoch": 1.42, "learning_rate": 2.6314461502629893e-05, "loss": 0.0009, "step": 73246 }, { "epoch": 1.42, "learning_rate": 2.6313814542372662e-05, "loss": 0.0, "step": 73248 }, { "epoch": 1.42, "learning_rate": 2.6313167582115435e-05, "loss": 0.0, "step": 73250 }, { "epoch": 1.42, "learning_rate": 2.63125206218582e-05, "loss": 0.0, "step": 73252 }, { "epoch": 1.42, "learning_rate": 2.6311873661600966e-05, "loss": 0.0, "step": 73254 }, { "epoch": 1.42, "learning_rate": 2.631122670134374e-05, "loss": 0.0, "step": 73256 }, { "epoch": 1.42, "learning_rate": 2.6310579741086504e-05, "loss": 0.0, "step": 73258 }, { "epoch": 1.42, "learning_rate": 2.6309932780829277e-05, "loss": 0.0, "step": 73260 }, { "epoch": 1.42, "learning_rate": 2.6309285820572042e-05, "loss": 0.0, "step": 73262 }, { "epoch": 1.42, "learning_rate": 2.630863886031481e-05, "loss": 0.0, "step": 73264 }, { "epoch": 1.42, "learning_rate": 2.6307991900057584e-05, "loss": 0.0, "step": 73266 }, { "epoch": 1.42, "learning_rate": 2.630734493980035e-05, "loss": 0.0, "step": 73268 }, { "epoch": 1.42, "learning_rate": 2.6306697979543115e-05, "loss": 0.0, "step": 73270 }, { "epoch": 1.42, "learning_rate": 2.6306051019285888e-05, "loss": 0.0, "step": 73272 }, { "epoch": 1.42, "learning_rate": 2.6305404059028654e-05, "loss": 0.0, "step": 73274 }, { "epoch": 1.42, "learning_rate": 2.6304757098771426e-05, "loss": 0.0, "step": 73276 }, { "epoch": 1.42, "learning_rate": 2.6304110138514192e-05, "loss": 0.0, "step": 73278 }, { "epoch": 1.42, "learning_rate": 2.630346317825696e-05, "loss": 0.0, "step": 73280 }, { "epoch": 1.42, "learning_rate": 2.6302816217999733e-05, "loss": 0.0, "step": 73282 }, { "epoch": 1.42, "learning_rate": 2.63021692577425e-05, "loss": 0.0, "step": 73284 }, { "epoch": 1.42, "learning_rate": 2.6301522297485265e-05, "loss": 0.0002, "step": 73286 }, { "epoch": 1.42, "learning_rate": 2.6300875337228037e-05, "loss": 0.0, "step": 73288 }, { "epoch": 1.42, "learning_rate": 2.6300228376970803e-05, "loss": 0.0, "step": 73290 }, { "epoch": 1.42, "learning_rate": 2.629958141671357e-05, "loss": 0.0007, "step": 73292 }, { "epoch": 1.42, "learning_rate": 2.629893445645634e-05, "loss": 0.0002, "step": 73294 }, { "epoch": 1.42, "learning_rate": 2.629828749619911e-05, "loss": 0.0, "step": 73296 }, { "epoch": 1.42, "learning_rate": 2.6297640535941883e-05, "loss": 0.0, "step": 73298 }, { "epoch": 1.42, "learning_rate": 2.629699357568465e-05, "loss": 0.0, "step": 73300 }, { "epoch": 1.42, "learning_rate": 2.6296346615427414e-05, "loss": 0.0084, "step": 73302 }, { "epoch": 1.42, "learning_rate": 2.6295699655170186e-05, "loss": 0.0, "step": 73304 }, { "epoch": 1.42, "learning_rate": 2.6295052694912952e-05, "loss": 0.0, "step": 73306 }, { "epoch": 1.42, "learning_rate": 2.6294405734655718e-05, "loss": 0.0, "step": 73308 }, { "epoch": 1.42, "learning_rate": 2.629375877439849e-05, "loss": 0.0, "step": 73310 }, { "epoch": 1.42, "learning_rate": 2.629311181414126e-05, "loss": 0.0, "step": 73312 }, { "epoch": 1.42, "learning_rate": 2.6292464853884025e-05, "loss": 0.0, "step": 73314 }, { "epoch": 1.42, "learning_rate": 2.6291817893626798e-05, "loss": 0.0, "step": 73316 }, { "epoch": 1.42, "learning_rate": 2.629149441349818e-05, "loss": 0.1316, "step": 73318 }, { "epoch": 1.42, "learning_rate": 2.6290847453240948e-05, "loss": 0.0, "step": 73320 }, { "epoch": 1.42, "learning_rate": 2.629020049298372e-05, "loss": 0.0, "step": 73322 }, { "epoch": 1.42, "learning_rate": 2.6289553532726486e-05, "loss": 0.0, "step": 73324 }, { "epoch": 1.42, "learning_rate": 2.6288906572469252e-05, "loss": 0.0095, "step": 73326 }, { "epoch": 1.42, "learning_rate": 2.6288259612212024e-05, "loss": 0.0, "step": 73328 }, { "epoch": 1.42, "learning_rate": 2.628761265195479e-05, "loss": 0.0, "step": 73330 }, { "epoch": 1.42, "learning_rate": 2.6286965691697562e-05, "loss": 0.0, "step": 73332 }, { "epoch": 1.42, "learning_rate": 2.6286318731440328e-05, "loss": 0.0001, "step": 73334 }, { "epoch": 1.42, "learning_rate": 2.6285671771183097e-05, "loss": 0.009, "step": 73336 }, { "epoch": 1.42, "learning_rate": 2.6285024810925866e-05, "loss": 0.0, "step": 73338 }, { "epoch": 1.42, "learning_rate": 2.6284377850668635e-05, "loss": 0.0002, "step": 73340 }, { "epoch": 1.42, "learning_rate": 2.62837308904114e-05, "loss": 0.0008, "step": 73342 }, { "epoch": 1.42, "learning_rate": 2.6283083930154173e-05, "loss": 0.0, "step": 73344 }, { "epoch": 1.42, "learning_rate": 2.628243696989694e-05, "loss": 0.0, "step": 73346 }, { "epoch": 1.42, "learning_rate": 2.628179000963971e-05, "loss": 0.0, "step": 73348 }, { "epoch": 1.42, "learning_rate": 2.6281143049382477e-05, "loss": 0.0, "step": 73350 }, { "epoch": 1.42, "learning_rate": 2.6280496089125246e-05, "loss": 0.0, "step": 73352 }, { "epoch": 1.42, "learning_rate": 2.6279849128868015e-05, "loss": 0.0, "step": 73354 }, { "epoch": 1.42, "learning_rate": 2.6279202168610785e-05, "loss": 0.0, "step": 73356 }, { "epoch": 1.42, "learning_rate": 2.627855520835355e-05, "loss": 0.0, "step": 73358 }, { "epoch": 1.42, "learning_rate": 2.6277908248096323e-05, "loss": 0.0, "step": 73360 }, { "epoch": 1.42, "learning_rate": 2.627726128783909e-05, "loss": 0.0001, "step": 73362 }, { "epoch": 1.42, "learning_rate": 2.6276614327581854e-05, "loss": 0.0, "step": 73364 }, { "epoch": 1.42, "learning_rate": 2.6275967367324627e-05, "loss": 0.0, "step": 73366 }, { "epoch": 1.42, "learning_rate": 2.6275320407067396e-05, "loss": 0.0, "step": 73368 }, { "epoch": 1.42, "learning_rate": 2.6274673446810165e-05, "loss": 0.0, "step": 73370 }, { "epoch": 1.42, "learning_rate": 2.6274026486552934e-05, "loss": 0.0, "step": 73372 }, { "epoch": 1.42, "learning_rate": 2.62733795262957e-05, "loss": 0.0, "step": 73374 }, { "epoch": 1.42, "learning_rate": 2.6272732566038472e-05, "loss": 0.0, "step": 73376 }, { "epoch": 1.42, "learning_rate": 2.6272085605781238e-05, "loss": 0.0034, "step": 73378 }, { "epoch": 1.42, "learning_rate": 2.6271438645524003e-05, "loss": 0.0002, "step": 73380 }, { "epoch": 1.42, "learning_rate": 2.6270791685266776e-05, "loss": 0.0, "step": 73382 }, { "epoch": 1.42, "learning_rate": 2.6270144725009545e-05, "loss": 0.0, "step": 73384 }, { "epoch": 1.42, "learning_rate": 2.626949776475231e-05, "loss": 0.0, "step": 73386 }, { "epoch": 1.42, "learning_rate": 2.6268850804495083e-05, "loss": 0.0134, "step": 73388 }, { "epoch": 1.42, "learning_rate": 2.626820384423785e-05, "loss": 0.0, "step": 73390 }, { "epoch": 1.42, "learning_rate": 2.626755688398062e-05, "loss": 0.0, "step": 73392 }, { "epoch": 1.42, "learning_rate": 2.6266909923723387e-05, "loss": 0.0, "step": 73394 }, { "epoch": 1.42, "learning_rate": 2.6266262963466153e-05, "loss": 0.0, "step": 73396 }, { "epoch": 1.42, "learning_rate": 2.6265616003208925e-05, "loss": 0.0, "step": 73398 }, { "epoch": 1.42, "learning_rate": 2.6264969042951694e-05, "loss": 0.0043, "step": 73400 }, { "epoch": 1.42, "learning_rate": 2.626432208269446e-05, "loss": 0.0, "step": 73402 }, { "epoch": 1.42, "learning_rate": 2.6263675122437232e-05, "loss": 0.0, "step": 73404 }, { "epoch": 1.42, "learning_rate": 2.6263028162179998e-05, "loss": 0.0, "step": 73406 }, { "epoch": 1.42, "learning_rate": 2.6262381201922764e-05, "loss": 0.0002, "step": 73408 }, { "epoch": 1.42, "learning_rate": 2.6261734241665536e-05, "loss": 0.0, "step": 73410 }, { "epoch": 1.42, "learning_rate": 2.6261087281408302e-05, "loss": 0.0, "step": 73412 }, { "epoch": 1.42, "learning_rate": 2.6260440321151074e-05, "loss": 0.0, "step": 73414 }, { "epoch": 1.42, "learning_rate": 2.625979336089384e-05, "loss": 0.0, "step": 73416 }, { "epoch": 1.42, "learning_rate": 2.625914640063661e-05, "loss": 0.0, "step": 73418 }, { "epoch": 1.42, "learning_rate": 2.6258499440379382e-05, "loss": 0.0, "step": 73420 }, { "epoch": 1.43, "learning_rate": 2.6257852480122147e-05, "loss": 0.0, "step": 73422 }, { "epoch": 1.43, "learning_rate": 2.6257205519864913e-05, "loss": 0.0, "step": 73424 }, { "epoch": 1.43, "learning_rate": 2.6256558559607686e-05, "loss": 0.0, "step": 73426 }, { "epoch": 1.43, "learning_rate": 2.625591159935045e-05, "loss": 0.0, "step": 73428 }, { "epoch": 1.43, "learning_rate": 2.6255264639093224e-05, "loss": 0.0, "step": 73430 }, { "epoch": 1.43, "learning_rate": 2.625461767883599e-05, "loss": 0.0, "step": 73432 }, { "epoch": 1.43, "learning_rate": 2.625397071857876e-05, "loss": 0.0, "step": 73434 }, { "epoch": 1.43, "learning_rate": 2.625332375832153e-05, "loss": 0.0001, "step": 73436 }, { "epoch": 1.43, "learning_rate": 2.6252676798064297e-05, "loss": 0.0, "step": 73438 }, { "epoch": 1.43, "learning_rate": 2.6252029837807062e-05, "loss": 0.0, "step": 73440 }, { "epoch": 1.43, "learning_rate": 2.6251382877549835e-05, "loss": 0.0028, "step": 73442 }, { "epoch": 1.43, "learning_rate": 2.62507359172926e-05, "loss": 0.0, "step": 73444 }, { "epoch": 1.43, "learning_rate": 2.625008895703537e-05, "loss": 0.0, "step": 73446 }, { "epoch": 1.43, "learning_rate": 2.624944199677814e-05, "loss": 0.0072, "step": 73448 }, { "epoch": 1.43, "learning_rate": 2.6248795036520908e-05, "loss": 0.0, "step": 73450 }, { "epoch": 1.43, "learning_rate": 2.624814807626368e-05, "loss": 0.0, "step": 73452 }, { "epoch": 1.43, "learning_rate": 2.6247501116006446e-05, "loss": 0.0, "step": 73454 }, { "epoch": 1.43, "learning_rate": 2.6246854155749212e-05, "loss": 0.0, "step": 73456 }, { "epoch": 1.43, "learning_rate": 2.6246207195491984e-05, "loss": 0.0, "step": 73458 }, { "epoch": 1.43, "learning_rate": 2.624556023523475e-05, "loss": 0.0, "step": 73460 }, { "epoch": 1.43, "learning_rate": 2.624491327497752e-05, "loss": 0.0, "step": 73462 }, { "epoch": 1.43, "learning_rate": 2.6244266314720288e-05, "loss": 0.0, "step": 73464 }, { "epoch": 1.43, "learning_rate": 2.6243619354463057e-05, "loss": 0.0, "step": 73466 }, { "epoch": 1.43, "learning_rate": 2.6242972394205823e-05, "loss": 0.0012, "step": 73468 }, { "epoch": 1.43, "learning_rate": 2.6242325433948595e-05, "loss": 0.0, "step": 73470 }, { "epoch": 1.43, "learning_rate": 2.624167847369136e-05, "loss": 0.0036, "step": 73472 }, { "epoch": 1.43, "learning_rate": 2.6241031513434133e-05, "loss": 0.0027, "step": 73474 }, { "epoch": 1.43, "learning_rate": 2.62403845531769e-05, "loss": 0.0, "step": 73476 }, { "epoch": 1.43, "learning_rate": 2.6239737592919665e-05, "loss": 0.0001, "step": 73478 }, { "epoch": 1.43, "learning_rate": 2.6239090632662437e-05, "loss": 0.0, "step": 73480 }, { "epoch": 1.43, "learning_rate": 2.6238443672405206e-05, "loss": 0.0, "step": 73482 }, { "epoch": 1.43, "learning_rate": 2.6237796712147972e-05, "loss": 0.0017, "step": 73484 }, { "epoch": 1.43, "learning_rate": 2.6237149751890745e-05, "loss": 0.0037, "step": 73486 }, { "epoch": 1.43, "learning_rate": 2.623650279163351e-05, "loss": 0.0, "step": 73488 }, { "epoch": 1.43, "learning_rate": 2.6235855831376283e-05, "loss": 0.0, "step": 73490 }, { "epoch": 1.43, "learning_rate": 2.623520887111905e-05, "loss": 0.0, "step": 73492 }, { "epoch": 1.43, "learning_rate": 2.6234561910861814e-05, "loss": 0.0, "step": 73494 }, { "epoch": 1.43, "learning_rate": 2.6233914950604587e-05, "loss": 0.0, "step": 73496 }, { "epoch": 1.43, "learning_rate": 2.6233267990347356e-05, "loss": 0.0, "step": 73498 }, { "epoch": 1.43, "learning_rate": 2.623262103009012e-05, "loss": 0.0, "step": 73500 }, { "epoch": 1.43, "learning_rate": 2.6231974069832894e-05, "loss": 0.0, "step": 73502 }, { "epoch": 1.43, "learning_rate": 2.623132710957566e-05, "loss": 0.0, "step": 73504 }, { "epoch": 1.43, "learning_rate": 2.6230680149318425e-05, "loss": 0.0, "step": 73506 }, { "epoch": 1.43, "learning_rate": 2.6230033189061198e-05, "loss": 0.0, "step": 73508 }, { "epoch": 1.43, "learning_rate": 2.6229386228803963e-05, "loss": 0.0013, "step": 73510 }, { "epoch": 1.43, "learning_rate": 2.6228739268546736e-05, "loss": 0.0, "step": 73512 }, { "epoch": 1.43, "learning_rate": 2.6228092308289505e-05, "loss": 0.0006, "step": 73514 }, { "epoch": 1.43, "learning_rate": 2.622744534803227e-05, "loss": 0.0, "step": 73516 }, { "epoch": 1.43, "learning_rate": 2.6226798387775043e-05, "loss": 0.0, "step": 73518 }, { "epoch": 1.43, "learning_rate": 2.622615142751781e-05, "loss": 0.0, "step": 73520 }, { "epoch": 1.43, "learning_rate": 2.6225504467260575e-05, "loss": 0.0, "step": 73522 }, { "epoch": 1.43, "learning_rate": 2.6224857507003347e-05, "loss": 0.0, "step": 73524 }, { "epoch": 1.43, "learning_rate": 2.6224210546746113e-05, "loss": 0.0, "step": 73526 }, { "epoch": 1.43, "learning_rate": 2.6223563586488882e-05, "loss": 0.0, "step": 73528 }, { "epoch": 1.43, "learning_rate": 2.6222916626231654e-05, "loss": 0.001, "step": 73530 }, { "epoch": 1.43, "learning_rate": 2.622226966597442e-05, "loss": 0.0, "step": 73532 }, { "epoch": 1.43, "learning_rate": 2.6221622705717192e-05, "loss": 0.0, "step": 73534 }, { "epoch": 1.43, "learning_rate": 2.6220975745459958e-05, "loss": 0.0, "step": 73536 }, { "epoch": 1.43, "learning_rate": 2.6220328785202724e-05, "loss": 0.0031, "step": 73538 }, { "epoch": 1.43, "learning_rate": 2.6219681824945496e-05, "loss": 0.0, "step": 73540 }, { "epoch": 1.43, "learning_rate": 2.6219034864688262e-05, "loss": 0.0012, "step": 73542 }, { "epoch": 1.43, "learning_rate": 2.621838790443103e-05, "loss": 0.0, "step": 73544 }, { "epoch": 1.43, "learning_rate": 2.6217740944173804e-05, "loss": 0.0, "step": 73546 }, { "epoch": 1.43, "learning_rate": 2.621709398391657e-05, "loss": 0.0001, "step": 73548 }, { "epoch": 1.43, "learning_rate": 2.6216447023659335e-05, "loss": 0.0, "step": 73550 }, { "epoch": 1.43, "learning_rate": 2.6215800063402107e-05, "loss": 0.0, "step": 73552 }, { "epoch": 1.43, "learning_rate": 2.6215153103144873e-05, "loss": 0.0, "step": 73554 }, { "epoch": 1.43, "learning_rate": 2.6214506142887646e-05, "loss": 0.0, "step": 73556 }, { "epoch": 1.43, "learning_rate": 2.621385918263041e-05, "loss": 0.0, "step": 73558 }, { "epoch": 1.43, "learning_rate": 2.621321222237318e-05, "loss": 0.0, "step": 73560 }, { "epoch": 1.43, "learning_rate": 2.6212565262115953e-05, "loss": 0.0, "step": 73562 }, { "epoch": 1.43, "learning_rate": 2.621191830185872e-05, "loss": 0.0, "step": 73564 }, { "epoch": 1.43, "learning_rate": 2.6211271341601484e-05, "loss": 0.0, "step": 73566 }, { "epoch": 1.43, "learning_rate": 2.6210624381344257e-05, "loss": 0.0013, "step": 73568 }, { "epoch": 1.43, "learning_rate": 2.6209977421087022e-05, "loss": 0.0174, "step": 73570 }, { "epoch": 1.43, "learning_rate": 2.6209330460829795e-05, "loss": 0.0, "step": 73572 }, { "epoch": 1.43, "learning_rate": 2.620868350057256e-05, "loss": 0.0001, "step": 73574 }, { "epoch": 1.43, "learning_rate": 2.620803654031533e-05, "loss": 0.0, "step": 73576 }, { "epoch": 1.43, "learning_rate": 2.62073895800581e-05, "loss": 0.0, "step": 73578 }, { "epoch": 1.43, "learning_rate": 2.6206742619800868e-05, "loss": 0.0, "step": 73580 }, { "epoch": 1.43, "learning_rate": 2.6206095659543634e-05, "loss": 0.0, "step": 73582 }, { "epoch": 1.43, "learning_rate": 2.6205448699286406e-05, "loss": 0.0, "step": 73584 }, { "epoch": 1.43, "learning_rate": 2.620480173902917e-05, "loss": 0.0, "step": 73586 }, { "epoch": 1.43, "learning_rate": 2.6204154778771937e-05, "loss": 0.0, "step": 73588 }, { "epoch": 1.43, "learning_rate": 2.620350781851471e-05, "loss": 0.0, "step": 73590 }, { "epoch": 1.43, "learning_rate": 2.620286085825748e-05, "loss": 0.0, "step": 73592 }, { "epoch": 1.43, "learning_rate": 2.6202213898000248e-05, "loss": 0.0, "step": 73594 }, { "epoch": 1.43, "learning_rate": 2.6201566937743017e-05, "loss": 0.0, "step": 73596 }, { "epoch": 1.43, "learning_rate": 2.6200919977485783e-05, "loss": 0.0138, "step": 73598 }, { "epoch": 1.43, "learning_rate": 2.6200273017228555e-05, "loss": 0.0, "step": 73600 }, { "epoch": 1.43, "learning_rate": 2.619962605697132e-05, "loss": 0.0, "step": 73602 }, { "epoch": 1.43, "learning_rate": 2.6198979096714087e-05, "loss": 0.0015, "step": 73604 }, { "epoch": 1.43, "learning_rate": 2.619833213645686e-05, "loss": 0.0001, "step": 73606 }, { "epoch": 1.43, "learning_rate": 2.6197685176199628e-05, "loss": 0.0, "step": 73608 }, { "epoch": 1.43, "learning_rate": 2.6197038215942394e-05, "loss": 0.0, "step": 73610 }, { "epoch": 1.43, "learning_rate": 2.6196391255685166e-05, "loss": 0.0, "step": 73612 }, { "epoch": 1.43, "learning_rate": 2.6195744295427932e-05, "loss": 0.0001, "step": 73614 }, { "epoch": 1.43, "learning_rate": 2.6195097335170705e-05, "loss": 0.0001, "step": 73616 }, { "epoch": 1.43, "learning_rate": 2.619445037491347e-05, "loss": 0.0, "step": 73618 }, { "epoch": 1.43, "learning_rate": 2.6193803414656236e-05, "loss": 0.0003, "step": 73620 }, { "epoch": 1.43, "learning_rate": 2.619315645439901e-05, "loss": 0.0, "step": 73622 }, { "epoch": 1.43, "learning_rate": 2.6192509494141778e-05, "loss": 0.0019, "step": 73624 }, { "epoch": 1.43, "learning_rate": 2.6191862533884543e-05, "loss": 0.0, "step": 73626 }, { "epoch": 1.43, "learning_rate": 2.6191215573627316e-05, "loss": 0.0001, "step": 73628 }, { "epoch": 1.43, "learning_rate": 2.619056861337008e-05, "loss": 0.0, "step": 73630 }, { "epoch": 1.43, "learning_rate": 2.6189921653112854e-05, "loss": 0.0, "step": 73632 }, { "epoch": 1.43, "learning_rate": 2.618927469285562e-05, "loss": 0.0, "step": 73634 }, { "epoch": 1.43, "learning_rate": 2.6188627732598385e-05, "loss": 0.0109, "step": 73636 }, { "epoch": 1.43, "learning_rate": 2.6187980772341158e-05, "loss": 0.0048, "step": 73638 }, { "epoch": 1.43, "learning_rate": 2.6187333812083927e-05, "loss": 0.0001, "step": 73640 }, { "epoch": 1.43, "learning_rate": 2.6186686851826692e-05, "loss": 0.0, "step": 73642 }, { "epoch": 1.43, "learning_rate": 2.6186039891569465e-05, "loss": 0.0, "step": 73644 }, { "epoch": 1.43, "learning_rate": 2.618539293131223e-05, "loss": 0.0001, "step": 73646 }, { "epoch": 1.43, "learning_rate": 2.6184745971054996e-05, "loss": 0.0002, "step": 73648 }, { "epoch": 1.43, "learning_rate": 2.618409901079777e-05, "loss": 0.0, "step": 73650 }, { "epoch": 1.43, "learning_rate": 2.6183452050540535e-05, "loss": 0.0001, "step": 73652 }, { "epoch": 1.43, "learning_rate": 2.6182805090283307e-05, "loss": 0.0, "step": 73654 }, { "epoch": 1.43, "learning_rate": 2.6182158130026073e-05, "loss": 0.0, "step": 73656 }, { "epoch": 1.43, "learning_rate": 2.6181511169768842e-05, "loss": 0.0079, "step": 73658 }, { "epoch": 1.43, "learning_rate": 2.6180864209511614e-05, "loss": 0.0011, "step": 73660 }, { "epoch": 1.43, "learning_rate": 2.618021724925438e-05, "loss": 0.0, "step": 73662 }, { "epoch": 1.43, "learning_rate": 2.6179570288997146e-05, "loss": 0.0, "step": 73664 }, { "epoch": 1.43, "learning_rate": 2.6178923328739918e-05, "loss": 0.0, "step": 73666 }, { "epoch": 1.43, "learning_rate": 2.6178276368482684e-05, "loss": 0.0, "step": 73668 }, { "epoch": 1.43, "learning_rate": 2.6177629408225453e-05, "loss": 0.0029, "step": 73670 }, { "epoch": 1.43, "learning_rate": 2.6176982447968222e-05, "loss": 0.0, "step": 73672 }, { "epoch": 1.43, "learning_rate": 2.617633548771099e-05, "loss": 0.0, "step": 73674 }, { "epoch": 1.43, "learning_rate": 2.6175688527453764e-05, "loss": 0.0, "step": 73676 }, { "epoch": 1.43, "learning_rate": 2.617504156719653e-05, "loss": 0.0, "step": 73678 }, { "epoch": 1.43, "learning_rate": 2.6174394606939295e-05, "loss": 0.0, "step": 73680 }, { "epoch": 1.43, "learning_rate": 2.6173747646682067e-05, "loss": 0.0, "step": 73682 }, { "epoch": 1.43, "learning_rate": 2.6173100686424833e-05, "loss": 0.0, "step": 73684 }, { "epoch": 1.43, "learning_rate": 2.6172453726167602e-05, "loss": 0.0, "step": 73686 }, { "epoch": 1.43, "learning_rate": 2.617180676591037e-05, "loss": 0.0003, "step": 73688 }, { "epoch": 1.43, "learning_rate": 2.617115980565314e-05, "loss": 0.0, "step": 73690 }, { "epoch": 1.43, "learning_rate": 2.6170512845395906e-05, "loss": 0.0, "step": 73692 }, { "epoch": 1.43, "learning_rate": 2.616986588513868e-05, "loss": 0.0, "step": 73694 }, { "epoch": 1.43, "learning_rate": 2.6169218924881444e-05, "loss": 0.0, "step": 73696 }, { "epoch": 1.43, "learning_rate": 2.6168571964624217e-05, "loss": 0.0015, "step": 73698 }, { "epoch": 1.43, "learning_rate": 2.6167925004366982e-05, "loss": 0.0, "step": 73700 }, { "epoch": 1.43, "learning_rate": 2.616727804410975e-05, "loss": 0.0, "step": 73702 }, { "epoch": 1.43, "learning_rate": 2.616663108385252e-05, "loss": 0.0, "step": 73704 }, { "epoch": 1.43, "learning_rate": 2.616598412359529e-05, "loss": 0.0082, "step": 73706 }, { "epoch": 1.43, "learning_rate": 2.6165337163338055e-05, "loss": 0.0, "step": 73708 }, { "epoch": 1.43, "learning_rate": 2.6164690203080828e-05, "loss": 0.0033, "step": 73710 }, { "epoch": 1.43, "learning_rate": 2.6164043242823594e-05, "loss": 0.0, "step": 73712 }, { "epoch": 1.43, "learning_rate": 2.6163396282566366e-05, "loss": 0.0001, "step": 73714 }, { "epoch": 1.43, "learning_rate": 2.616274932230913e-05, "loss": 0.0, "step": 73716 }, { "epoch": 1.43, "learning_rate": 2.6162102362051897e-05, "loss": 0.0, "step": 73718 }, { "epoch": 1.43, "learning_rate": 2.616145540179467e-05, "loss": 0.0, "step": 73720 }, { "epoch": 1.43, "learning_rate": 2.616080844153744e-05, "loss": 0.003, "step": 73722 }, { "epoch": 1.43, "learning_rate": 2.6160161481280205e-05, "loss": 0.0, "step": 73724 }, { "epoch": 1.43, "learning_rate": 2.6159514521022977e-05, "loss": 0.0, "step": 73726 }, { "epoch": 1.43, "learning_rate": 2.6158867560765743e-05, "loss": 0.0007, "step": 73728 }, { "epoch": 1.43, "learning_rate": 2.615822060050851e-05, "loss": 0.0001, "step": 73730 }, { "epoch": 1.43, "learning_rate": 2.615757364025128e-05, "loss": 0.0, "step": 73732 }, { "epoch": 1.43, "learning_rate": 2.6156926679994047e-05, "loss": 0.0, "step": 73734 }, { "epoch": 1.43, "learning_rate": 2.615627971973682e-05, "loss": 0.0, "step": 73736 }, { "epoch": 1.43, "learning_rate": 2.6155632759479588e-05, "loss": 0.0, "step": 73738 }, { "epoch": 1.43, "learning_rate": 2.6154985799222354e-05, "loss": 0.0003, "step": 73740 }, { "epoch": 1.43, "learning_rate": 2.6154338838965126e-05, "loss": 0.0, "step": 73742 }, { "epoch": 1.43, "learning_rate": 2.6153691878707892e-05, "loss": 0.0, "step": 73744 }, { "epoch": 1.43, "learning_rate": 2.6153044918450658e-05, "loss": 0.0002, "step": 73746 }, { "epoch": 1.43, "learning_rate": 2.615239795819343e-05, "loss": 0.0022, "step": 73748 }, { "epoch": 1.43, "learning_rate": 2.6151750997936196e-05, "loss": 0.0, "step": 73750 }, { "epoch": 1.43, "learning_rate": 2.6151104037678965e-05, "loss": 0.0001, "step": 73752 }, { "epoch": 1.43, "learning_rate": 2.6150457077421738e-05, "loss": 0.0, "step": 73754 }, { "epoch": 1.43, "learning_rate": 2.6149810117164503e-05, "loss": 0.013, "step": 73756 }, { "epoch": 1.43, "learning_rate": 2.6149163156907276e-05, "loss": 0.0, "step": 73758 }, { "epoch": 1.43, "learning_rate": 2.614851619665004e-05, "loss": 0.0, "step": 73760 }, { "epoch": 1.43, "learning_rate": 2.6147869236392807e-05, "loss": 0.0, "step": 73762 }, { "epoch": 1.43, "learning_rate": 2.614722227613558e-05, "loss": 0.0, "step": 73764 }, { "epoch": 1.43, "learning_rate": 2.6146575315878345e-05, "loss": 0.0005, "step": 73766 }, { "epoch": 1.43, "learning_rate": 2.6145928355621114e-05, "loss": 0.007, "step": 73768 }, { "epoch": 1.43, "learning_rate": 2.6145281395363887e-05, "loss": 0.0, "step": 73770 }, { "epoch": 1.43, "learning_rate": 2.6144634435106652e-05, "loss": 0.0, "step": 73772 }, { "epoch": 1.43, "learning_rate": 2.6143987474849425e-05, "loss": 0.0, "step": 73774 }, { "epoch": 1.43, "learning_rate": 2.614334051459219e-05, "loss": 0.0, "step": 73776 }, { "epoch": 1.43, "learning_rate": 2.6142693554334956e-05, "loss": 0.0, "step": 73778 }, { "epoch": 1.43, "learning_rate": 2.614204659407773e-05, "loss": 0.0001, "step": 73780 }, { "epoch": 1.43, "learning_rate": 2.6141399633820495e-05, "loss": 0.0, "step": 73782 }, { "epoch": 1.43, "learning_rate": 2.6140752673563264e-05, "loss": 0.0, "step": 73784 }, { "epoch": 1.43, "learning_rate": 2.6140105713306036e-05, "loss": 0.0, "step": 73786 }, { "epoch": 1.43, "learning_rate": 2.6139458753048802e-05, "loss": 0.0123, "step": 73788 }, { "epoch": 1.43, "learning_rate": 2.6138811792791567e-05, "loss": 0.0, "step": 73790 }, { "epoch": 1.43, "learning_rate": 2.613816483253434e-05, "loss": 0.0, "step": 73792 }, { "epoch": 1.43, "learning_rate": 2.6137517872277106e-05, "loss": 0.0001, "step": 73794 }, { "epoch": 1.43, "learning_rate": 2.6136870912019878e-05, "loss": 0.0, "step": 73796 }, { "epoch": 1.43, "learning_rate": 2.6136223951762644e-05, "loss": 0.0, "step": 73798 }, { "epoch": 1.43, "learning_rate": 2.6135576991505413e-05, "loss": 0.0, "step": 73800 }, { "epoch": 1.43, "learning_rate": 2.6134930031248185e-05, "loss": 0.0, "step": 73802 }, { "epoch": 1.43, "learning_rate": 2.613428307099095e-05, "loss": 0.0001, "step": 73804 }, { "epoch": 1.43, "learning_rate": 2.6133636110733717e-05, "loss": 0.0001, "step": 73806 }, { "epoch": 1.43, "learning_rate": 2.613298915047649e-05, "loss": 0.0265, "step": 73808 }, { "epoch": 1.43, "learning_rate": 2.6132342190219255e-05, "loss": 0.0016, "step": 73810 }, { "epoch": 1.43, "learning_rate": 2.613169522996202e-05, "loss": 0.0, "step": 73812 }, { "epoch": 1.43, "learning_rate": 2.6131048269704793e-05, "loss": 0.0074, "step": 73814 }, { "epoch": 1.43, "learning_rate": 2.6130401309447562e-05, "loss": 0.0, "step": 73816 }, { "epoch": 1.43, "learning_rate": 2.612975434919033e-05, "loss": 0.0, "step": 73818 }, { "epoch": 1.43, "learning_rate": 2.61291073889331e-05, "loss": 0.0, "step": 73820 }, { "epoch": 1.43, "learning_rate": 2.6128460428675866e-05, "loss": 0.0004, "step": 73822 }, { "epoch": 1.43, "learning_rate": 2.612781346841864e-05, "loss": 0.0004, "step": 73824 }, { "epoch": 1.43, "learning_rate": 2.6127166508161404e-05, "loss": 0.0, "step": 73826 }, { "epoch": 1.43, "learning_rate": 2.612651954790417e-05, "loss": 0.0, "step": 73828 }, { "epoch": 1.43, "learning_rate": 2.6125872587646942e-05, "loss": 0.0, "step": 73830 }, { "epoch": 1.43, "learning_rate": 2.612522562738971e-05, "loss": 0.0, "step": 73832 }, { "epoch": 1.43, "learning_rate": 2.6124578667132477e-05, "loss": 0.0, "step": 73834 }, { "epoch": 1.43, "learning_rate": 2.612393170687525e-05, "loss": 0.0, "step": 73836 }, { "epoch": 1.43, "learning_rate": 2.6123284746618015e-05, "loss": 0.0013, "step": 73838 }, { "epoch": 1.43, "learning_rate": 2.6122637786360788e-05, "loss": 0.0006, "step": 73840 }, { "epoch": 1.43, "learning_rate": 2.6121990826103553e-05, "loss": 0.0001, "step": 73842 }, { "epoch": 1.43, "learning_rate": 2.612134386584632e-05, "loss": 0.0, "step": 73844 }, { "epoch": 1.43, "learning_rate": 2.612069690558909e-05, "loss": 0.0041, "step": 73846 }, { "epoch": 1.43, "learning_rate": 2.612004994533186e-05, "loss": 0.0, "step": 73848 }, { "epoch": 1.43, "learning_rate": 2.6119402985074626e-05, "loss": 0.0, "step": 73850 }, { "epoch": 1.43, "learning_rate": 2.61187560248174e-05, "loss": 0.0034, "step": 73852 }, { "epoch": 1.43, "learning_rate": 2.6118109064560165e-05, "loss": 0.0, "step": 73854 }, { "epoch": 1.43, "learning_rate": 2.6117462104302937e-05, "loss": 0.0, "step": 73856 }, { "epoch": 1.43, "learning_rate": 2.6116815144045703e-05, "loss": 0.0003, "step": 73858 }, { "epoch": 1.43, "learning_rate": 2.611616818378847e-05, "loss": 0.0, "step": 73860 }, { "epoch": 1.43, "learning_rate": 2.611552122353124e-05, "loss": 0.0, "step": 73862 }, { "epoch": 1.43, "learning_rate": 2.611487426327401e-05, "loss": 0.0, "step": 73864 }, { "epoch": 1.43, "learning_rate": 2.6114227303016776e-05, "loss": 0.0, "step": 73866 }, { "epoch": 1.43, "learning_rate": 2.6113580342759548e-05, "loss": 0.0008, "step": 73868 }, { "epoch": 1.43, "learning_rate": 2.6112933382502314e-05, "loss": 0.0001, "step": 73870 }, { "epoch": 1.43, "learning_rate": 2.611228642224508e-05, "loss": 0.0001, "step": 73872 }, { "epoch": 1.43, "learning_rate": 2.6111639461987852e-05, "loss": 0.0, "step": 73874 }, { "epoch": 1.43, "learning_rate": 2.6110992501730618e-05, "loss": 0.0001, "step": 73876 }, { "epoch": 1.43, "learning_rate": 2.611034554147339e-05, "loss": 0.0, "step": 73878 }, { "epoch": 1.43, "learning_rate": 2.610969858121616e-05, "loss": 0.0377, "step": 73880 }, { "epoch": 1.43, "learning_rate": 2.6109051620958925e-05, "loss": 0.01, "step": 73882 }, { "epoch": 1.43, "learning_rate": 2.6108404660701697e-05, "loss": 0.0001, "step": 73884 }, { "epoch": 1.43, "learning_rate": 2.6107757700444463e-05, "loss": 0.0, "step": 73886 }, { "epoch": 1.43, "learning_rate": 2.610711074018723e-05, "loss": 0.0, "step": 73888 }, { "epoch": 1.43, "learning_rate": 2.610646377993e-05, "loss": 0.0, "step": 73890 }, { "epoch": 1.43, "learning_rate": 2.6105816819672767e-05, "loss": 0.0001, "step": 73892 }, { "epoch": 1.43, "learning_rate": 2.6105169859415536e-05, "loss": 0.0, "step": 73894 }, { "epoch": 1.43, "learning_rate": 2.6104522899158305e-05, "loss": 0.0001, "step": 73896 }, { "epoch": 1.43, "learning_rate": 2.6103875938901074e-05, "loss": 0.0, "step": 73898 }, { "epoch": 1.43, "learning_rate": 2.6103228978643847e-05, "loss": 0.0002, "step": 73900 }, { "epoch": 1.43, "learning_rate": 2.6102582018386612e-05, "loss": 0.0065, "step": 73902 }, { "epoch": 1.43, "learning_rate": 2.6101935058129378e-05, "loss": 0.0, "step": 73904 }, { "epoch": 1.43, "learning_rate": 2.610128809787215e-05, "loss": 0.0, "step": 73906 }, { "epoch": 1.43, "learning_rate": 2.6100641137614916e-05, "loss": 0.0, "step": 73908 }, { "epoch": 1.43, "learning_rate": 2.6099994177357685e-05, "loss": 0.001, "step": 73910 }, { "epoch": 1.43, "learning_rate": 2.6099347217100455e-05, "loss": 0.0, "step": 73912 }, { "epoch": 1.43, "learning_rate": 2.6098700256843224e-05, "loss": 0.0, "step": 73914 }, { "epoch": 1.43, "learning_rate": 2.6098053296585996e-05, "loss": 0.0, "step": 73916 }, { "epoch": 1.43, "learning_rate": 2.6097406336328762e-05, "loss": 0.0003, "step": 73918 }, { "epoch": 1.43, "learning_rate": 2.6096759376071527e-05, "loss": 0.0, "step": 73920 }, { "epoch": 1.43, "learning_rate": 2.60961124158143e-05, "loss": 0.0, "step": 73922 }, { "epoch": 1.43, "learning_rate": 2.6095465455557066e-05, "loss": 0.0, "step": 73924 }, { "epoch": 1.43, "learning_rate": 2.6094818495299835e-05, "loss": 0.0, "step": 73926 }, { "epoch": 1.43, "learning_rate": 2.6094171535042604e-05, "loss": 0.0, "step": 73928 }, { "epoch": 1.43, "learning_rate": 2.6093524574785373e-05, "loss": 0.0, "step": 73930 }, { "epoch": 1.43, "learning_rate": 2.609287761452814e-05, "loss": 0.0, "step": 73932 }, { "epoch": 1.43, "learning_rate": 2.609223065427091e-05, "loss": 0.0, "step": 73934 }, { "epoch": 1.44, "learning_rate": 2.6091583694013677e-05, "loss": 0.0, "step": 73936 }, { "epoch": 1.44, "learning_rate": 2.609093673375645e-05, "loss": 0.0021, "step": 73938 }, { "epoch": 1.44, "learning_rate": 2.6090289773499215e-05, "loss": 0.0001, "step": 73940 }, { "epoch": 1.44, "learning_rate": 2.6089642813241984e-05, "loss": 0.0, "step": 73942 }, { "epoch": 1.44, "learning_rate": 2.6088995852984753e-05, "loss": 0.0, "step": 73944 }, { "epoch": 1.44, "learning_rate": 2.6088348892727522e-05, "loss": 0.0, "step": 73946 }, { "epoch": 1.44, "learning_rate": 2.6087701932470288e-05, "loss": 0.0, "step": 73948 }, { "epoch": 1.44, "learning_rate": 2.608705497221306e-05, "loss": 0.0, "step": 73950 }, { "epoch": 1.44, "learning_rate": 2.6086408011955826e-05, "loss": 0.0, "step": 73952 }, { "epoch": 1.44, "learning_rate": 2.6085761051698592e-05, "loss": 0.0, "step": 73954 }, { "epoch": 1.44, "learning_rate": 2.6085114091441364e-05, "loss": 0.0, "step": 73956 }, { "epoch": 1.44, "learning_rate": 2.608446713118413e-05, "loss": 0.0, "step": 73958 }, { "epoch": 1.44, "learning_rate": 2.6083820170926902e-05, "loss": 0.0, "step": 73960 }, { "epoch": 1.44, "learning_rate": 2.608317321066967e-05, "loss": 0.0, "step": 73962 }, { "epoch": 1.44, "learning_rate": 2.6082526250412437e-05, "loss": 0.0, "step": 73964 }, { "epoch": 1.44, "learning_rate": 2.608187929015521e-05, "loss": 0.0, "step": 73966 }, { "epoch": 1.44, "learning_rate": 2.6081232329897975e-05, "loss": 0.0, "step": 73968 }, { "epoch": 1.44, "learning_rate": 2.608058536964074e-05, "loss": 0.0, "step": 73970 }, { "epoch": 1.44, "learning_rate": 2.6079938409383513e-05, "loss": 0.0, "step": 73972 }, { "epoch": 1.44, "learning_rate": 2.607929144912628e-05, "loss": 0.0, "step": 73974 }, { "epoch": 1.44, "learning_rate": 2.6078644488869048e-05, "loss": 0.0, "step": 73976 }, { "epoch": 1.44, "learning_rate": 2.607799752861182e-05, "loss": 0.0, "step": 73978 }, { "epoch": 1.44, "learning_rate": 2.6077350568354586e-05, "loss": 0.0001, "step": 73980 }, { "epoch": 1.44, "learning_rate": 2.607670360809736e-05, "loss": 0.0002, "step": 73982 }, { "epoch": 1.44, "learning_rate": 2.6076056647840125e-05, "loss": 0.0064, "step": 73984 }, { "epoch": 1.44, "learning_rate": 2.607540968758289e-05, "loss": 0.0, "step": 73986 }, { "epoch": 1.44, "learning_rate": 2.6074762727325663e-05, "loss": 0.0, "step": 73988 }, { "epoch": 1.44, "learning_rate": 2.607411576706843e-05, "loss": 0.0, "step": 73990 }, { "epoch": 1.44, "learning_rate": 2.6073468806811198e-05, "loss": 0.0, "step": 73992 }, { "epoch": 1.44, "learning_rate": 2.607282184655397e-05, "loss": 0.0, "step": 73994 }, { "epoch": 1.44, "learning_rate": 2.6072174886296736e-05, "loss": 0.0, "step": 73996 }, { "epoch": 1.44, "learning_rate": 2.6071527926039508e-05, "loss": 0.0034, "step": 73998 }, { "epoch": 1.44, "learning_rate": 2.6070880965782274e-05, "loss": 0.0, "step": 74000 }, { "epoch": 1.44, "learning_rate": 2.607023400552504e-05, "loss": 0.0, "step": 74002 }, { "epoch": 1.44, "learning_rate": 2.6069587045267812e-05, "loss": 0.0, "step": 74004 }, { "epoch": 1.44, "learning_rate": 2.6068940085010578e-05, "loss": 0.002, "step": 74006 }, { "epoch": 1.44, "learning_rate": 2.6068293124753347e-05, "loss": 0.0128, "step": 74008 }, { "epoch": 1.44, "learning_rate": 2.606764616449612e-05, "loss": 0.0, "step": 74010 }, { "epoch": 1.44, "learning_rate": 2.6066999204238885e-05, "loss": 0.0, "step": 74012 }, { "epoch": 1.44, "learning_rate": 2.606635224398165e-05, "loss": 0.0, "step": 74014 }, { "epoch": 1.44, "learning_rate": 2.6065705283724423e-05, "loss": 0.0, "step": 74016 }, { "epoch": 1.44, "learning_rate": 2.606505832346719e-05, "loss": 0.0, "step": 74018 }, { "epoch": 1.44, "learning_rate": 2.606441136320996e-05, "loss": 0.0013, "step": 74020 }, { "epoch": 1.44, "learning_rate": 2.6063764402952727e-05, "loss": 0.0, "step": 74022 }, { "epoch": 1.44, "learning_rate": 2.6063117442695496e-05, "loss": 0.0002, "step": 74024 }, { "epoch": 1.44, "learning_rate": 2.606247048243827e-05, "loss": 0.0, "step": 74026 }, { "epoch": 1.44, "learning_rate": 2.6061823522181034e-05, "loss": 0.0, "step": 74028 }, { "epoch": 1.44, "learning_rate": 2.60611765619238e-05, "loss": 0.0001, "step": 74030 }, { "epoch": 1.44, "learning_rate": 2.6060529601666572e-05, "loss": 0.0, "step": 74032 }, { "epoch": 1.44, "learning_rate": 2.6059882641409338e-05, "loss": 0.0001, "step": 74034 }, { "epoch": 1.44, "learning_rate": 2.6059235681152104e-05, "loss": 0.0, "step": 74036 }, { "epoch": 1.44, "learning_rate": 2.6058588720894876e-05, "loss": 0.0, "step": 74038 }, { "epoch": 1.44, "learning_rate": 2.6057941760637645e-05, "loss": 0.0, "step": 74040 }, { "epoch": 1.44, "learning_rate": 2.6057294800380418e-05, "loss": 0.0005, "step": 74042 }, { "epoch": 1.44, "learning_rate": 2.6056647840123184e-05, "loss": 0.0, "step": 74044 }, { "epoch": 1.44, "learning_rate": 2.605600087986595e-05, "loss": 0.0, "step": 74046 }, { "epoch": 1.44, "learning_rate": 2.6055353919608722e-05, "loss": 0.0014, "step": 74048 }, { "epoch": 1.44, "learning_rate": 2.6054706959351487e-05, "loss": 0.0, "step": 74050 }, { "epoch": 1.44, "learning_rate": 2.6054059999094253e-05, "loss": 0.0, "step": 74052 }, { "epoch": 1.44, "learning_rate": 2.6053413038837026e-05, "loss": 0.0, "step": 74054 }, { "epoch": 1.44, "learning_rate": 2.6052766078579795e-05, "loss": 0.0, "step": 74056 }, { "epoch": 1.44, "learning_rate": 2.6052119118322564e-05, "loss": 0.0005, "step": 74058 }, { "epoch": 1.44, "learning_rate": 2.6051472158065333e-05, "loss": 0.0, "step": 74060 }, { "epoch": 1.44, "learning_rate": 2.60508251978081e-05, "loss": 0.0014, "step": 74062 }, { "epoch": 1.44, "learning_rate": 2.605017823755087e-05, "loss": 0.0, "step": 74064 }, { "epoch": 1.44, "learning_rate": 2.6049531277293637e-05, "loss": 0.0076, "step": 74066 }, { "epoch": 1.44, "learning_rate": 2.6048884317036402e-05, "loss": 0.0, "step": 74068 }, { "epoch": 1.44, "learning_rate": 2.6048237356779175e-05, "loss": 0.0054, "step": 74070 }, { "epoch": 1.44, "learning_rate": 2.6047590396521944e-05, "loss": 0.0, "step": 74072 }, { "epoch": 1.44, "learning_rate": 2.604694343626471e-05, "loss": 0.0, "step": 74074 }, { "epoch": 1.44, "learning_rate": 2.6046296476007482e-05, "loss": 0.0, "step": 74076 }, { "epoch": 1.44, "learning_rate": 2.6045649515750248e-05, "loss": 0.0, "step": 74078 }, { "epoch": 1.44, "learning_rate": 2.604500255549302e-05, "loss": 0.0, "step": 74080 }, { "epoch": 1.44, "learning_rate": 2.6044355595235786e-05, "loss": 0.0, "step": 74082 }, { "epoch": 1.44, "learning_rate": 2.6043708634978552e-05, "loss": 0.0, "step": 74084 }, { "epoch": 1.44, "learning_rate": 2.6043061674721324e-05, "loss": 0.0, "step": 74086 }, { "epoch": 1.44, "learning_rate": 2.6042414714464093e-05, "loss": 0.0, "step": 74088 }, { "epoch": 1.44, "learning_rate": 2.604176775420686e-05, "loss": 0.0, "step": 74090 }, { "epoch": 1.44, "learning_rate": 2.604112079394963e-05, "loss": 0.0, "step": 74092 }, { "epoch": 1.44, "learning_rate": 2.6040473833692397e-05, "loss": 0.0, "step": 74094 }, { "epoch": 1.44, "learning_rate": 2.6039826873435163e-05, "loss": 0.0, "step": 74096 }, { "epoch": 1.44, "learning_rate": 2.6039179913177935e-05, "loss": 0.0, "step": 74098 }, { "epoch": 1.44, "learning_rate": 2.60385329529207e-05, "loss": 0.0, "step": 74100 }, { "epoch": 1.44, "learning_rate": 2.6037885992663473e-05, "loss": 0.0001, "step": 74102 }, { "epoch": 1.44, "learning_rate": 2.6037239032406243e-05, "loss": 0.0, "step": 74104 }, { "epoch": 1.44, "learning_rate": 2.6036592072149008e-05, "loss": 0.0, "step": 74106 }, { "epoch": 1.44, "learning_rate": 2.603594511189178e-05, "loss": 0.0004, "step": 74108 }, { "epoch": 1.44, "learning_rate": 2.6035298151634546e-05, "loss": 0.0, "step": 74110 }, { "epoch": 1.44, "learning_rate": 2.6034651191377312e-05, "loss": 0.0001, "step": 74112 }, { "epoch": 1.44, "learning_rate": 2.6034004231120085e-05, "loss": 0.0, "step": 74114 }, { "epoch": 1.44, "learning_rate": 2.603335727086285e-05, "loss": 0.0, "step": 74116 }, { "epoch": 1.44, "learning_rate": 2.603271031060562e-05, "loss": 0.0124, "step": 74118 }, { "epoch": 1.44, "learning_rate": 2.6032063350348392e-05, "loss": 0.0002, "step": 74120 }, { "epoch": 1.44, "learning_rate": 2.6031416390091158e-05, "loss": 0.0, "step": 74122 }, { "epoch": 1.44, "learning_rate": 2.603076942983393e-05, "loss": 0.0, "step": 74124 }, { "epoch": 1.44, "learning_rate": 2.6030122469576696e-05, "loss": 0.0, "step": 74126 }, { "epoch": 1.44, "learning_rate": 2.602947550931946e-05, "loss": 0.0003, "step": 74128 }, { "epoch": 1.44, "learning_rate": 2.6028828549062234e-05, "loss": 0.0, "step": 74130 }, { "epoch": 1.44, "learning_rate": 2.6028181588805e-05, "loss": 0.0, "step": 74132 }, { "epoch": 1.44, "learning_rate": 2.602753462854777e-05, "loss": 0.0, "step": 74134 }, { "epoch": 1.44, "learning_rate": 2.6026887668290538e-05, "loss": 0.0075, "step": 74136 }, { "epoch": 1.44, "learning_rate": 2.6026240708033307e-05, "loss": 0.0, "step": 74138 }, { "epoch": 1.44, "learning_rate": 2.602559374777608e-05, "loss": 0.0, "step": 74140 }, { "epoch": 1.44, "learning_rate": 2.6024946787518845e-05, "loss": 0.0, "step": 74142 }, { "epoch": 1.44, "learning_rate": 2.602429982726161e-05, "loss": 0.0, "step": 74144 }, { "epoch": 1.44, "learning_rate": 2.6023652867004383e-05, "loss": 0.0, "step": 74146 }, { "epoch": 1.44, "learning_rate": 2.602300590674715e-05, "loss": 0.0001, "step": 74148 }, { "epoch": 1.44, "learning_rate": 2.6022358946489918e-05, "loss": 0.0031, "step": 74150 }, { "epoch": 1.44, "learning_rate": 2.6021711986232687e-05, "loss": 0.0, "step": 74152 }, { "epoch": 1.44, "learning_rate": 2.6021065025975456e-05, "loss": 0.0011, "step": 74154 }, { "epoch": 1.44, "learning_rate": 2.6020418065718222e-05, "loss": 0.0, "step": 74156 }, { "epoch": 1.44, "learning_rate": 2.6019771105460994e-05, "loss": 0.0, "step": 74158 }, { "epoch": 1.44, "learning_rate": 2.601912414520376e-05, "loss": 0.0, "step": 74160 }, { "epoch": 1.44, "learning_rate": 2.6018477184946532e-05, "loss": 0.0003, "step": 74162 }, { "epoch": 1.44, "learning_rate": 2.6017830224689298e-05, "loss": 0.0, "step": 74164 }, { "epoch": 1.44, "learning_rate": 2.6017183264432067e-05, "loss": 0.0, "step": 74166 }, { "epoch": 1.44, "learning_rate": 2.6016536304174836e-05, "loss": 0.0, "step": 74168 }, { "epoch": 1.44, "learning_rate": 2.6015889343917605e-05, "loss": 0.0, "step": 74170 }, { "epoch": 1.44, "learning_rate": 2.601524238366037e-05, "loss": 0.0, "step": 74172 }, { "epoch": 1.44, "learning_rate": 2.6014595423403144e-05, "loss": 0.0, "step": 74174 }, { "epoch": 1.44, "learning_rate": 2.601394846314591e-05, "loss": 0.0, "step": 74176 }, { "epoch": 1.44, "learning_rate": 2.6013301502888675e-05, "loss": 0.0, "step": 74178 }, { "epoch": 1.44, "learning_rate": 2.6012654542631447e-05, "loss": 0.0, "step": 74180 }, { "epoch": 1.44, "learning_rate": 2.6012007582374217e-05, "loss": 0.0, "step": 74182 }, { "epoch": 1.44, "learning_rate": 2.6011360622116986e-05, "loss": 0.0001, "step": 74184 }, { "epoch": 1.44, "learning_rate": 2.6010713661859755e-05, "loss": 0.0169, "step": 74186 }, { "epoch": 1.44, "learning_rate": 2.601006670160252e-05, "loss": 0.0, "step": 74188 }, { "epoch": 1.44, "learning_rate": 2.6009419741345293e-05, "loss": 0.0, "step": 74190 }, { "epoch": 1.44, "learning_rate": 2.600877278108806e-05, "loss": 0.0, "step": 74192 }, { "epoch": 1.44, "learning_rate": 2.6008125820830824e-05, "loss": 0.0, "step": 74194 }, { "epoch": 1.44, "learning_rate": 2.6007478860573597e-05, "loss": 0.0423, "step": 74196 }, { "epoch": 1.44, "learning_rate": 2.6006831900316362e-05, "loss": 0.0, "step": 74198 }, { "epoch": 1.44, "learning_rate": 2.600618494005913e-05, "loss": 0.004, "step": 74200 }, { "epoch": 1.44, "learning_rate": 2.6005537979801904e-05, "loss": 0.0, "step": 74202 }, { "epoch": 1.44, "learning_rate": 2.600489101954467e-05, "loss": 0.0001, "step": 74204 }, { "epoch": 1.44, "learning_rate": 2.6004244059287442e-05, "loss": 0.0, "step": 74206 }, { "epoch": 1.44, "learning_rate": 2.6003597099030208e-05, "loss": 0.0, "step": 74208 }, { "epoch": 1.44, "learning_rate": 2.6002950138772974e-05, "loss": 0.0, "step": 74210 }, { "epoch": 1.44, "learning_rate": 2.6002303178515746e-05, "loss": 0.0, "step": 74212 }, { "epoch": 1.44, "learning_rate": 2.6001656218258512e-05, "loss": 0.0, "step": 74214 }, { "epoch": 1.44, "learning_rate": 2.600100925800128e-05, "loss": 0.0, "step": 74216 }, { "epoch": 1.44, "learning_rate": 2.6000362297744053e-05, "loss": 0.0, "step": 74218 }, { "epoch": 1.44, "learning_rate": 2.599971533748682e-05, "loss": 0.0, "step": 74220 }, { "epoch": 1.44, "learning_rate": 2.599906837722959e-05, "loss": 0.0, "step": 74222 }, { "epoch": 1.44, "learning_rate": 2.5998421416972357e-05, "loss": 0.0, "step": 74224 }, { "epoch": 1.44, "learning_rate": 2.5997774456715123e-05, "loss": 0.0, "step": 74226 }, { "epoch": 1.44, "learning_rate": 2.5997127496457895e-05, "loss": 0.0, "step": 74228 }, { "epoch": 1.44, "learning_rate": 2.599648053620066e-05, "loss": 0.0, "step": 74230 }, { "epoch": 1.44, "learning_rate": 2.599583357594343e-05, "loss": 0.0, "step": 74232 }, { "epoch": 1.44, "learning_rate": 2.5995186615686203e-05, "loss": 0.0003, "step": 74234 }, { "epoch": 1.44, "learning_rate": 2.5994539655428968e-05, "loss": 0.0002, "step": 74236 }, { "epoch": 1.44, "learning_rate": 2.5993892695171734e-05, "loss": 0.0, "step": 74238 }, { "epoch": 1.44, "learning_rate": 2.5993245734914506e-05, "loss": 0.0, "step": 74240 }, { "epoch": 1.44, "learning_rate": 2.5992598774657272e-05, "loss": 0.0, "step": 74242 }, { "epoch": 1.44, "learning_rate": 2.5991951814400045e-05, "loss": 0.0002, "step": 74244 }, { "epoch": 1.44, "learning_rate": 2.599130485414281e-05, "loss": 0.0, "step": 74246 }, { "epoch": 1.44, "learning_rate": 2.599065789388558e-05, "loss": 0.0, "step": 74248 }, { "epoch": 1.44, "learning_rate": 2.5990010933628352e-05, "loss": 0.0, "step": 74250 }, { "epoch": 1.44, "learning_rate": 2.5989363973371118e-05, "loss": 0.0029, "step": 74252 }, { "epoch": 1.44, "learning_rate": 2.5988717013113883e-05, "loss": 0.0144, "step": 74254 }, { "epoch": 1.44, "learning_rate": 2.5988070052856656e-05, "loss": 0.0126, "step": 74256 }, { "epoch": 1.44, "learning_rate": 2.598742309259942e-05, "loss": 0.0, "step": 74258 }, { "epoch": 1.44, "learning_rate": 2.5986776132342187e-05, "loss": 0.0, "step": 74260 }, { "epoch": 1.44, "learning_rate": 2.598612917208496e-05, "loss": 0.0, "step": 74262 }, { "epoch": 1.44, "learning_rate": 2.598548221182773e-05, "loss": 0.0001, "step": 74264 }, { "epoch": 1.44, "learning_rate": 2.59848352515705e-05, "loss": 0.0002, "step": 74266 }, { "epoch": 1.44, "learning_rate": 2.5984188291313267e-05, "loss": 0.0, "step": 74268 }, { "epoch": 1.44, "learning_rate": 2.5983541331056033e-05, "loss": 0.0, "step": 74270 }, { "epoch": 1.44, "learning_rate": 2.5982894370798805e-05, "loss": 0.0, "step": 74272 }, { "epoch": 1.44, "learning_rate": 2.598224741054157e-05, "loss": 0.0, "step": 74274 }, { "epoch": 1.44, "learning_rate": 2.5981600450284336e-05, "loss": 0.0, "step": 74276 }, { "epoch": 1.44, "learning_rate": 2.598095349002711e-05, "loss": 0.0, "step": 74278 }, { "epoch": 1.44, "learning_rate": 2.5980306529769878e-05, "loss": 0.0001, "step": 74280 }, { "epoch": 1.44, "learning_rate": 2.597965956951265e-05, "loss": 0.0, "step": 74282 }, { "epoch": 1.44, "learning_rate": 2.5979012609255416e-05, "loss": 0.0, "step": 74284 }, { "epoch": 1.44, "learning_rate": 2.5978365648998182e-05, "loss": 0.004, "step": 74286 }, { "epoch": 1.44, "learning_rate": 2.5977718688740954e-05, "loss": 0.0, "step": 74288 }, { "epoch": 1.44, "learning_rate": 2.597707172848372e-05, "loss": 0.0, "step": 74290 }, { "epoch": 1.44, "learning_rate": 2.5976424768226486e-05, "loss": 0.0, "step": 74292 }, { "epoch": 1.44, "learning_rate": 2.5975777807969258e-05, "loss": 0.0, "step": 74294 }, { "epoch": 1.44, "learning_rate": 2.5975130847712027e-05, "loss": 0.0, "step": 74296 }, { "epoch": 1.44, "learning_rate": 2.5974483887454793e-05, "loss": 0.0, "step": 74298 }, { "epoch": 1.44, "learning_rate": 2.5973836927197565e-05, "loss": 0.0, "step": 74300 }, { "epoch": 1.44, "learning_rate": 2.597318996694033e-05, "loss": 0.0001, "step": 74302 }, { "epoch": 1.44, "learning_rate": 2.5972543006683104e-05, "loss": 0.0, "step": 74304 }, { "epoch": 1.44, "learning_rate": 2.597189604642587e-05, "loss": 0.0, "step": 74306 }, { "epoch": 1.44, "learning_rate": 2.5971249086168635e-05, "loss": 0.0, "step": 74308 }, { "epoch": 1.44, "learning_rate": 2.5970602125911407e-05, "loss": 0.0, "step": 74310 }, { "epoch": 1.44, "learning_rate": 2.5969955165654177e-05, "loss": 0.0, "step": 74312 }, { "epoch": 1.44, "learning_rate": 2.5969308205396942e-05, "loss": 0.0, "step": 74314 }, { "epoch": 1.44, "learning_rate": 2.5968661245139715e-05, "loss": 0.0, "step": 74316 }, { "epoch": 1.44, "learning_rate": 2.596801428488248e-05, "loss": 0.036, "step": 74318 }, { "epoch": 1.44, "learning_rate": 2.5967367324625246e-05, "loss": 0.0054, "step": 74320 }, { "epoch": 1.44, "learning_rate": 2.596672036436802e-05, "loss": 0.0, "step": 74322 }, { "epoch": 1.44, "learning_rate": 2.5966073404110784e-05, "loss": 0.0, "step": 74324 }, { "epoch": 1.44, "learning_rate": 2.5965426443853557e-05, "loss": 0.0, "step": 74326 }, { "epoch": 1.44, "learning_rate": 2.5964779483596326e-05, "loss": 0.0, "step": 74328 }, { "epoch": 1.44, "learning_rate": 2.596413252333909e-05, "loss": 0.0005, "step": 74330 }, { "epoch": 1.44, "learning_rate": 2.5963485563081864e-05, "loss": 0.0093, "step": 74332 }, { "epoch": 1.44, "learning_rate": 2.596283860282463e-05, "loss": 0.0, "step": 74334 }, { "epoch": 1.44, "learning_rate": 2.5962191642567395e-05, "loss": 0.0, "step": 74336 }, { "epoch": 1.44, "learning_rate": 2.5961544682310168e-05, "loss": 0.0, "step": 74338 }, { "epoch": 1.44, "learning_rate": 2.5960897722052934e-05, "loss": 0.0, "step": 74340 }, { "epoch": 1.44, "learning_rate": 2.5960250761795703e-05, "loss": 0.0, "step": 74342 }, { "epoch": 1.44, "learning_rate": 2.5959603801538475e-05, "loss": 0.0, "step": 74344 }, { "epoch": 1.44, "learning_rate": 2.595895684128124e-05, "loss": 0.0, "step": 74346 }, { "epoch": 1.44, "learning_rate": 2.5958309881024013e-05, "loss": 0.0162, "step": 74348 }, { "epoch": 1.44, "learning_rate": 2.595766292076678e-05, "loss": 0.0, "step": 74350 }, { "epoch": 1.44, "learning_rate": 2.5957015960509545e-05, "loss": 0.0, "step": 74352 }, { "epoch": 1.44, "learning_rate": 2.5956369000252317e-05, "loss": 0.0039, "step": 74354 }, { "epoch": 1.44, "learning_rate": 2.5955722039995083e-05, "loss": 0.0, "step": 74356 }, { "epoch": 1.44, "learning_rate": 2.5955075079737852e-05, "loss": 0.0, "step": 74358 }, { "epoch": 1.44, "learning_rate": 2.5954428119480624e-05, "loss": 0.0, "step": 74360 }, { "epoch": 1.44, "learning_rate": 2.595378115922339e-05, "loss": 0.0, "step": 74362 }, { "epoch": 1.44, "learning_rate": 2.5953134198966163e-05, "loss": 0.0, "step": 74364 }, { "epoch": 1.44, "learning_rate": 2.5952487238708928e-05, "loss": 0.0, "step": 74366 }, { "epoch": 1.44, "learning_rate": 2.5951840278451694e-05, "loss": 0.0, "step": 74368 }, { "epoch": 1.44, "learning_rate": 2.5951193318194466e-05, "loss": 0.0001, "step": 74370 }, { "epoch": 1.44, "learning_rate": 2.5950546357937232e-05, "loss": 0.0, "step": 74372 }, { "epoch": 1.44, "learning_rate": 2.594989939768e-05, "loss": 0.0, "step": 74374 }, { "epoch": 1.44, "learning_rate": 2.594925243742277e-05, "loss": 0.0001, "step": 74376 }, { "epoch": 1.44, "learning_rate": 2.594860547716554e-05, "loss": 0.0, "step": 74378 }, { "epoch": 1.44, "learning_rate": 2.5947958516908305e-05, "loss": 0.0, "step": 74380 }, { "epoch": 1.44, "learning_rate": 2.5947311556651078e-05, "loss": 0.0003, "step": 74382 }, { "epoch": 1.44, "learning_rate": 2.5946664596393843e-05, "loss": 0.0, "step": 74384 }, { "epoch": 1.44, "learning_rate": 2.5946017636136616e-05, "loss": 0.0, "step": 74386 }, { "epoch": 1.44, "learning_rate": 2.594537067587938e-05, "loss": 0.0, "step": 74388 }, { "epoch": 1.44, "learning_rate": 2.594472371562215e-05, "loss": 0.0, "step": 74390 }, { "epoch": 1.44, "learning_rate": 2.594407675536492e-05, "loss": 0.0, "step": 74392 }, { "epoch": 1.44, "learning_rate": 2.594342979510769e-05, "loss": 0.0, "step": 74394 }, { "epoch": 1.44, "learning_rate": 2.5942782834850454e-05, "loss": 0.0, "step": 74396 }, { "epoch": 1.44, "learning_rate": 2.5942135874593227e-05, "loss": 0.0, "step": 74398 }, { "epoch": 1.44, "learning_rate": 2.5941488914335993e-05, "loss": 0.0002, "step": 74400 }, { "epoch": 1.44, "learning_rate": 2.5940841954078758e-05, "loss": 0.0, "step": 74402 }, { "epoch": 1.44, "learning_rate": 2.594019499382153e-05, "loss": 0.0, "step": 74404 }, { "epoch": 1.44, "learning_rate": 2.59395480335643e-05, "loss": 0.0, "step": 74406 }, { "epoch": 1.44, "learning_rate": 2.593890107330707e-05, "loss": 0.0003, "step": 74408 }, { "epoch": 1.44, "learning_rate": 2.5938254113049838e-05, "loss": 0.0, "step": 74410 }, { "epoch": 1.44, "learning_rate": 2.5937607152792604e-05, "loss": 0.0, "step": 74412 }, { "epoch": 1.44, "learning_rate": 2.5936960192535376e-05, "loss": 0.0, "step": 74414 }, { "epoch": 1.44, "learning_rate": 2.5936313232278142e-05, "loss": 0.0, "step": 74416 }, { "epoch": 1.44, "learning_rate": 2.5935666272020907e-05, "loss": 0.0, "step": 74418 }, { "epoch": 1.44, "learning_rate": 2.593501931176368e-05, "loss": 0.0006, "step": 74420 }, { "epoch": 1.44, "learning_rate": 2.593437235150645e-05, "loss": 0.0, "step": 74422 }, { "epoch": 1.44, "learning_rate": 2.5933725391249218e-05, "loss": 0.0, "step": 74424 }, { "epoch": 1.44, "learning_rate": 2.5933078430991987e-05, "loss": 0.0, "step": 74426 }, { "epoch": 1.44, "learning_rate": 2.5932431470734753e-05, "loss": 0.0, "step": 74428 }, { "epoch": 1.44, "learning_rate": 2.5931784510477525e-05, "loss": 0.0136, "step": 74430 }, { "epoch": 1.44, "learning_rate": 2.593113755022029e-05, "loss": 0.0001, "step": 74432 }, { "epoch": 1.44, "learning_rate": 2.5930490589963057e-05, "loss": 0.0024, "step": 74434 }, { "epoch": 1.44, "learning_rate": 2.592984362970583e-05, "loss": 0.0, "step": 74436 }, { "epoch": 1.44, "learning_rate": 2.5929196669448595e-05, "loss": 0.0115, "step": 74438 }, { "epoch": 1.44, "learning_rate": 2.5928549709191364e-05, "loss": 0.0107, "step": 74440 }, { "epoch": 1.44, "learning_rate": 2.5927902748934137e-05, "loss": 0.0615, "step": 74442 }, { "epoch": 1.44, "learning_rate": 2.5927255788676902e-05, "loss": 0.0, "step": 74444 }, { "epoch": 1.44, "learning_rate": 2.5926608828419675e-05, "loss": 0.0, "step": 74446 }, { "epoch": 1.44, "learning_rate": 2.592596186816244e-05, "loss": 0.0, "step": 74448 }, { "epoch": 1.44, "learning_rate": 2.5925314907905206e-05, "loss": 0.0, "step": 74450 }, { "epoch": 1.45, "learning_rate": 2.592466794764798e-05, "loss": 0.0004, "step": 74452 }, { "epoch": 1.45, "learning_rate": 2.5924020987390744e-05, "loss": 0.0, "step": 74454 }, { "epoch": 1.45, "learning_rate": 2.5923374027133513e-05, "loss": 0.011, "step": 74456 }, { "epoch": 1.45, "learning_rate": 2.5922727066876286e-05, "loss": 0.0112, "step": 74458 }, { "epoch": 1.45, "learning_rate": 2.592208010661905e-05, "loss": 0.0001, "step": 74460 }, { "epoch": 1.45, "learning_rate": 2.5921433146361817e-05, "loss": 0.0, "step": 74462 }, { "epoch": 1.45, "learning_rate": 2.592078618610459e-05, "loss": 0.0003, "step": 74464 }, { "epoch": 1.45, "learning_rate": 2.5920139225847355e-05, "loss": 0.0, "step": 74466 }, { "epoch": 1.45, "learning_rate": 2.5919492265590128e-05, "loss": 0.0, "step": 74468 }, { "epoch": 1.45, "learning_rate": 2.5918845305332894e-05, "loss": 0.0, "step": 74470 }, { "epoch": 1.45, "learning_rate": 2.5918198345075663e-05, "loss": 0.0, "step": 74472 }, { "epoch": 1.45, "learning_rate": 2.5917551384818435e-05, "loss": 0.0, "step": 74474 }, { "epoch": 1.45, "learning_rate": 2.59169044245612e-05, "loss": 0.0055, "step": 74476 }, { "epoch": 1.45, "learning_rate": 2.5916257464303966e-05, "loss": 0.0, "step": 74478 }, { "epoch": 1.45, "learning_rate": 2.591561050404674e-05, "loss": 0.0, "step": 74480 }, { "epoch": 1.45, "learning_rate": 2.5914963543789505e-05, "loss": 0.0, "step": 74482 }, { "epoch": 1.45, "learning_rate": 2.5914316583532274e-05, "loss": 0.0038, "step": 74484 }, { "epoch": 1.45, "learning_rate": 2.5913669623275043e-05, "loss": 0.0, "step": 74486 }, { "epoch": 1.45, "learning_rate": 2.5913022663017812e-05, "loss": 0.0075, "step": 74488 }, { "epoch": 1.45, "learning_rate": 2.5912375702760584e-05, "loss": 0.0, "step": 74490 }, { "epoch": 1.45, "learning_rate": 2.591172874250335e-05, "loss": 0.0, "step": 74492 }, { "epoch": 1.45, "learning_rate": 2.5911081782246116e-05, "loss": 0.0001, "step": 74494 }, { "epoch": 1.45, "learning_rate": 2.5910434821988888e-05, "loss": 0.0, "step": 74496 }, { "epoch": 1.45, "learning_rate": 2.5909787861731654e-05, "loss": 0.0042, "step": 74498 }, { "epoch": 1.45, "learning_rate": 2.590914090147442e-05, "loss": 0.0, "step": 74500 }, { "epoch": 1.45, "learning_rate": 2.5908493941217192e-05, "loss": 0.0002, "step": 74502 }, { "epoch": 1.45, "learning_rate": 2.590784698095996e-05, "loss": 0.0, "step": 74504 }, { "epoch": 1.45, "learning_rate": 2.5907200020702734e-05, "loss": 0.0009, "step": 74506 }, { "epoch": 1.45, "learning_rate": 2.59065530604455e-05, "loss": 0.0001, "step": 74508 }, { "epoch": 1.45, "learning_rate": 2.5905906100188265e-05, "loss": 0.0001, "step": 74510 }, { "epoch": 1.45, "learning_rate": 2.5905259139931038e-05, "loss": 0.0, "step": 74512 }, { "epoch": 1.45, "learning_rate": 2.5904612179673803e-05, "loss": 0.0002, "step": 74514 }, { "epoch": 1.45, "learning_rate": 2.590396521941657e-05, "loss": 0.0, "step": 74516 }, { "epoch": 1.45, "learning_rate": 2.590331825915934e-05, "loss": 0.0, "step": 74518 }, { "epoch": 1.45, "learning_rate": 2.590267129890211e-05, "loss": 0.0, "step": 74520 }, { "epoch": 1.45, "learning_rate": 2.5902024338644876e-05, "loss": 0.0, "step": 74522 }, { "epoch": 1.45, "learning_rate": 2.590137737838765e-05, "loss": 0.0, "step": 74524 }, { "epoch": 1.45, "learning_rate": 2.5900730418130414e-05, "loss": 0.0, "step": 74526 }, { "epoch": 1.45, "learning_rate": 2.5900083457873187e-05, "loss": 0.0, "step": 74528 }, { "epoch": 1.45, "learning_rate": 2.5899436497615953e-05, "loss": 0.0002, "step": 74530 }, { "epoch": 1.45, "learning_rate": 2.5898789537358718e-05, "loss": 0.0, "step": 74532 }, { "epoch": 1.45, "learning_rate": 2.589814257710149e-05, "loss": 0.0, "step": 74534 }, { "epoch": 1.45, "learning_rate": 2.589749561684426e-05, "loss": 0.0, "step": 74536 }, { "epoch": 1.45, "learning_rate": 2.5896848656587025e-05, "loss": 0.0, "step": 74538 }, { "epoch": 1.45, "learning_rate": 2.5896201696329798e-05, "loss": 0.0059, "step": 74540 }, { "epoch": 1.45, "learning_rate": 2.5895554736072564e-05, "loss": 0.0, "step": 74542 }, { "epoch": 1.45, "learning_rate": 2.589490777581533e-05, "loss": 0.0, "step": 74544 }, { "epoch": 1.45, "learning_rate": 2.5894260815558102e-05, "loss": 0.0, "step": 74546 }, { "epoch": 1.45, "learning_rate": 2.5893613855300867e-05, "loss": 0.0001, "step": 74548 }, { "epoch": 1.45, "learning_rate": 2.589296689504364e-05, "loss": 0.0, "step": 74550 }, { "epoch": 1.45, "learning_rate": 2.589231993478641e-05, "loss": 0.0058, "step": 74552 }, { "epoch": 1.45, "learning_rate": 2.5891672974529175e-05, "loss": 0.0, "step": 74554 }, { "epoch": 1.45, "learning_rate": 2.5891026014271947e-05, "loss": 0.0, "step": 74556 }, { "epoch": 1.45, "learning_rate": 2.5890379054014713e-05, "loss": 0.0001, "step": 74558 }, { "epoch": 1.45, "learning_rate": 2.588973209375748e-05, "loss": 0.0, "step": 74560 }, { "epoch": 1.45, "learning_rate": 2.588908513350025e-05, "loss": 0.0001, "step": 74562 }, { "epoch": 1.45, "learning_rate": 2.5888438173243017e-05, "loss": 0.0, "step": 74564 }, { "epoch": 1.45, "learning_rate": 2.588779121298579e-05, "loss": 0.0, "step": 74566 }, { "epoch": 1.45, "learning_rate": 2.588714425272856e-05, "loss": 0.0, "step": 74568 }, { "epoch": 1.45, "learning_rate": 2.5886497292471324e-05, "loss": 0.0001, "step": 74570 }, { "epoch": 1.45, "learning_rate": 2.5885850332214096e-05, "loss": 0.0028, "step": 74572 }, { "epoch": 1.45, "learning_rate": 2.5885203371956862e-05, "loss": 0.0, "step": 74574 }, { "epoch": 1.45, "learning_rate": 2.5884556411699628e-05, "loss": 0.0, "step": 74576 }, { "epoch": 1.45, "learning_rate": 2.58839094514424e-05, "loss": 0.0, "step": 74578 }, { "epoch": 1.45, "learning_rate": 2.5883262491185166e-05, "loss": 0.0, "step": 74580 }, { "epoch": 1.45, "learning_rate": 2.5882615530927935e-05, "loss": 0.0, "step": 74582 }, { "epoch": 1.45, "learning_rate": 2.5881968570670708e-05, "loss": 0.0, "step": 74584 }, { "epoch": 1.45, "learning_rate": 2.5881321610413473e-05, "loss": 0.0031, "step": 74586 }, { "epoch": 1.45, "learning_rate": 2.5880674650156246e-05, "loss": 0.0, "step": 74588 }, { "epoch": 1.45, "learning_rate": 2.588002768989901e-05, "loss": 0.0, "step": 74590 }, { "epoch": 1.45, "learning_rate": 2.5879380729641777e-05, "loss": 0.0, "step": 74592 }, { "epoch": 1.45, "learning_rate": 2.587873376938455e-05, "loss": 0.0, "step": 74594 }, { "epoch": 1.45, "learning_rate": 2.5878086809127315e-05, "loss": 0.0, "step": 74596 }, { "epoch": 1.45, "learning_rate": 2.5877439848870084e-05, "loss": 0.0, "step": 74598 }, { "epoch": 1.45, "learning_rate": 2.5876792888612857e-05, "loss": 0.0, "step": 74600 }, { "epoch": 1.45, "learning_rate": 2.5876145928355623e-05, "loss": 0.0001, "step": 74602 }, { "epoch": 1.45, "learning_rate": 2.5875498968098388e-05, "loss": 0.0, "step": 74604 }, { "epoch": 1.45, "learning_rate": 2.587485200784116e-05, "loss": 0.0018, "step": 74606 }, { "epoch": 1.45, "learning_rate": 2.5874205047583926e-05, "loss": 0.0, "step": 74608 }, { "epoch": 1.45, "learning_rate": 2.58735580873267e-05, "loss": 0.0, "step": 74610 }, { "epoch": 1.45, "learning_rate": 2.5872911127069465e-05, "loss": 0.0092, "step": 74612 }, { "epoch": 1.45, "learning_rate": 2.5872264166812234e-05, "loss": 0.0, "step": 74614 }, { "epoch": 1.45, "learning_rate": 2.5871617206555003e-05, "loss": 0.0, "step": 74616 }, { "epoch": 1.45, "learning_rate": 2.5870970246297772e-05, "loss": 0.0, "step": 74618 }, { "epoch": 1.45, "learning_rate": 2.5870323286040538e-05, "loss": 0.0, "step": 74620 }, { "epoch": 1.45, "learning_rate": 2.586967632578331e-05, "loss": 0.0, "step": 74622 }, { "epoch": 1.45, "learning_rate": 2.5869029365526076e-05, "loss": 0.0009, "step": 74624 }, { "epoch": 1.45, "learning_rate": 2.586838240526884e-05, "loss": 0.0089, "step": 74626 }, { "epoch": 1.45, "learning_rate": 2.5867735445011614e-05, "loss": 0.0, "step": 74628 }, { "epoch": 1.45, "learning_rate": 2.5867088484754383e-05, "loss": 0.0, "step": 74630 }, { "epoch": 1.45, "learning_rate": 2.5866441524497152e-05, "loss": 0.0, "step": 74632 }, { "epoch": 1.45, "learning_rate": 2.586579456423992e-05, "loss": 0.0, "step": 74634 }, { "epoch": 1.45, "learning_rate": 2.5865147603982687e-05, "loss": 0.0, "step": 74636 }, { "epoch": 1.45, "learning_rate": 2.586450064372546e-05, "loss": 0.0, "step": 74638 }, { "epoch": 1.45, "learning_rate": 2.5863853683468225e-05, "loss": 0.0, "step": 74640 }, { "epoch": 1.45, "learning_rate": 2.586320672321099e-05, "loss": 0.0, "step": 74642 }, { "epoch": 1.45, "learning_rate": 2.5862559762953763e-05, "loss": 0.0, "step": 74644 }, { "epoch": 1.45, "learning_rate": 2.5861912802696532e-05, "loss": 0.0001, "step": 74646 }, { "epoch": 1.45, "learning_rate": 2.58612658424393e-05, "loss": 0.0, "step": 74648 }, { "epoch": 1.45, "learning_rate": 2.586061888218207e-05, "loss": 0.0, "step": 74650 }, { "epoch": 1.45, "learning_rate": 2.5859971921924836e-05, "loss": 0.0, "step": 74652 }, { "epoch": 1.45, "learning_rate": 2.585932496166761e-05, "loss": 0.0, "step": 74654 }, { "epoch": 1.45, "learning_rate": 2.5858678001410374e-05, "loss": 0.0, "step": 74656 }, { "epoch": 1.45, "learning_rate": 2.585803104115314e-05, "loss": 0.0094, "step": 74658 }, { "epoch": 1.45, "learning_rate": 2.5857384080895912e-05, "loss": 0.0034, "step": 74660 }, { "epoch": 1.45, "learning_rate": 2.585673712063868e-05, "loss": 0.0, "step": 74662 }, { "epoch": 1.45, "learning_rate": 2.5856090160381447e-05, "loss": 0.0001, "step": 74664 }, { "epoch": 1.45, "learning_rate": 2.585544320012422e-05, "loss": 0.0006, "step": 74666 }, { "epoch": 1.45, "learning_rate": 2.5854796239866985e-05, "loss": 0.0, "step": 74668 }, { "epoch": 1.45, "learning_rate": 2.5854149279609758e-05, "loss": 0.0, "step": 74670 }, { "epoch": 1.45, "learning_rate": 2.5853502319352524e-05, "loss": 0.0, "step": 74672 }, { "epoch": 1.45, "learning_rate": 2.585285535909529e-05, "loss": 0.0, "step": 74674 }, { "epoch": 1.45, "learning_rate": 2.5852208398838062e-05, "loss": 0.0001, "step": 74676 }, { "epoch": 1.45, "learning_rate": 2.5851561438580827e-05, "loss": 0.0, "step": 74678 }, { "epoch": 1.45, "learning_rate": 2.5850914478323597e-05, "loss": 0.0, "step": 74680 }, { "epoch": 1.45, "learning_rate": 2.585026751806637e-05, "loss": 0.0, "step": 74682 }, { "epoch": 1.45, "learning_rate": 2.5849620557809135e-05, "loss": 0.0, "step": 74684 }, { "epoch": 1.45, "learning_rate": 2.58489735975519e-05, "loss": 0.0007, "step": 74686 }, { "epoch": 1.45, "learning_rate": 2.5848326637294673e-05, "loss": 0.0, "step": 74688 }, { "epoch": 1.45, "learning_rate": 2.584767967703744e-05, "loss": 0.0, "step": 74690 }, { "epoch": 1.45, "learning_rate": 2.584703271678021e-05, "loss": 0.0, "step": 74692 }, { "epoch": 1.45, "learning_rate": 2.5846385756522977e-05, "loss": 0.0, "step": 74694 }, { "epoch": 1.45, "learning_rate": 2.5845738796265746e-05, "loss": 0.0, "step": 74696 }, { "epoch": 1.45, "learning_rate": 2.584509183600852e-05, "loss": 0.0002, "step": 74698 }, { "epoch": 1.45, "learning_rate": 2.5844444875751284e-05, "loss": 0.0, "step": 74700 }, { "epoch": 1.45, "learning_rate": 2.584379791549405e-05, "loss": 0.0, "step": 74702 }, { "epoch": 1.45, "learning_rate": 2.5843150955236822e-05, "loss": 0.0, "step": 74704 }, { "epoch": 1.45, "learning_rate": 2.5842503994979588e-05, "loss": 0.0094, "step": 74706 }, { "epoch": 1.45, "learning_rate": 2.584185703472236e-05, "loss": 0.0005, "step": 74708 }, { "epoch": 1.45, "learning_rate": 2.5841210074465126e-05, "loss": 0.0, "step": 74710 }, { "epoch": 1.45, "learning_rate": 2.5840563114207895e-05, "loss": 0.0, "step": 74712 }, { "epoch": 1.45, "learning_rate": 2.5839916153950668e-05, "loss": 0.0, "step": 74714 }, { "epoch": 1.45, "learning_rate": 2.5839269193693433e-05, "loss": 0.0, "step": 74716 }, { "epoch": 1.45, "learning_rate": 2.58386222334362e-05, "loss": 0.0, "step": 74718 }, { "epoch": 1.45, "learning_rate": 2.583797527317897e-05, "loss": 0.0, "step": 74720 }, { "epoch": 1.45, "learning_rate": 2.5837328312921737e-05, "loss": 0.0002, "step": 74722 }, { "epoch": 1.45, "learning_rate": 2.5836681352664506e-05, "loss": 0.0, "step": 74724 }, { "epoch": 1.45, "learning_rate": 2.5836034392407275e-05, "loss": 0.0, "step": 74726 }, { "epoch": 1.45, "learning_rate": 2.5835387432150044e-05, "loss": 0.0005, "step": 74728 }, { "epoch": 1.45, "learning_rate": 2.5834740471892817e-05, "loss": 0.0, "step": 74730 }, { "epoch": 1.45, "learning_rate": 2.5834093511635583e-05, "loss": 0.0, "step": 74732 }, { "epoch": 1.45, "learning_rate": 2.5833446551378348e-05, "loss": 0.0001, "step": 74734 }, { "epoch": 1.45, "learning_rate": 2.583279959112112e-05, "loss": 0.0, "step": 74736 }, { "epoch": 1.45, "learning_rate": 2.5832152630863886e-05, "loss": 0.0001, "step": 74738 }, { "epoch": 1.45, "learning_rate": 2.5831505670606652e-05, "loss": 0.0, "step": 74740 }, { "epoch": 1.45, "learning_rate": 2.5830858710349425e-05, "loss": 0.0001, "step": 74742 }, { "epoch": 1.45, "learning_rate": 2.5830211750092194e-05, "loss": 0.0, "step": 74744 }, { "epoch": 1.45, "learning_rate": 2.582956478983496e-05, "loss": 0.0, "step": 74746 }, { "epoch": 1.45, "learning_rate": 2.5828917829577732e-05, "loss": 0.0, "step": 74748 }, { "epoch": 1.45, "learning_rate": 2.5828270869320498e-05, "loss": 0.0, "step": 74750 }, { "epoch": 1.45, "learning_rate": 2.582762390906327e-05, "loss": 0.0262, "step": 74752 }, { "epoch": 1.45, "learning_rate": 2.5826976948806036e-05, "loss": 0.0, "step": 74754 }, { "epoch": 1.45, "learning_rate": 2.58263299885488e-05, "loss": 0.0001, "step": 74756 }, { "epoch": 1.45, "learning_rate": 2.5825683028291574e-05, "loss": 0.0, "step": 74758 }, { "epoch": 1.45, "learning_rate": 2.5825036068034343e-05, "loss": 0.0001, "step": 74760 }, { "epoch": 1.45, "learning_rate": 2.582438910777711e-05, "loss": 0.0, "step": 74762 }, { "epoch": 1.45, "learning_rate": 2.582374214751988e-05, "loss": 0.0, "step": 74764 }, { "epoch": 1.45, "learning_rate": 2.5823095187262647e-05, "loss": 0.0, "step": 74766 }, { "epoch": 1.45, "learning_rate": 2.5822448227005413e-05, "loss": 0.0, "step": 74768 }, { "epoch": 1.45, "learning_rate": 2.5821801266748185e-05, "loss": 0.0, "step": 74770 }, { "epoch": 1.45, "learning_rate": 2.582115430649095e-05, "loss": 0.0, "step": 74772 }, { "epoch": 1.45, "learning_rate": 2.5820507346233723e-05, "loss": 0.0, "step": 74774 }, { "epoch": 1.45, "learning_rate": 2.5819860385976492e-05, "loss": 0.0, "step": 74776 }, { "epoch": 1.45, "learning_rate": 2.5819213425719258e-05, "loss": 0.0, "step": 74778 }, { "epoch": 1.45, "learning_rate": 2.581856646546203e-05, "loss": 0.0, "step": 74780 }, { "epoch": 1.45, "learning_rate": 2.5817919505204796e-05, "loss": 0.0, "step": 74782 }, { "epoch": 1.45, "learning_rate": 2.5817272544947562e-05, "loss": 0.0004, "step": 74784 }, { "epoch": 1.45, "learning_rate": 2.5816625584690334e-05, "loss": 0.0, "step": 74786 }, { "epoch": 1.45, "learning_rate": 2.58159786244331e-05, "loss": 0.0001, "step": 74788 }, { "epoch": 1.45, "learning_rate": 2.5815331664175872e-05, "loss": 0.001, "step": 74790 }, { "epoch": 1.45, "learning_rate": 2.581468470391864e-05, "loss": 0.0, "step": 74792 }, { "epoch": 1.45, "learning_rate": 2.5814037743661407e-05, "loss": 0.0, "step": 74794 }, { "epoch": 1.45, "learning_rate": 2.581339078340418e-05, "loss": 0.006, "step": 74796 }, { "epoch": 1.45, "learning_rate": 2.5812743823146945e-05, "loss": 0.0, "step": 74798 }, { "epoch": 1.45, "learning_rate": 2.581209686288971e-05, "loss": 0.0046, "step": 74800 }, { "epoch": 1.45, "learning_rate": 2.5811449902632484e-05, "loss": 0.0, "step": 74802 }, { "epoch": 1.45, "learning_rate": 2.581080294237525e-05, "loss": 0.0, "step": 74804 }, { "epoch": 1.45, "learning_rate": 2.581015598211802e-05, "loss": 0.0012, "step": 74806 }, { "epoch": 1.45, "learning_rate": 2.580950902186079e-05, "loss": 0.0047, "step": 74808 }, { "epoch": 1.45, "learning_rate": 2.5808862061603557e-05, "loss": 0.0, "step": 74810 }, { "epoch": 1.45, "learning_rate": 2.580821510134633e-05, "loss": 0.0, "step": 74812 }, { "epoch": 1.45, "learning_rate": 2.5807568141089095e-05, "loss": 0.0, "step": 74814 }, { "epoch": 1.45, "learning_rate": 2.580692118083186e-05, "loss": 0.0, "step": 74816 }, { "epoch": 1.45, "learning_rate": 2.5806274220574633e-05, "loss": 0.0, "step": 74818 }, { "epoch": 1.45, "learning_rate": 2.58056272603174e-05, "loss": 0.0, "step": 74820 }, { "epoch": 1.45, "learning_rate": 2.5804980300060168e-05, "loss": 0.0023, "step": 74822 }, { "epoch": 1.45, "learning_rate": 2.580433333980294e-05, "loss": 0.0003, "step": 74824 }, { "epoch": 1.45, "learning_rate": 2.5803686379545706e-05, "loss": 0.0, "step": 74826 }, { "epoch": 1.45, "learning_rate": 2.580303941928847e-05, "loss": 0.0, "step": 74828 }, { "epoch": 1.45, "learning_rate": 2.5802392459031244e-05, "loss": 0.0, "step": 74830 }, { "epoch": 1.45, "learning_rate": 2.580174549877401e-05, "loss": 0.0, "step": 74832 }, { "epoch": 1.45, "learning_rate": 2.5801098538516782e-05, "loss": 0.0, "step": 74834 }, { "epoch": 1.45, "learning_rate": 2.5800451578259548e-05, "loss": 0.0049, "step": 74836 }, { "epoch": 1.45, "learning_rate": 2.5799804618002317e-05, "loss": 0.0, "step": 74838 }, { "epoch": 1.45, "learning_rate": 2.579915765774509e-05, "loss": 0.0, "step": 74840 }, { "epoch": 1.45, "learning_rate": 2.5798510697487855e-05, "loss": 0.0, "step": 74842 }, { "epoch": 1.45, "learning_rate": 2.579786373723062e-05, "loss": 0.0002, "step": 74844 }, { "epoch": 1.45, "learning_rate": 2.5797216776973393e-05, "loss": 0.0, "step": 74846 }, { "epoch": 1.45, "learning_rate": 2.579656981671616e-05, "loss": 0.0, "step": 74848 }, { "epoch": 1.45, "learning_rate": 2.579592285645893e-05, "loss": 0.0, "step": 74850 }, { "epoch": 1.45, "learning_rate": 2.5795275896201697e-05, "loss": 0.0, "step": 74852 }, { "epoch": 1.45, "learning_rate": 2.5794628935944466e-05, "loss": 0.0, "step": 74854 }, { "epoch": 1.45, "learning_rate": 2.5793981975687235e-05, "loss": 0.0, "step": 74856 }, { "epoch": 1.45, "learning_rate": 2.5793335015430004e-05, "loss": 0.004, "step": 74858 }, { "epoch": 1.45, "learning_rate": 2.579268805517277e-05, "loss": 0.0002, "step": 74860 }, { "epoch": 1.45, "learning_rate": 2.5792041094915543e-05, "loss": 0.0, "step": 74862 }, { "epoch": 1.45, "learning_rate": 2.5791394134658308e-05, "loss": 0.0001, "step": 74864 }, { "epoch": 1.45, "learning_rate": 2.5790747174401074e-05, "loss": 0.0001, "step": 74866 }, { "epoch": 1.45, "learning_rate": 2.5790100214143846e-05, "loss": 0.0, "step": 74868 }, { "epoch": 1.45, "learning_rate": 2.5789453253886616e-05, "loss": 0.0, "step": 74870 }, { "epoch": 1.45, "learning_rate": 2.5788806293629385e-05, "loss": 0.0, "step": 74872 }, { "epoch": 1.45, "learning_rate": 2.5788159333372154e-05, "loss": 0.0, "step": 74874 }, { "epoch": 1.45, "learning_rate": 2.578751237311492e-05, "loss": 0.0, "step": 74876 }, { "epoch": 1.45, "learning_rate": 2.5786865412857692e-05, "loss": 0.0, "step": 74878 }, { "epoch": 1.45, "learning_rate": 2.5786218452600458e-05, "loss": 0.0, "step": 74880 }, { "epoch": 1.45, "learning_rate": 2.5785571492343223e-05, "loss": 0.0, "step": 74882 }, { "epoch": 1.45, "learning_rate": 2.5784924532085996e-05, "loss": 0.0011, "step": 74884 }, { "epoch": 1.45, "learning_rate": 2.5784277571828765e-05, "loss": 0.0, "step": 74886 }, { "epoch": 1.45, "learning_rate": 2.578363061157153e-05, "loss": 0.0131, "step": 74888 }, { "epoch": 1.45, "learning_rate": 2.5782983651314303e-05, "loss": 0.0003, "step": 74890 }, { "epoch": 1.45, "learning_rate": 2.578233669105707e-05, "loss": 0.0, "step": 74892 }, { "epoch": 1.45, "learning_rate": 2.578168973079984e-05, "loss": 0.0002, "step": 74894 }, { "epoch": 1.45, "learning_rate": 2.5781042770542607e-05, "loss": 0.0, "step": 74896 }, { "epoch": 1.45, "learning_rate": 2.5780395810285373e-05, "loss": 0.0, "step": 74898 }, { "epoch": 1.45, "learning_rate": 2.5779748850028145e-05, "loss": 0.0004, "step": 74900 }, { "epoch": 1.45, "learning_rate": 2.5779101889770914e-05, "loss": 0.0, "step": 74902 }, { "epoch": 1.45, "learning_rate": 2.577845492951368e-05, "loss": 0.0, "step": 74904 }, { "epoch": 1.45, "learning_rate": 2.5777807969256452e-05, "loss": 0.0, "step": 74906 }, { "epoch": 1.45, "learning_rate": 2.5777161008999218e-05, "loss": 0.0, "step": 74908 }, { "epoch": 1.45, "learning_rate": 2.5776514048741984e-05, "loss": 0.0, "step": 74910 }, { "epoch": 1.45, "learning_rate": 2.5775867088484756e-05, "loss": 0.0, "step": 74912 }, { "epoch": 1.45, "learning_rate": 2.5775220128227522e-05, "loss": 0.0, "step": 74914 }, { "epoch": 1.45, "learning_rate": 2.5774573167970294e-05, "loss": 0.0, "step": 74916 }, { "epoch": 1.45, "learning_rate": 2.577392620771306e-05, "loss": 0.0, "step": 74918 }, { "epoch": 1.45, "learning_rate": 2.577327924745583e-05, "loss": 0.0, "step": 74920 }, { "epoch": 1.45, "learning_rate": 2.57726322871986e-05, "loss": 0.0, "step": 74922 }, { "epoch": 1.45, "learning_rate": 2.5771985326941367e-05, "loss": 0.0, "step": 74924 }, { "epoch": 1.45, "learning_rate": 2.5771338366684133e-05, "loss": 0.0, "step": 74926 }, { "epoch": 1.45, "learning_rate": 2.5770691406426905e-05, "loss": 0.0, "step": 74928 }, { "epoch": 1.45, "learning_rate": 2.577004444616967e-05, "loss": 0.0, "step": 74930 }, { "epoch": 1.45, "learning_rate": 2.5769397485912444e-05, "loss": 0.0, "step": 74932 }, { "epoch": 1.45, "learning_rate": 2.576875052565521e-05, "loss": 0.0, "step": 74934 }, { "epoch": 1.45, "learning_rate": 2.576810356539798e-05, "loss": 0.0, "step": 74936 }, { "epoch": 1.45, "learning_rate": 2.576745660514075e-05, "loss": 0.0, "step": 74938 }, { "epoch": 1.45, "learning_rate": 2.5766809644883517e-05, "loss": 0.0098, "step": 74940 }, { "epoch": 1.45, "learning_rate": 2.5766162684626282e-05, "loss": 0.0, "step": 74942 }, { "epoch": 1.45, "learning_rate": 2.5765515724369055e-05, "loss": 0.0, "step": 74944 }, { "epoch": 1.45, "learning_rate": 2.576486876411182e-05, "loss": 0.0, "step": 74946 }, { "epoch": 1.45, "learning_rate": 2.576422180385459e-05, "loss": 0.0, "step": 74948 }, { "epoch": 1.45, "learning_rate": 2.576357484359736e-05, "loss": 0.0, "step": 74950 }, { "epoch": 1.45, "learning_rate": 2.5762927883340128e-05, "loss": 0.0, "step": 74952 }, { "epoch": 1.45, "learning_rate": 2.57622809230829e-05, "loss": 0.0, "step": 74954 }, { "epoch": 1.45, "learning_rate": 2.5761633962825666e-05, "loss": 0.0004, "step": 74956 }, { "epoch": 1.45, "learning_rate": 2.576098700256843e-05, "loss": 0.0, "step": 74958 }, { "epoch": 1.45, "learning_rate": 2.5760340042311204e-05, "loss": 0.0, "step": 74960 }, { "epoch": 1.45, "learning_rate": 2.575969308205397e-05, "loss": 0.0, "step": 74962 }, { "epoch": 1.45, "learning_rate": 2.575904612179674e-05, "loss": 0.0, "step": 74964 }, { "epoch": 1.46, "learning_rate": 2.5758399161539508e-05, "loss": 0.0, "step": 74966 }, { "epoch": 1.46, "learning_rate": 2.5757752201282277e-05, "loss": 0.0, "step": 74968 }, { "epoch": 1.46, "learning_rate": 2.5757105241025043e-05, "loss": 0.0, "step": 74970 }, { "epoch": 1.46, "learning_rate": 2.5756458280767815e-05, "loss": 0.0, "step": 74972 }, { "epoch": 1.46, "learning_rate": 2.575581132051058e-05, "loss": 0.0058, "step": 74974 }, { "epoch": 1.46, "learning_rate": 2.5755164360253353e-05, "loss": 0.0, "step": 74976 }, { "epoch": 1.46, "learning_rate": 2.575451739999612e-05, "loss": 0.0, "step": 74978 }, { "epoch": 1.46, "learning_rate": 2.5753870439738885e-05, "loss": 0.0, "step": 74980 }, { "epoch": 1.46, "learning_rate": 2.5753223479481657e-05, "loss": 0.0, "step": 74982 }, { "epoch": 1.46, "learning_rate": 2.5752576519224426e-05, "loss": 0.0, "step": 74984 }, { "epoch": 1.46, "learning_rate": 2.5751929558967192e-05, "loss": 0.0, "step": 74986 }, { "epoch": 1.46, "learning_rate": 2.5751282598709964e-05, "loss": 0.0, "step": 74988 }, { "epoch": 1.46, "learning_rate": 2.575063563845273e-05, "loss": 0.0002, "step": 74990 }, { "epoch": 1.46, "learning_rate": 2.5749988678195503e-05, "loss": 0.0, "step": 74992 }, { "epoch": 1.46, "learning_rate": 2.5749341717938268e-05, "loss": 0.0001, "step": 74994 }, { "epoch": 1.46, "learning_rate": 2.5748694757681034e-05, "loss": 0.0, "step": 74996 }, { "epoch": 1.46, "learning_rate": 2.5748047797423806e-05, "loss": 0.0, "step": 74998 }, { "epoch": 1.46, "learning_rate": 2.5747400837166576e-05, "loss": 0.0, "step": 75000 }, { "epoch": 1.46, "learning_rate": 2.574675387690934e-05, "loss": 0.0068, "step": 75002 }, { "epoch": 1.46, "learning_rate": 2.5746106916652114e-05, "loss": 0.0, "step": 75004 }, { "epoch": 1.46, "learning_rate": 2.574545995639488e-05, "loss": 0.0026, "step": 75006 }, { "epoch": 1.46, "learning_rate": 2.5744812996137645e-05, "loss": 0.0172, "step": 75008 }, { "epoch": 1.46, "learning_rate": 2.5744166035880418e-05, "loss": 0.0, "step": 75010 }, { "epoch": 1.46, "learning_rate": 2.5743519075623183e-05, "loss": 0.0001, "step": 75012 }, { "epoch": 1.46, "learning_rate": 2.5742872115365956e-05, "loss": 0.0, "step": 75014 }, { "epoch": 1.46, "learning_rate": 2.5742225155108725e-05, "loss": 0.0, "step": 75016 }, { "epoch": 1.46, "learning_rate": 2.574157819485149e-05, "loss": 0.0, "step": 75018 }, { "epoch": 1.46, "learning_rate": 2.5740931234594263e-05, "loss": 0.0047, "step": 75020 }, { "epoch": 1.46, "learning_rate": 2.574028427433703e-05, "loss": 0.0, "step": 75022 }, { "epoch": 1.46, "learning_rate": 2.5739637314079794e-05, "loss": 0.0, "step": 75024 }, { "epoch": 1.46, "learning_rate": 2.5738990353822567e-05, "loss": 0.0, "step": 75026 }, { "epoch": 1.46, "learning_rate": 2.5738343393565333e-05, "loss": 0.0, "step": 75028 }, { "epoch": 1.46, "learning_rate": 2.57376964333081e-05, "loss": 0.0086, "step": 75030 }, { "epoch": 1.46, "learning_rate": 2.5737049473050874e-05, "loss": 0.0, "step": 75032 }, { "epoch": 1.46, "learning_rate": 2.573640251279364e-05, "loss": 0.0001, "step": 75034 }, { "epoch": 1.46, "learning_rate": 2.5735755552536412e-05, "loss": 0.0, "step": 75036 }, { "epoch": 1.46, "learning_rate": 2.5735108592279178e-05, "loss": 0.0, "step": 75038 }, { "epoch": 1.46, "learning_rate": 2.5734461632021944e-05, "loss": 0.0, "step": 75040 }, { "epoch": 1.46, "learning_rate": 2.5733814671764716e-05, "loss": 0.0034, "step": 75042 }, { "epoch": 1.46, "learning_rate": 2.5733167711507482e-05, "loss": 0.0, "step": 75044 }, { "epoch": 1.46, "learning_rate": 2.573252075125025e-05, "loss": 0.0, "step": 75046 }, { "epoch": 1.46, "learning_rate": 2.5731873790993023e-05, "loss": 0.0001, "step": 75048 }, { "epoch": 1.46, "learning_rate": 2.573122683073579e-05, "loss": 0.0, "step": 75050 }, { "epoch": 1.46, "learning_rate": 2.5730579870478555e-05, "loss": 0.0, "step": 75052 }, { "epoch": 1.46, "learning_rate": 2.5729932910221327e-05, "loss": 0.0, "step": 75054 }, { "epoch": 1.46, "learning_rate": 2.5729285949964093e-05, "loss": 0.0, "step": 75056 }, { "epoch": 1.46, "learning_rate": 2.5728638989706865e-05, "loss": 0.0, "step": 75058 }, { "epoch": 1.46, "learning_rate": 2.572799202944963e-05, "loss": 0.0, "step": 75060 }, { "epoch": 1.46, "learning_rate": 2.57273450691924e-05, "loss": 0.0032, "step": 75062 }, { "epoch": 1.46, "learning_rate": 2.5726698108935173e-05, "loss": 0.0001, "step": 75064 }, { "epoch": 1.46, "learning_rate": 2.572605114867794e-05, "loss": 0.0, "step": 75066 }, { "epoch": 1.46, "learning_rate": 2.5725404188420704e-05, "loss": 0.0001, "step": 75068 }, { "epoch": 1.46, "learning_rate": 2.5724757228163477e-05, "loss": 0.0, "step": 75070 }, { "epoch": 1.46, "learning_rate": 2.5724110267906242e-05, "loss": 0.0, "step": 75072 }, { "epoch": 1.46, "learning_rate": 2.5723463307649015e-05, "loss": 0.0, "step": 75074 }, { "epoch": 1.46, "learning_rate": 2.572281634739178e-05, "loss": 0.0, "step": 75076 }, { "epoch": 1.46, "learning_rate": 2.572216938713455e-05, "loss": 0.0, "step": 75078 }, { "epoch": 1.46, "learning_rate": 2.5721522426877322e-05, "loss": 0.0042, "step": 75080 }, { "epoch": 1.46, "learning_rate": 2.5720875466620088e-05, "loss": 0.0, "step": 75082 }, { "epoch": 1.46, "learning_rate": 2.5720228506362853e-05, "loss": 0.0028, "step": 75084 }, { "epoch": 1.46, "learning_rate": 2.5719581546105626e-05, "loss": 0.0009, "step": 75086 }, { "epoch": 1.46, "learning_rate": 2.571893458584839e-05, "loss": 0.0, "step": 75088 }, { "epoch": 1.46, "learning_rate": 2.5718287625591157e-05, "loss": 0.0023, "step": 75090 }, { "epoch": 1.46, "learning_rate": 2.571764066533393e-05, "loss": 0.0, "step": 75092 }, { "epoch": 1.46, "learning_rate": 2.57169937050767e-05, "loss": 0.0, "step": 75094 }, { "epoch": 1.46, "learning_rate": 2.5716346744819468e-05, "loss": 0.0058, "step": 75096 }, { "epoch": 1.46, "learning_rate": 2.5715699784562237e-05, "loss": 0.0, "step": 75098 }, { "epoch": 1.46, "learning_rate": 2.5715052824305003e-05, "loss": 0.0, "step": 75100 }, { "epoch": 1.46, "learning_rate": 2.5714405864047775e-05, "loss": 0.0005, "step": 75102 }, { "epoch": 1.46, "learning_rate": 2.571375890379054e-05, "loss": 0.0, "step": 75104 }, { "epoch": 1.46, "learning_rate": 2.5713111943533307e-05, "loss": 0.0051, "step": 75106 }, { "epoch": 1.46, "learning_rate": 2.571246498327608e-05, "loss": 0.0, "step": 75108 }, { "epoch": 1.46, "learning_rate": 2.5711818023018848e-05, "loss": 0.0002, "step": 75110 }, { "epoch": 1.46, "learning_rate": 2.5711171062761614e-05, "loss": 0.0, "step": 75112 }, { "epoch": 1.46, "learning_rate": 2.5710524102504386e-05, "loss": 0.0, "step": 75114 }, { "epoch": 1.46, "learning_rate": 2.5709877142247152e-05, "loss": 0.0, "step": 75116 }, { "epoch": 1.46, "learning_rate": 2.5709230181989924e-05, "loss": 0.0, "step": 75118 }, { "epoch": 1.46, "learning_rate": 2.570858322173269e-05, "loss": 0.001, "step": 75120 }, { "epoch": 1.46, "learning_rate": 2.5707936261475456e-05, "loss": 0.0209, "step": 75122 }, { "epoch": 1.46, "learning_rate": 2.5707289301218228e-05, "loss": 0.0, "step": 75124 }, { "epoch": 1.46, "learning_rate": 2.5706642340960997e-05, "loss": 0.0, "step": 75126 }, { "epoch": 1.46, "learning_rate": 2.5705995380703763e-05, "loss": 0.0, "step": 75128 }, { "epoch": 1.46, "learning_rate": 2.5705348420446536e-05, "loss": 0.0, "step": 75130 }, { "epoch": 1.46, "learning_rate": 2.57047014601893e-05, "loss": 0.0001, "step": 75132 }, { "epoch": 1.46, "learning_rate": 2.5704054499932074e-05, "loss": 0.0, "step": 75134 }, { "epoch": 1.46, "learning_rate": 2.570340753967484e-05, "loss": 0.006, "step": 75136 }, { "epoch": 1.46, "learning_rate": 2.5702760579417605e-05, "loss": 0.0006, "step": 75138 }, { "epoch": 1.46, "learning_rate": 2.5702113619160378e-05, "loss": 0.0, "step": 75140 }, { "epoch": 1.46, "learning_rate": 2.5701466658903147e-05, "loss": 0.0001, "step": 75142 }, { "epoch": 1.46, "learning_rate": 2.5700819698645912e-05, "loss": 0.0, "step": 75144 }, { "epoch": 1.46, "learning_rate": 2.5700172738388685e-05, "loss": 0.0, "step": 75146 }, { "epoch": 1.46, "learning_rate": 2.569952577813145e-05, "loss": 0.0001, "step": 75148 }, { "epoch": 1.46, "learning_rate": 2.5698878817874216e-05, "loss": 0.0, "step": 75150 }, { "epoch": 1.46, "learning_rate": 2.569823185761699e-05, "loss": 0.0, "step": 75152 }, { "epoch": 1.46, "learning_rate": 2.5697584897359754e-05, "loss": 0.0003, "step": 75154 }, { "epoch": 1.46, "learning_rate": 2.5696937937102527e-05, "loss": 0.0, "step": 75156 }, { "epoch": 1.46, "learning_rate": 2.5696290976845293e-05, "loss": 0.0, "step": 75158 }, { "epoch": 1.46, "learning_rate": 2.569564401658806e-05, "loss": 0.0003, "step": 75160 }, { "epoch": 1.46, "learning_rate": 2.5694997056330834e-05, "loss": 0.0, "step": 75162 }, { "epoch": 1.46, "learning_rate": 2.56943500960736e-05, "loss": 0.0, "step": 75164 }, { "epoch": 1.46, "learning_rate": 2.5693703135816365e-05, "loss": 0.0, "step": 75166 }, { "epoch": 1.46, "learning_rate": 2.5693056175559138e-05, "loss": 0.0001, "step": 75168 }, { "epoch": 1.46, "learning_rate": 2.5692409215301904e-05, "loss": 0.0001, "step": 75170 }, { "epoch": 1.46, "learning_rate": 2.5691762255044673e-05, "loss": 0.0006, "step": 75172 }, { "epoch": 1.46, "learning_rate": 2.5691115294787442e-05, "loss": 0.0, "step": 75174 }, { "epoch": 1.46, "learning_rate": 2.569046833453021e-05, "loss": 0.0, "step": 75176 }, { "epoch": 1.46, "learning_rate": 2.5689821374272983e-05, "loss": 0.0103, "step": 75178 }, { "epoch": 1.46, "learning_rate": 2.568917441401575e-05, "loss": 0.0, "step": 75180 }, { "epoch": 1.46, "learning_rate": 2.5688527453758515e-05, "loss": 0.0047, "step": 75182 }, { "epoch": 1.46, "learning_rate": 2.5687880493501287e-05, "loss": 0.0, "step": 75184 }, { "epoch": 1.46, "learning_rate": 2.5687233533244053e-05, "loss": 0.0, "step": 75186 }, { "epoch": 1.46, "learning_rate": 2.5686586572986822e-05, "loss": 0.0038, "step": 75188 }, { "epoch": 1.46, "learning_rate": 2.568593961272959e-05, "loss": 0.0, "step": 75190 }, { "epoch": 1.46, "learning_rate": 2.568529265247236e-05, "loss": 0.0, "step": 75192 }, { "epoch": 1.46, "learning_rate": 2.5684645692215126e-05, "loss": 0.0, "step": 75194 }, { "epoch": 1.46, "learning_rate": 2.56839987319579e-05, "loss": 0.0013, "step": 75196 }, { "epoch": 1.46, "learning_rate": 2.5683351771700664e-05, "loss": 0.0, "step": 75198 }, { "epoch": 1.46, "learning_rate": 2.5682704811443437e-05, "loss": 0.0009, "step": 75200 }, { "epoch": 1.46, "learning_rate": 2.5682057851186202e-05, "loss": 0.0, "step": 75202 }, { "epoch": 1.46, "learning_rate": 2.568141089092897e-05, "loss": 0.0, "step": 75204 }, { "epoch": 1.46, "learning_rate": 2.568076393067174e-05, "loss": 0.0, "step": 75206 }, { "epoch": 1.46, "learning_rate": 2.568011697041451e-05, "loss": 0.0, "step": 75208 }, { "epoch": 1.46, "learning_rate": 2.5679470010157275e-05, "loss": 0.0005, "step": 75210 }, { "epoch": 1.46, "learning_rate": 2.5678823049900048e-05, "loss": 0.0, "step": 75212 }, { "epoch": 1.46, "learning_rate": 2.5678176089642813e-05, "loss": 0.0, "step": 75214 }, { "epoch": 1.46, "learning_rate": 2.5677529129385586e-05, "loss": 0.0, "step": 75216 }, { "epoch": 1.46, "learning_rate": 2.567688216912835e-05, "loss": 0.0, "step": 75218 }, { "epoch": 1.46, "learning_rate": 2.5676235208871117e-05, "loss": 0.0, "step": 75220 }, { "epoch": 1.46, "learning_rate": 2.567558824861389e-05, "loss": 0.0001, "step": 75222 }, { "epoch": 1.46, "learning_rate": 2.567494128835666e-05, "loss": 0.0, "step": 75224 }, { "epoch": 1.46, "learning_rate": 2.5674294328099424e-05, "loss": 0.0035, "step": 75226 }, { "epoch": 1.46, "learning_rate": 2.5673647367842197e-05, "loss": 0.0, "step": 75228 }, { "epoch": 1.46, "learning_rate": 2.5673000407584963e-05, "loss": 0.0073, "step": 75230 }, { "epoch": 1.46, "learning_rate": 2.567235344732773e-05, "loss": 0.0, "step": 75232 }, { "epoch": 1.46, "learning_rate": 2.56717064870705e-05, "loss": 0.0, "step": 75234 }, { "epoch": 1.46, "learning_rate": 2.5671059526813266e-05, "loss": 0.0, "step": 75236 }, { "epoch": 1.46, "learning_rate": 2.567041256655604e-05, "loss": 0.0, "step": 75238 }, { "epoch": 1.46, "learning_rate": 2.5669765606298808e-05, "loss": 0.0, "step": 75240 }, { "epoch": 1.46, "learning_rate": 2.5669118646041574e-05, "loss": 0.0, "step": 75242 }, { "epoch": 1.46, "learning_rate": 2.5668471685784346e-05, "loss": 0.0001, "step": 75244 }, { "epoch": 1.46, "learning_rate": 2.5667824725527112e-05, "loss": 0.0, "step": 75246 }, { "epoch": 1.46, "learning_rate": 2.5667177765269878e-05, "loss": 0.0, "step": 75248 }, { "epoch": 1.46, "learning_rate": 2.566653080501265e-05, "loss": 0.0001, "step": 75250 }, { "epoch": 1.46, "learning_rate": 2.5665883844755416e-05, "loss": 0.0, "step": 75252 }, { "epoch": 1.46, "learning_rate": 2.5665236884498185e-05, "loss": 0.0, "step": 75254 }, { "epoch": 1.46, "learning_rate": 2.5664589924240957e-05, "loss": 0.0, "step": 75256 }, { "epoch": 1.46, "learning_rate": 2.5663942963983723e-05, "loss": 0.0, "step": 75258 }, { "epoch": 1.46, "learning_rate": 2.5663296003726496e-05, "loss": 0.0, "step": 75260 }, { "epoch": 1.46, "learning_rate": 2.566264904346926e-05, "loss": 0.0, "step": 75262 }, { "epoch": 1.46, "learning_rate": 2.5662002083212027e-05, "loss": 0.0, "step": 75264 }, { "epoch": 1.46, "learning_rate": 2.56613551229548e-05, "loss": 0.0, "step": 75266 }, { "epoch": 1.46, "learning_rate": 2.5660708162697565e-05, "loss": 0.0001, "step": 75268 }, { "epoch": 1.46, "learning_rate": 2.5660061202440334e-05, "loss": 0.0001, "step": 75270 }, { "epoch": 1.46, "learning_rate": 2.5659414242183107e-05, "loss": 0.0002, "step": 75272 }, { "epoch": 1.46, "learning_rate": 2.5658767281925872e-05, "loss": 0.0024, "step": 75274 }, { "epoch": 1.46, "learning_rate": 2.5658120321668645e-05, "loss": 0.0001, "step": 75276 }, { "epoch": 1.46, "learning_rate": 2.565747336141141e-05, "loss": 0.0, "step": 75278 }, { "epoch": 1.46, "learning_rate": 2.5656826401154176e-05, "loss": 0.0, "step": 75280 }, { "epoch": 1.46, "learning_rate": 2.565617944089695e-05, "loss": 0.0, "step": 75282 }, { "epoch": 1.46, "learning_rate": 2.5655532480639714e-05, "loss": 0.0, "step": 75284 }, { "epoch": 1.46, "learning_rate": 2.5654885520382483e-05, "loss": 0.0, "step": 75286 }, { "epoch": 1.46, "learning_rate": 2.5654238560125256e-05, "loss": 0.0, "step": 75288 }, { "epoch": 1.46, "learning_rate": 2.565359159986802e-05, "loss": 0.0, "step": 75290 }, { "epoch": 1.46, "learning_rate": 2.5652944639610787e-05, "loss": 0.001, "step": 75292 }, { "epoch": 1.46, "learning_rate": 2.565229767935356e-05, "loss": 0.0002, "step": 75294 }, { "epoch": 1.46, "learning_rate": 2.5651650719096325e-05, "loss": 0.0002, "step": 75296 }, { "epoch": 1.46, "learning_rate": 2.5651003758839098e-05, "loss": 0.0, "step": 75298 }, { "epoch": 1.46, "learning_rate": 2.5650356798581864e-05, "loss": 0.0, "step": 75300 }, { "epoch": 1.46, "learning_rate": 2.5649709838324633e-05, "loss": 0.0, "step": 75302 }, { "epoch": 1.46, "learning_rate": 2.5649062878067405e-05, "loss": 0.0, "step": 75304 }, { "epoch": 1.46, "learning_rate": 2.564841591781017e-05, "loss": 0.0002, "step": 75306 }, { "epoch": 1.46, "learning_rate": 2.5647768957552937e-05, "loss": 0.0, "step": 75308 }, { "epoch": 1.46, "learning_rate": 2.564712199729571e-05, "loss": 0.0, "step": 75310 }, { "epoch": 1.46, "learning_rate": 2.5646475037038475e-05, "loss": 0.0011, "step": 75312 }, { "epoch": 1.46, "learning_rate": 2.564582807678124e-05, "loss": 0.0, "step": 75314 }, { "epoch": 1.46, "learning_rate": 2.5645181116524013e-05, "loss": 0.0001, "step": 75316 }, { "epoch": 1.46, "learning_rate": 2.5644534156266782e-05, "loss": 0.0, "step": 75318 }, { "epoch": 1.46, "learning_rate": 2.5643887196009554e-05, "loss": 0.0, "step": 75320 }, { "epoch": 1.46, "learning_rate": 2.564324023575232e-05, "loss": 0.0, "step": 75322 }, { "epoch": 1.46, "learning_rate": 2.5642593275495086e-05, "loss": 0.0, "step": 75324 }, { "epoch": 1.46, "learning_rate": 2.564194631523786e-05, "loss": 0.0, "step": 75326 }, { "epoch": 1.46, "learning_rate": 2.5641299354980624e-05, "loss": 0.0, "step": 75328 }, { "epoch": 1.46, "learning_rate": 2.564065239472339e-05, "loss": 0.0001, "step": 75330 }, { "epoch": 1.46, "learning_rate": 2.5640005434466162e-05, "loss": 0.0, "step": 75332 }, { "epoch": 1.46, "learning_rate": 2.563935847420893e-05, "loss": 0.0, "step": 75334 }, { "epoch": 1.46, "learning_rate": 2.5638711513951697e-05, "loss": 0.0083, "step": 75336 }, { "epoch": 1.46, "learning_rate": 2.563806455369447e-05, "loss": 0.0, "step": 75338 }, { "epoch": 1.46, "learning_rate": 2.5637417593437235e-05, "loss": 0.0, "step": 75340 }, { "epoch": 1.46, "learning_rate": 2.5636770633180008e-05, "loss": 0.0, "step": 75342 }, { "epoch": 1.46, "learning_rate": 2.5636123672922773e-05, "loss": 0.0, "step": 75344 }, { "epoch": 1.46, "learning_rate": 2.563547671266554e-05, "loss": 0.0005, "step": 75346 }, { "epoch": 1.46, "learning_rate": 2.563482975240831e-05, "loss": 0.0, "step": 75348 }, { "epoch": 1.46, "learning_rate": 2.563418279215108e-05, "loss": 0.0019, "step": 75350 }, { "epoch": 1.46, "learning_rate": 2.5633535831893846e-05, "loss": 0.0, "step": 75352 }, { "epoch": 1.46, "learning_rate": 2.563288887163662e-05, "loss": 0.0001, "step": 75354 }, { "epoch": 1.46, "learning_rate": 2.5632241911379384e-05, "loss": 0.0, "step": 75356 }, { "epoch": 1.46, "learning_rate": 2.5631594951122157e-05, "loss": 0.0, "step": 75358 }, { "epoch": 1.46, "learning_rate": 2.5630947990864923e-05, "loss": 0.0, "step": 75360 }, { "epoch": 1.46, "learning_rate": 2.563030103060769e-05, "loss": 0.0, "step": 75362 }, { "epoch": 1.46, "learning_rate": 2.562965407035046e-05, "loss": 0.0, "step": 75364 }, { "epoch": 1.46, "learning_rate": 2.562900711009323e-05, "loss": 0.0, "step": 75366 }, { "epoch": 1.46, "learning_rate": 2.5628360149835996e-05, "loss": 0.0005, "step": 75368 }, { "epoch": 1.46, "learning_rate": 2.5627713189578768e-05, "loss": 0.0, "step": 75370 }, { "epoch": 1.46, "learning_rate": 2.5627066229321534e-05, "loss": 0.0, "step": 75372 }, { "epoch": 1.46, "learning_rate": 2.56264192690643e-05, "loss": 0.0, "step": 75374 }, { "epoch": 1.46, "learning_rate": 2.5625772308807072e-05, "loss": 0.0001, "step": 75376 }, { "epoch": 1.46, "learning_rate": 2.5625125348549838e-05, "loss": 0.0006, "step": 75378 }, { "epoch": 1.46, "learning_rate": 2.562447838829261e-05, "loss": 0.0, "step": 75380 }, { "epoch": 1.46, "learning_rate": 2.562383142803538e-05, "loss": 0.0, "step": 75382 }, { "epoch": 1.46, "learning_rate": 2.5623184467778145e-05, "loss": 0.0, "step": 75384 }, { "epoch": 1.46, "learning_rate": 2.5622537507520917e-05, "loss": 0.0112, "step": 75386 }, { "epoch": 1.46, "learning_rate": 2.5621890547263683e-05, "loss": 0.0019, "step": 75388 }, { "epoch": 1.46, "learning_rate": 2.562124358700645e-05, "loss": 0.0, "step": 75390 }, { "epoch": 1.46, "learning_rate": 2.562059662674922e-05, "loss": 0.0, "step": 75392 }, { "epoch": 1.46, "learning_rate": 2.5619949666491987e-05, "loss": 0.0002, "step": 75394 }, { "epoch": 1.46, "learning_rate": 2.5619302706234756e-05, "loss": 0.0, "step": 75396 }, { "epoch": 1.46, "learning_rate": 2.5618655745977525e-05, "loss": 0.0, "step": 75398 }, { "epoch": 1.46, "learning_rate": 2.5618008785720294e-05, "loss": 0.0001, "step": 75400 }, { "epoch": 1.46, "learning_rate": 2.5617361825463067e-05, "loss": 0.0, "step": 75402 }, { "epoch": 1.46, "learning_rate": 2.5616714865205832e-05, "loss": 0.0, "step": 75404 }, { "epoch": 1.46, "learning_rate": 2.5616067904948598e-05, "loss": 0.0001, "step": 75406 }, { "epoch": 1.46, "learning_rate": 2.561542094469137e-05, "loss": 0.0, "step": 75408 }, { "epoch": 1.46, "learning_rate": 2.5614773984434136e-05, "loss": 0.0007, "step": 75410 }, { "epoch": 1.46, "learning_rate": 2.5614127024176905e-05, "loss": 0.0, "step": 75412 }, { "epoch": 1.46, "learning_rate": 2.5613480063919674e-05, "loss": 0.0, "step": 75414 }, { "epoch": 1.46, "learning_rate": 2.5612833103662443e-05, "loss": 0.0, "step": 75416 }, { "epoch": 1.46, "learning_rate": 2.5612186143405216e-05, "loss": 0.0, "step": 75418 }, { "epoch": 1.46, "learning_rate": 2.561153918314798e-05, "loss": 0.0, "step": 75420 }, { "epoch": 1.46, "learning_rate": 2.5610892222890747e-05, "loss": 0.0, "step": 75422 }, { "epoch": 1.46, "learning_rate": 2.561024526263352e-05, "loss": 0.0, "step": 75424 }, { "epoch": 1.46, "learning_rate": 2.5609598302376285e-05, "loss": 0.0, "step": 75426 }, { "epoch": 1.46, "learning_rate": 2.5608951342119055e-05, "loss": 0.0002, "step": 75428 }, { "epoch": 1.46, "learning_rate": 2.5608304381861824e-05, "loss": 0.0001, "step": 75430 }, { "epoch": 1.46, "learning_rate": 2.5607657421604593e-05, "loss": 0.0, "step": 75432 }, { "epoch": 1.46, "learning_rate": 2.560701046134736e-05, "loss": 0.0, "step": 75434 }, { "epoch": 1.46, "learning_rate": 2.560636350109013e-05, "loss": 0.0, "step": 75436 }, { "epoch": 1.46, "learning_rate": 2.5605716540832897e-05, "loss": 0.0008, "step": 75438 }, { "epoch": 1.46, "learning_rate": 2.560506958057567e-05, "loss": 0.0, "step": 75440 }, { "epoch": 1.46, "learning_rate": 2.5604422620318435e-05, "loss": 0.0, "step": 75442 }, { "epoch": 1.46, "learning_rate": 2.5603775660061204e-05, "loss": 0.0, "step": 75444 }, { "epoch": 1.46, "learning_rate": 2.5603128699803973e-05, "loss": 0.0, "step": 75446 }, { "epoch": 1.46, "learning_rate": 2.5602481739546742e-05, "loss": 0.0067, "step": 75448 }, { "epoch": 1.46, "learning_rate": 2.5601834779289508e-05, "loss": 0.0, "step": 75450 }, { "epoch": 1.46, "learning_rate": 2.560118781903228e-05, "loss": 0.0, "step": 75452 }, { "epoch": 1.46, "learning_rate": 2.5600540858775046e-05, "loss": 0.0, "step": 75454 }, { "epoch": 1.46, "learning_rate": 2.559989389851781e-05, "loss": 0.0, "step": 75456 }, { "epoch": 1.46, "learning_rate": 2.5599246938260584e-05, "loss": 0.0002, "step": 75458 }, { "epoch": 1.46, "learning_rate": 2.559859997800335e-05, "loss": 0.0001, "step": 75460 }, { "epoch": 1.46, "learning_rate": 2.5597953017746122e-05, "loss": 0.0, "step": 75462 }, { "epoch": 1.46, "learning_rate": 2.559730605748889e-05, "loss": 0.0, "step": 75464 }, { "epoch": 1.46, "learning_rate": 2.5596659097231657e-05, "loss": 0.0, "step": 75466 }, { "epoch": 1.46, "learning_rate": 2.559601213697443e-05, "loss": 0.0014, "step": 75468 }, { "epoch": 1.46, "learning_rate": 2.5595365176717195e-05, "loss": 0.009, "step": 75470 }, { "epoch": 1.46, "learning_rate": 2.559471821645996e-05, "loss": 0.0, "step": 75472 }, { "epoch": 1.46, "learning_rate": 2.5594071256202733e-05, "loss": 0.0, "step": 75474 }, { "epoch": 1.46, "learning_rate": 2.55934242959455e-05, "loss": 0.0, "step": 75476 }, { "epoch": 1.46, "learning_rate": 2.5592777335688268e-05, "loss": 0.0, "step": 75478 }, { "epoch": 1.46, "learning_rate": 2.559213037543104e-05, "loss": 0.0003, "step": 75480 }, { "epoch": 1.47, "learning_rate": 2.5591483415173806e-05, "loss": 0.0, "step": 75482 }, { "epoch": 1.47, "learning_rate": 2.559083645491658e-05, "loss": 0.0005, "step": 75484 }, { "epoch": 1.47, "learning_rate": 2.5590189494659344e-05, "loss": 0.0, "step": 75486 }, { "epoch": 1.47, "learning_rate": 2.558954253440211e-05, "loss": 0.0, "step": 75488 }, { "epoch": 1.47, "learning_rate": 2.5588895574144883e-05, "loss": 0.0154, "step": 75490 }, { "epoch": 1.47, "learning_rate": 2.5588248613887648e-05, "loss": 0.0004, "step": 75492 }, { "epoch": 1.47, "learning_rate": 2.5587601653630417e-05, "loss": 0.0037, "step": 75494 }, { "epoch": 1.47, "learning_rate": 2.558695469337319e-05, "loss": 0.0, "step": 75496 }, { "epoch": 1.47, "learning_rate": 2.5586307733115956e-05, "loss": 0.0, "step": 75498 }, { "epoch": 1.47, "learning_rate": 2.5585660772858728e-05, "loss": 0.0, "step": 75500 }, { "epoch": 1.47, "learning_rate": 2.5585013812601494e-05, "loss": 0.0, "step": 75502 }, { "epoch": 1.47, "learning_rate": 2.558436685234426e-05, "loss": 0.0, "step": 75504 }, { "epoch": 1.47, "learning_rate": 2.5583719892087032e-05, "loss": 0.0, "step": 75506 }, { "epoch": 1.47, "learning_rate": 2.5583072931829798e-05, "loss": 0.0, "step": 75508 }, { "epoch": 1.47, "learning_rate": 2.5582425971572567e-05, "loss": 0.0, "step": 75510 }, { "epoch": 1.47, "learning_rate": 2.558177901131534e-05, "loss": 0.0, "step": 75512 }, { "epoch": 1.47, "learning_rate": 2.5581132051058105e-05, "loss": 0.0, "step": 75514 }, { "epoch": 1.47, "learning_rate": 2.558048509080087e-05, "loss": 0.0, "step": 75516 }, { "epoch": 1.47, "learning_rate": 2.5579838130543643e-05, "loss": 0.0, "step": 75518 }, { "epoch": 1.47, "learning_rate": 2.557919117028641e-05, "loss": 0.0, "step": 75520 }, { "epoch": 1.47, "learning_rate": 2.557854421002918e-05, "loss": 0.0001, "step": 75522 }, { "epoch": 1.47, "learning_rate": 2.5577897249771947e-05, "loss": 0.0, "step": 75524 }, { "epoch": 1.47, "learning_rate": 2.5577250289514716e-05, "loss": 0.0, "step": 75526 }, { "epoch": 1.47, "learning_rate": 2.557660332925749e-05, "loss": 0.0, "step": 75528 }, { "epoch": 1.47, "learning_rate": 2.5575956369000254e-05, "loss": 0.0, "step": 75530 }, { "epoch": 1.47, "learning_rate": 2.557530940874302e-05, "loss": 0.0, "step": 75532 }, { "epoch": 1.47, "learning_rate": 2.5574662448485792e-05, "loss": 0.0, "step": 75534 }, { "epoch": 1.47, "learning_rate": 2.5574015488228558e-05, "loss": 0.0, "step": 75536 }, { "epoch": 1.47, "learning_rate": 2.5573368527971324e-05, "loss": 0.0065, "step": 75538 }, { "epoch": 1.47, "learning_rate": 2.5572721567714096e-05, "loss": 0.0, "step": 75540 }, { "epoch": 1.47, "learning_rate": 2.5572074607456865e-05, "loss": 0.0, "step": 75542 }, { "epoch": 1.47, "learning_rate": 2.5571427647199638e-05, "loss": 0.0, "step": 75544 }, { "epoch": 1.47, "learning_rate": 2.5570780686942403e-05, "loss": 0.0001, "step": 75546 }, { "epoch": 1.47, "learning_rate": 2.557013372668517e-05, "loss": 0.0, "step": 75548 }, { "epoch": 1.47, "learning_rate": 2.556948676642794e-05, "loss": 0.0, "step": 75550 }, { "epoch": 1.47, "learning_rate": 2.5568839806170707e-05, "loss": 0.0001, "step": 75552 }, { "epoch": 1.47, "learning_rate": 2.5568192845913473e-05, "loss": 0.0001, "step": 75554 }, { "epoch": 1.47, "learning_rate": 2.5567545885656245e-05, "loss": 0.0, "step": 75556 }, { "epoch": 1.47, "learning_rate": 2.5566898925399015e-05, "loss": 0.0052, "step": 75558 }, { "epoch": 1.47, "learning_rate": 2.5566251965141787e-05, "loss": 0.0, "step": 75560 }, { "epoch": 1.47, "learning_rate": 2.5565605004884553e-05, "loss": 0.0043, "step": 75562 }, { "epoch": 1.47, "learning_rate": 2.556495804462732e-05, "loss": 0.0, "step": 75564 }, { "epoch": 1.47, "learning_rate": 2.556431108437009e-05, "loss": 0.0001, "step": 75566 }, { "epoch": 1.47, "learning_rate": 2.5563664124112857e-05, "loss": 0.0, "step": 75568 }, { "epoch": 1.47, "learning_rate": 2.5563017163855622e-05, "loss": 0.0, "step": 75570 }, { "epoch": 1.47, "learning_rate": 2.5562370203598395e-05, "loss": 0.0, "step": 75572 }, { "epoch": 1.47, "learning_rate": 2.5561723243341164e-05, "loss": 0.0, "step": 75574 }, { "epoch": 1.47, "learning_rate": 2.556107628308393e-05, "loss": 0.0002, "step": 75576 }, { "epoch": 1.47, "learning_rate": 2.5560429322826702e-05, "loss": 0.0, "step": 75578 }, { "epoch": 1.47, "learning_rate": 2.5559782362569468e-05, "loss": 0.0043, "step": 75580 }, { "epoch": 1.47, "learning_rate": 2.555913540231224e-05, "loss": 0.0, "step": 75582 }, { "epoch": 1.47, "learning_rate": 2.5558488442055006e-05, "loss": 0.0, "step": 75584 }, { "epoch": 1.47, "learning_rate": 2.555784148179777e-05, "loss": 0.0, "step": 75586 }, { "epoch": 1.47, "learning_rate": 2.5557194521540544e-05, "loss": 0.0, "step": 75588 }, { "epoch": 1.47, "learning_rate": 2.5556547561283313e-05, "loss": 0.0, "step": 75590 }, { "epoch": 1.47, "learning_rate": 2.555590060102608e-05, "loss": 0.0, "step": 75592 }, { "epoch": 1.47, "learning_rate": 2.555525364076885e-05, "loss": 0.0, "step": 75594 }, { "epoch": 1.47, "learning_rate": 2.5554606680511617e-05, "loss": 0.0, "step": 75596 }, { "epoch": 1.47, "learning_rate": 2.5553959720254383e-05, "loss": 0.0, "step": 75598 }, { "epoch": 1.47, "learning_rate": 2.5553312759997155e-05, "loss": 0.0, "step": 75600 }, { "epoch": 1.47, "learning_rate": 2.555266579973992e-05, "loss": 0.0, "step": 75602 }, { "epoch": 1.47, "learning_rate": 2.5552018839482693e-05, "loss": 0.0001, "step": 75604 }, { "epoch": 1.47, "learning_rate": 2.5551371879225462e-05, "loss": 0.0, "step": 75606 }, { "epoch": 1.47, "learning_rate": 2.5550724918968228e-05, "loss": 0.0, "step": 75608 }, { "epoch": 1.47, "learning_rate": 2.5550077958711e-05, "loss": 0.0, "step": 75610 }, { "epoch": 1.47, "learning_rate": 2.5549430998453766e-05, "loss": 0.0061, "step": 75612 }, { "epoch": 1.47, "learning_rate": 2.5548784038196532e-05, "loss": 0.007, "step": 75614 }, { "epoch": 1.47, "learning_rate": 2.5548137077939304e-05, "loss": 0.0, "step": 75616 }, { "epoch": 1.47, "learning_rate": 2.554749011768207e-05, "loss": 0.0001, "step": 75618 }, { "epoch": 1.47, "learning_rate": 2.554684315742484e-05, "loss": 0.0, "step": 75620 }, { "epoch": 1.47, "learning_rate": 2.554619619716761e-05, "loss": 0.0003, "step": 75622 }, { "epoch": 1.47, "learning_rate": 2.5545549236910377e-05, "loss": 0.0, "step": 75624 }, { "epoch": 1.47, "learning_rate": 2.554490227665315e-05, "loss": 0.0017, "step": 75626 }, { "epoch": 1.47, "learning_rate": 2.5544255316395916e-05, "loss": 0.0, "step": 75628 }, { "epoch": 1.47, "learning_rate": 2.554360835613868e-05, "loss": 0.0, "step": 75630 }, { "epoch": 1.47, "learning_rate": 2.5542961395881454e-05, "loss": 0.0, "step": 75632 }, { "epoch": 1.47, "learning_rate": 2.554231443562422e-05, "loss": 0.001, "step": 75634 }, { "epoch": 1.47, "learning_rate": 2.554166747536699e-05, "loss": 0.0, "step": 75636 }, { "epoch": 1.47, "learning_rate": 2.5541020515109758e-05, "loss": 0.0, "step": 75638 }, { "epoch": 1.47, "learning_rate": 2.5540373554852527e-05, "loss": 0.0, "step": 75640 }, { "epoch": 1.47, "learning_rate": 2.55397265945953e-05, "loss": 0.0, "step": 75642 }, { "epoch": 1.47, "learning_rate": 2.5539079634338065e-05, "loss": 0.0, "step": 75644 }, { "epoch": 1.47, "learning_rate": 2.553843267408083e-05, "loss": 0.0, "step": 75646 }, { "epoch": 1.47, "learning_rate": 2.5537785713823603e-05, "loss": 0.0, "step": 75648 }, { "epoch": 1.47, "learning_rate": 2.553713875356637e-05, "loss": 0.0, "step": 75650 }, { "epoch": 1.47, "learning_rate": 2.5536491793309138e-05, "loss": 0.0458, "step": 75652 }, { "epoch": 1.47, "learning_rate": 2.5535844833051907e-05, "loss": 0.0175, "step": 75654 }, { "epoch": 1.47, "learning_rate": 2.5535197872794676e-05, "loss": 0.0, "step": 75656 }, { "epoch": 1.47, "learning_rate": 2.553455091253744e-05, "loss": 0.0016, "step": 75658 }, { "epoch": 1.47, "learning_rate": 2.5533903952280214e-05, "loss": 0.0, "step": 75660 }, { "epoch": 1.47, "learning_rate": 2.553325699202298e-05, "loss": 0.0001, "step": 75662 }, { "epoch": 1.47, "learning_rate": 2.5532610031765752e-05, "loss": 0.0001, "step": 75664 }, { "epoch": 1.47, "learning_rate": 2.5531963071508518e-05, "loss": 0.0, "step": 75666 }, { "epoch": 1.47, "learning_rate": 2.5531316111251287e-05, "loss": 0.0, "step": 75668 }, { "epoch": 1.47, "learning_rate": 2.5530669150994056e-05, "loss": 0.0, "step": 75670 }, { "epoch": 1.47, "learning_rate": 2.5530022190736825e-05, "loss": 0.0, "step": 75672 }, { "epoch": 1.47, "learning_rate": 2.552937523047959e-05, "loss": 0.0, "step": 75674 }, { "epoch": 1.47, "learning_rate": 2.5528728270222363e-05, "loss": 0.0116, "step": 75676 }, { "epoch": 1.47, "learning_rate": 2.552808130996513e-05, "loss": 0.0, "step": 75678 }, { "epoch": 1.47, "learning_rate": 2.5527434349707895e-05, "loss": 0.0, "step": 75680 }, { "epoch": 1.47, "learning_rate": 2.5526787389450667e-05, "loss": 0.0034, "step": 75682 }, { "epoch": 1.47, "learning_rate": 2.5526140429193436e-05, "loss": 0.0, "step": 75684 }, { "epoch": 1.47, "learning_rate": 2.5525493468936205e-05, "loss": 0.0, "step": 75686 }, { "epoch": 1.47, "learning_rate": 2.5524846508678975e-05, "loss": 0.0001, "step": 75688 }, { "epoch": 1.47, "learning_rate": 2.552419954842174e-05, "loss": 0.0, "step": 75690 }, { "epoch": 1.47, "learning_rate": 2.5523552588164513e-05, "loss": 0.0, "step": 75692 }, { "epoch": 1.47, "learning_rate": 2.552290562790728e-05, "loss": 0.0005, "step": 75694 }, { "epoch": 1.47, "learning_rate": 2.5522258667650044e-05, "loss": 0.0055, "step": 75696 }, { "epoch": 1.47, "learning_rate": 2.5521611707392817e-05, "loss": 0.0, "step": 75698 }, { "epoch": 1.47, "learning_rate": 2.5520964747135582e-05, "loss": 0.0172, "step": 75700 }, { "epoch": 1.47, "learning_rate": 2.5520317786878355e-05, "loss": 0.0, "step": 75702 }, { "epoch": 1.47, "learning_rate": 2.5519670826621124e-05, "loss": 0.0, "step": 75704 }, { "epoch": 1.47, "learning_rate": 2.551902386636389e-05, "loss": 0.0, "step": 75706 }, { "epoch": 1.47, "learning_rate": 2.5518376906106662e-05, "loss": 0.0, "step": 75708 }, { "epoch": 1.47, "learning_rate": 2.5517729945849428e-05, "loss": 0.0, "step": 75710 }, { "epoch": 1.47, "learning_rate": 2.5517082985592193e-05, "loss": 0.0081, "step": 75712 }, { "epoch": 1.47, "learning_rate": 2.5516436025334966e-05, "loss": 0.0, "step": 75714 }, { "epoch": 1.47, "learning_rate": 2.551578906507773e-05, "loss": 0.0, "step": 75716 }, { "epoch": 1.47, "learning_rate": 2.55151421048205e-05, "loss": 0.0052, "step": 75718 }, { "epoch": 1.47, "learning_rate": 2.5514495144563273e-05, "loss": 0.0001, "step": 75720 }, { "epoch": 1.47, "learning_rate": 2.551384818430604e-05, "loss": 0.0, "step": 75722 }, { "epoch": 1.47, "learning_rate": 2.551320122404881e-05, "loss": 0.0, "step": 75724 }, { "epoch": 1.47, "learning_rate": 2.5512554263791577e-05, "loss": 0.0, "step": 75726 }, { "epoch": 1.47, "learning_rate": 2.5511907303534343e-05, "loss": 0.0, "step": 75728 }, { "epoch": 1.47, "learning_rate": 2.5511260343277115e-05, "loss": 0.0, "step": 75730 }, { "epoch": 1.47, "learning_rate": 2.551061338301988e-05, "loss": 0.0, "step": 75732 }, { "epoch": 1.47, "learning_rate": 2.550996642276265e-05, "loss": 0.0, "step": 75734 }, { "epoch": 1.47, "learning_rate": 2.5509319462505422e-05, "loss": 0.0001, "step": 75736 }, { "epoch": 1.47, "learning_rate": 2.5508672502248188e-05, "loss": 0.0, "step": 75738 }, { "epoch": 1.47, "learning_rate": 2.5508025541990954e-05, "loss": 0.0, "step": 75740 }, { "epoch": 1.47, "learning_rate": 2.5507378581733726e-05, "loss": 0.0036, "step": 75742 }, { "epoch": 1.47, "learning_rate": 2.5506731621476492e-05, "loss": 0.0, "step": 75744 }, { "epoch": 1.47, "learning_rate": 2.5506084661219264e-05, "loss": 0.0, "step": 75746 }, { "epoch": 1.47, "learning_rate": 2.550543770096203e-05, "loss": 0.0, "step": 75748 }, { "epoch": 1.47, "learning_rate": 2.55047907407048e-05, "loss": 0.0, "step": 75750 }, { "epoch": 1.47, "learning_rate": 2.550414378044757e-05, "loss": 0.0, "step": 75752 }, { "epoch": 1.47, "learning_rate": 2.5503496820190337e-05, "loss": 0.0, "step": 75754 }, { "epoch": 1.47, "learning_rate": 2.5502849859933103e-05, "loss": 0.0002, "step": 75756 }, { "epoch": 1.47, "learning_rate": 2.5502202899675876e-05, "loss": 0.0, "step": 75758 }, { "epoch": 1.47, "learning_rate": 2.550155593941864e-05, "loss": 0.0, "step": 75760 }, { "epoch": 1.47, "learning_rate": 2.550090897916141e-05, "loss": 0.0, "step": 75762 }, { "epoch": 1.47, "learning_rate": 2.550026201890418e-05, "loss": 0.0, "step": 75764 }, { "epoch": 1.47, "learning_rate": 2.549961505864695e-05, "loss": 0.0, "step": 75766 }, { "epoch": 1.47, "learning_rate": 2.549896809838972e-05, "loss": 0.0, "step": 75768 }, { "epoch": 1.47, "learning_rate": 2.5498321138132487e-05, "loss": 0.0, "step": 75770 }, { "epoch": 1.47, "learning_rate": 2.5497674177875252e-05, "loss": 0.0, "step": 75772 }, { "epoch": 1.47, "learning_rate": 2.5497027217618025e-05, "loss": 0.0, "step": 75774 }, { "epoch": 1.47, "learning_rate": 2.549638025736079e-05, "loss": 0.0, "step": 75776 }, { "epoch": 1.47, "learning_rate": 2.5495733297103556e-05, "loss": 0.0, "step": 75778 }, { "epoch": 1.47, "learning_rate": 2.549508633684633e-05, "loss": 0.0, "step": 75780 }, { "epoch": 1.47, "learning_rate": 2.5494439376589098e-05, "loss": 0.0001, "step": 75782 }, { "epoch": 1.47, "learning_rate": 2.549379241633187e-05, "loss": 0.0, "step": 75784 }, { "epoch": 1.47, "learning_rate": 2.5493145456074636e-05, "loss": 0.0, "step": 75786 }, { "epoch": 1.47, "learning_rate": 2.54924984958174e-05, "loss": 0.0, "step": 75788 }, { "epoch": 1.47, "learning_rate": 2.5491851535560174e-05, "loss": 0.0032, "step": 75790 }, { "epoch": 1.47, "learning_rate": 2.549120457530294e-05, "loss": 0.0009, "step": 75792 }, { "epoch": 1.47, "learning_rate": 2.5490557615045706e-05, "loss": 0.0, "step": 75794 }, { "epoch": 1.47, "learning_rate": 2.5489910654788478e-05, "loss": 0.0079, "step": 75796 }, { "epoch": 1.47, "learning_rate": 2.5489263694531247e-05, "loss": 0.0, "step": 75798 }, { "epoch": 1.47, "learning_rate": 2.5488616734274013e-05, "loss": 0.0, "step": 75800 }, { "epoch": 1.47, "learning_rate": 2.5487969774016785e-05, "loss": 0.0, "step": 75802 }, { "epoch": 1.47, "learning_rate": 2.548732281375955e-05, "loss": 0.0, "step": 75804 }, { "epoch": 1.47, "learning_rate": 2.5486675853502323e-05, "loss": 0.0132, "step": 75806 }, { "epoch": 1.47, "learning_rate": 2.548602889324509e-05, "loss": 0.0, "step": 75808 }, { "epoch": 1.47, "learning_rate": 2.5485381932987855e-05, "loss": 0.0, "step": 75810 }, { "epoch": 1.47, "learning_rate": 2.5484734972730627e-05, "loss": 0.0, "step": 75812 }, { "epoch": 1.47, "learning_rate": 2.5484088012473396e-05, "loss": 0.0039, "step": 75814 }, { "epoch": 1.47, "learning_rate": 2.5483441052216162e-05, "loss": 0.0, "step": 75816 }, { "epoch": 1.47, "learning_rate": 2.5482794091958935e-05, "loss": 0.0, "step": 75818 }, { "epoch": 1.47, "learning_rate": 2.54821471317017e-05, "loss": 0.0, "step": 75820 }, { "epoch": 1.47, "learning_rate": 2.5481500171444466e-05, "loss": 0.0, "step": 75822 }, { "epoch": 1.47, "learning_rate": 2.548085321118724e-05, "loss": 0.0, "step": 75824 }, { "epoch": 1.47, "learning_rate": 2.5480206250930004e-05, "loss": 0.0045, "step": 75826 }, { "epoch": 1.47, "learning_rate": 2.5479559290672777e-05, "loss": 0.0, "step": 75828 }, { "epoch": 1.47, "learning_rate": 2.5478912330415546e-05, "loss": 0.0, "step": 75830 }, { "epoch": 1.47, "learning_rate": 2.547826537015831e-05, "loss": 0.0, "step": 75832 }, { "epoch": 1.47, "learning_rate": 2.5477618409901084e-05, "loss": 0.0, "step": 75834 }, { "epoch": 1.47, "learning_rate": 2.547697144964385e-05, "loss": 0.0, "step": 75836 }, { "epoch": 1.47, "learning_rate": 2.5476324489386615e-05, "loss": 0.0, "step": 75838 }, { "epoch": 1.47, "learning_rate": 2.5475677529129388e-05, "loss": 0.0, "step": 75840 }, { "epoch": 1.47, "learning_rate": 2.5475030568872153e-05, "loss": 0.0, "step": 75842 }, { "epoch": 1.47, "learning_rate": 2.5474383608614922e-05, "loss": 0.0, "step": 75844 }, { "epoch": 1.47, "learning_rate": 2.5473736648357695e-05, "loss": 0.0, "step": 75846 }, { "epoch": 1.47, "learning_rate": 2.547308968810046e-05, "loss": 0.0058, "step": 75848 }, { "epoch": 1.47, "learning_rate": 2.5472442727843233e-05, "loss": 0.0, "step": 75850 }, { "epoch": 1.47, "learning_rate": 2.5471795767586e-05, "loss": 0.0, "step": 75852 }, { "epoch": 1.47, "learning_rate": 2.5471148807328764e-05, "loss": 0.0, "step": 75854 }, { "epoch": 1.47, "learning_rate": 2.5470501847071537e-05, "loss": 0.0, "step": 75856 }, { "epoch": 1.47, "learning_rate": 2.5469854886814303e-05, "loss": 0.0118, "step": 75858 }, { "epoch": 1.47, "learning_rate": 2.5469207926557072e-05, "loss": 0.0, "step": 75860 }, { "epoch": 1.47, "learning_rate": 2.5468560966299844e-05, "loss": 0.0, "step": 75862 }, { "epoch": 1.47, "learning_rate": 2.546791400604261e-05, "loss": 0.011, "step": 75864 }, { "epoch": 1.47, "learning_rate": 2.5467267045785382e-05, "loss": 0.0001, "step": 75866 }, { "epoch": 1.47, "learning_rate": 2.5466620085528148e-05, "loss": 0.0, "step": 75868 }, { "epoch": 1.47, "learning_rate": 2.5465973125270914e-05, "loss": 0.0, "step": 75870 }, { "epoch": 1.47, "learning_rate": 2.5465326165013686e-05, "loss": 0.0, "step": 75872 }, { "epoch": 1.47, "learning_rate": 2.5464679204756452e-05, "loss": 0.0, "step": 75874 }, { "epoch": 1.47, "learning_rate": 2.546403224449922e-05, "loss": 0.0038, "step": 75876 }, { "epoch": 1.47, "learning_rate": 2.546338528424199e-05, "loss": 0.0084, "step": 75878 }, { "epoch": 1.47, "learning_rate": 2.546273832398476e-05, "loss": 0.0, "step": 75880 }, { "epoch": 1.47, "learning_rate": 2.5462091363727525e-05, "loss": 0.0, "step": 75882 }, { "epoch": 1.47, "learning_rate": 2.5461444403470297e-05, "loss": 0.0001, "step": 75884 }, { "epoch": 1.47, "learning_rate": 2.5460797443213063e-05, "loss": 0.0001, "step": 75886 }, { "epoch": 1.47, "learning_rate": 2.5460150482955836e-05, "loss": 0.0001, "step": 75888 }, { "epoch": 1.47, "learning_rate": 2.54595035226986e-05, "loss": 0.0, "step": 75890 }, { "epoch": 1.47, "learning_rate": 2.545885656244137e-05, "loss": 0.0, "step": 75892 }, { "epoch": 1.47, "learning_rate": 2.545820960218414e-05, "loss": 0.0023, "step": 75894 }, { "epoch": 1.47, "learning_rate": 2.545756264192691e-05, "loss": 0.0, "step": 75896 }, { "epoch": 1.47, "learning_rate": 2.5456915681669674e-05, "loss": 0.0, "step": 75898 }, { "epoch": 1.47, "learning_rate": 2.5456268721412447e-05, "loss": 0.0, "step": 75900 }, { "epoch": 1.47, "learning_rate": 2.5455621761155212e-05, "loss": 0.0, "step": 75902 }, { "epoch": 1.47, "learning_rate": 2.5454974800897978e-05, "loss": 0.0, "step": 75904 }, { "epoch": 1.47, "learning_rate": 2.545432784064075e-05, "loss": 0.0001, "step": 75906 }, { "epoch": 1.47, "learning_rate": 2.545368088038352e-05, "loss": 0.0, "step": 75908 }, { "epoch": 1.47, "learning_rate": 2.545303392012629e-05, "loss": 0.0001, "step": 75910 }, { "epoch": 1.47, "learning_rate": 2.5452386959869058e-05, "loss": 0.0, "step": 75912 }, { "epoch": 1.47, "learning_rate": 2.5451739999611823e-05, "loss": 0.0, "step": 75914 }, { "epoch": 1.47, "learning_rate": 2.5451093039354596e-05, "loss": 0.0, "step": 75916 }, { "epoch": 1.47, "learning_rate": 2.545044607909736e-05, "loss": 0.0, "step": 75918 }, { "epoch": 1.47, "learning_rate": 2.5449799118840127e-05, "loss": 0.0, "step": 75920 }, { "epoch": 1.47, "learning_rate": 2.54491521585829e-05, "loss": 0.0, "step": 75922 }, { "epoch": 1.47, "learning_rate": 2.544850519832567e-05, "loss": 0.0, "step": 75924 }, { "epoch": 1.47, "learning_rate": 2.5447858238068438e-05, "loss": 0.0, "step": 75926 }, { "epoch": 1.47, "learning_rate": 2.5447211277811207e-05, "loss": 0.0003, "step": 75928 }, { "epoch": 1.47, "learning_rate": 2.5446564317553973e-05, "loss": 0.0, "step": 75930 }, { "epoch": 1.47, "learning_rate": 2.5445917357296745e-05, "loss": 0.0, "step": 75932 }, { "epoch": 1.47, "learning_rate": 2.544527039703951e-05, "loss": 0.0002, "step": 75934 }, { "epoch": 1.47, "learning_rate": 2.5444623436782277e-05, "loss": 0.0, "step": 75936 }, { "epoch": 1.47, "learning_rate": 2.544397647652505e-05, "loss": 0.0, "step": 75938 }, { "epoch": 1.47, "learning_rate": 2.5443329516267815e-05, "loss": 0.0, "step": 75940 }, { "epoch": 1.47, "learning_rate": 2.5442682556010584e-05, "loss": 0.0, "step": 75942 }, { "epoch": 1.47, "learning_rate": 2.5442035595753356e-05, "loss": 0.0, "step": 75944 }, { "epoch": 1.47, "learning_rate": 2.5441388635496122e-05, "loss": 0.0, "step": 75946 }, { "epoch": 1.47, "learning_rate": 2.5440741675238895e-05, "loss": 0.0, "step": 75948 }, { "epoch": 1.47, "learning_rate": 2.544009471498166e-05, "loss": 0.0002, "step": 75950 }, { "epoch": 1.47, "learning_rate": 2.5439447754724426e-05, "loss": 0.0, "step": 75952 }, { "epoch": 1.47, "learning_rate": 2.54388007944672e-05, "loss": 0.0, "step": 75954 }, { "epoch": 1.47, "learning_rate": 2.5438153834209964e-05, "loss": 0.0, "step": 75956 }, { "epoch": 1.47, "learning_rate": 2.5437506873952733e-05, "loss": 0.0, "step": 75958 }, { "epoch": 1.47, "learning_rate": 2.5436859913695506e-05, "loss": 0.0001, "step": 75960 }, { "epoch": 1.47, "learning_rate": 2.543621295343827e-05, "loss": 0.0, "step": 75962 }, { "epoch": 1.47, "learning_rate": 2.5435565993181037e-05, "loss": 0.0, "step": 75964 }, { "epoch": 1.47, "learning_rate": 2.543491903292381e-05, "loss": 0.0001, "step": 75966 }, { "epoch": 1.47, "learning_rate": 2.5434272072666575e-05, "loss": 0.0006, "step": 75968 }, { "epoch": 1.47, "learning_rate": 2.5433625112409348e-05, "loss": 0.0, "step": 75970 }, { "epoch": 1.47, "learning_rate": 2.5432978152152113e-05, "loss": 0.0, "step": 75972 }, { "epoch": 1.47, "learning_rate": 2.5432331191894882e-05, "loss": 0.0, "step": 75974 }, { "epoch": 1.47, "learning_rate": 2.5431684231637655e-05, "loss": 0.0, "step": 75976 }, { "epoch": 1.47, "learning_rate": 2.543103727138042e-05, "loss": 0.0, "step": 75978 }, { "epoch": 1.47, "learning_rate": 2.5430390311123186e-05, "loss": 0.0, "step": 75980 }, { "epoch": 1.47, "learning_rate": 2.542974335086596e-05, "loss": 0.0, "step": 75982 }, { "epoch": 1.47, "learning_rate": 2.5429096390608724e-05, "loss": 0.0, "step": 75984 }, { "epoch": 1.47, "learning_rate": 2.5428449430351494e-05, "loss": 0.0003, "step": 75986 }, { "epoch": 1.47, "learning_rate": 2.5427802470094263e-05, "loss": 0.0, "step": 75988 }, { "epoch": 1.47, "learning_rate": 2.5427155509837032e-05, "loss": 0.0, "step": 75990 }, { "epoch": 1.47, "learning_rate": 2.5426508549579804e-05, "loss": 0.0009, "step": 75992 }, { "epoch": 1.47, "learning_rate": 2.542586158932257e-05, "loss": 0.0, "step": 75994 }, { "epoch": 1.47, "learning_rate": 2.5425214629065336e-05, "loss": 0.0, "step": 75996 }, { "epoch": 1.48, "learning_rate": 2.5424567668808108e-05, "loss": 0.0, "step": 75998 }, { "epoch": 1.48, "learning_rate": 2.5423920708550874e-05, "loss": 0.0, "step": 76000 }, { "epoch": 1.48, "learning_rate": 2.5423273748293643e-05, "loss": 0.0, "step": 76002 }, { "epoch": 1.48, "learning_rate": 2.5422626788036412e-05, "loss": 0.0, "step": 76004 }, { "epoch": 1.48, "learning_rate": 2.542197982777918e-05, "loss": 0.0002, "step": 76006 }, { "epoch": 1.48, "learning_rate": 2.5421332867521953e-05, "loss": 0.0, "step": 76008 }, { "epoch": 1.48, "learning_rate": 2.542068590726472e-05, "loss": 0.0, "step": 76010 }, { "epoch": 1.48, "learning_rate": 2.5420038947007485e-05, "loss": 0.0001, "step": 76012 }, { "epoch": 1.48, "learning_rate": 2.5419391986750257e-05, "loss": 0.0, "step": 76014 }, { "epoch": 1.48, "learning_rate": 2.5418745026493023e-05, "loss": 0.0, "step": 76016 }, { "epoch": 1.48, "learning_rate": 2.541809806623579e-05, "loss": 0.0, "step": 76018 }, { "epoch": 1.48, "learning_rate": 2.541745110597856e-05, "loss": 0.0, "step": 76020 }, { "epoch": 1.48, "learning_rate": 2.541680414572133e-05, "loss": 0.0, "step": 76022 }, { "epoch": 1.48, "learning_rate": 2.5416157185464096e-05, "loss": 0.0, "step": 76024 }, { "epoch": 1.48, "learning_rate": 2.541551022520687e-05, "loss": 0.0, "step": 76026 }, { "epoch": 1.48, "learning_rate": 2.5414863264949634e-05, "loss": 0.0, "step": 76028 }, { "epoch": 1.48, "learning_rate": 2.5414216304692407e-05, "loss": 0.0, "step": 76030 }, { "epoch": 1.48, "learning_rate": 2.5413569344435172e-05, "loss": 0.0008, "step": 76032 }, { "epoch": 1.48, "learning_rate": 2.5412922384177938e-05, "loss": 0.0001, "step": 76034 }, { "epoch": 1.48, "learning_rate": 2.541227542392071e-05, "loss": 0.0, "step": 76036 }, { "epoch": 1.48, "learning_rate": 2.541162846366348e-05, "loss": 0.0, "step": 76038 }, { "epoch": 1.48, "learning_rate": 2.5410981503406245e-05, "loss": 0.0, "step": 76040 }, { "epoch": 1.48, "learning_rate": 2.5410334543149018e-05, "loss": 0.0, "step": 76042 }, { "epoch": 1.48, "learning_rate": 2.5409687582891783e-05, "loss": 0.0, "step": 76044 }, { "epoch": 1.48, "learning_rate": 2.540904062263455e-05, "loss": 0.0, "step": 76046 }, { "epoch": 1.48, "learning_rate": 2.540839366237732e-05, "loss": 0.0, "step": 76048 }, { "epoch": 1.48, "learning_rate": 2.5407746702120087e-05, "loss": 0.0, "step": 76050 }, { "epoch": 1.48, "learning_rate": 2.540709974186286e-05, "loss": 0.0093, "step": 76052 }, { "epoch": 1.48, "learning_rate": 2.540645278160563e-05, "loss": 0.0, "step": 76054 }, { "epoch": 1.48, "learning_rate": 2.5405805821348395e-05, "loss": 0.0, "step": 76056 }, { "epoch": 1.48, "learning_rate": 2.5405158861091167e-05, "loss": 0.0, "step": 76058 }, { "epoch": 1.48, "learning_rate": 2.5404511900833933e-05, "loss": 0.0, "step": 76060 }, { "epoch": 1.48, "learning_rate": 2.54038649405767e-05, "loss": 0.0005, "step": 76062 }, { "epoch": 1.48, "learning_rate": 2.540321798031947e-05, "loss": 0.0, "step": 76064 }, { "epoch": 1.48, "learning_rate": 2.5402571020062237e-05, "loss": 0.0, "step": 76066 }, { "epoch": 1.48, "learning_rate": 2.540192405980501e-05, "loss": 0.0, "step": 76068 }, { "epoch": 1.48, "learning_rate": 2.5401277099547778e-05, "loss": 0.0, "step": 76070 }, { "epoch": 1.48, "learning_rate": 2.5400630139290544e-05, "loss": 0.0, "step": 76072 }, { "epoch": 1.48, "learning_rate": 2.5399983179033316e-05, "loss": 0.0, "step": 76074 }, { "epoch": 1.48, "learning_rate": 2.5399336218776082e-05, "loss": 0.0, "step": 76076 }, { "epoch": 1.48, "learning_rate": 2.5398689258518848e-05, "loss": 0.0001, "step": 76078 }, { "epoch": 1.48, "learning_rate": 2.539804229826162e-05, "loss": 0.0, "step": 76080 }, { "epoch": 1.48, "learning_rate": 2.5397395338004386e-05, "loss": 0.0, "step": 76082 }, { "epoch": 1.48, "learning_rate": 2.5396748377747155e-05, "loss": 0.0, "step": 76084 }, { "epoch": 1.48, "learning_rate": 2.5396101417489927e-05, "loss": 0.0045, "step": 76086 }, { "epoch": 1.48, "learning_rate": 2.5395454457232693e-05, "loss": 0.0, "step": 76088 }, { "epoch": 1.48, "learning_rate": 2.5394807496975466e-05, "loss": 0.0, "step": 76090 }, { "epoch": 1.48, "learning_rate": 2.539416053671823e-05, "loss": 0.0, "step": 76092 }, { "epoch": 1.48, "learning_rate": 2.5393513576460997e-05, "loss": 0.0016, "step": 76094 }, { "epoch": 1.48, "learning_rate": 2.539286661620377e-05, "loss": 0.0003, "step": 76096 }, { "epoch": 1.48, "learning_rate": 2.5392219655946535e-05, "loss": 0.0, "step": 76098 }, { "epoch": 1.48, "learning_rate": 2.5391572695689304e-05, "loss": 0.0, "step": 76100 }, { "epoch": 1.48, "learning_rate": 2.5390925735432077e-05, "loss": 0.0, "step": 76102 }, { "epoch": 1.48, "learning_rate": 2.5390278775174842e-05, "loss": 0.008, "step": 76104 }, { "epoch": 1.48, "learning_rate": 2.5389631814917608e-05, "loss": 0.0, "step": 76106 }, { "epoch": 1.48, "learning_rate": 2.538898485466038e-05, "loss": 0.0003, "step": 76108 }, { "epoch": 1.48, "learning_rate": 2.5388337894403146e-05, "loss": 0.002, "step": 76110 }, { "epoch": 1.48, "learning_rate": 2.538769093414592e-05, "loss": 0.0, "step": 76112 }, { "epoch": 1.48, "learning_rate": 2.5387043973888684e-05, "loss": 0.0005, "step": 76114 }, { "epoch": 1.48, "learning_rate": 2.5386397013631454e-05, "loss": 0.0, "step": 76116 }, { "epoch": 1.48, "learning_rate": 2.5385750053374223e-05, "loss": 0.0437, "step": 76118 }, { "epoch": 1.48, "learning_rate": 2.5385103093116992e-05, "loss": 0.0, "step": 76120 }, { "epoch": 1.48, "learning_rate": 2.5384456132859757e-05, "loss": 0.0, "step": 76122 }, { "epoch": 1.48, "learning_rate": 2.538380917260253e-05, "loss": 0.0, "step": 76124 }, { "epoch": 1.48, "learning_rate": 2.5383162212345296e-05, "loss": 0.0, "step": 76126 }, { "epoch": 1.48, "learning_rate": 2.538251525208806e-05, "loss": 0.007, "step": 76128 }, { "epoch": 1.48, "learning_rate": 2.5381868291830834e-05, "loss": 0.0, "step": 76130 }, { "epoch": 1.48, "learning_rate": 2.5381221331573603e-05, "loss": 0.0055, "step": 76132 }, { "epoch": 1.48, "learning_rate": 2.5380574371316372e-05, "loss": 0.0003, "step": 76134 }, { "epoch": 1.48, "learning_rate": 2.537992741105914e-05, "loss": 0.0, "step": 76136 }, { "epoch": 1.48, "learning_rate": 2.5379280450801907e-05, "loss": 0.0039, "step": 76138 }, { "epoch": 1.48, "learning_rate": 2.537863349054468e-05, "loss": 0.0, "step": 76140 }, { "epoch": 1.48, "learning_rate": 2.5377986530287445e-05, "loss": 0.0095, "step": 76142 }, { "epoch": 1.48, "learning_rate": 2.537733957003021e-05, "loss": 0.0, "step": 76144 }, { "epoch": 1.48, "learning_rate": 2.5376692609772983e-05, "loss": 0.0, "step": 76146 }, { "epoch": 1.48, "learning_rate": 2.5376045649515752e-05, "loss": 0.0, "step": 76148 }, { "epoch": 1.48, "learning_rate": 2.537539868925852e-05, "loss": 0.0, "step": 76150 }, { "epoch": 1.48, "learning_rate": 2.537475172900129e-05, "loss": 0.0001, "step": 76152 }, { "epoch": 1.48, "learning_rate": 2.5374104768744056e-05, "loss": 0.0, "step": 76154 }, { "epoch": 1.48, "learning_rate": 2.537345780848683e-05, "loss": 0.0, "step": 76156 }, { "epoch": 1.48, "learning_rate": 2.5372810848229594e-05, "loss": 0.0, "step": 76158 }, { "epoch": 1.48, "learning_rate": 2.537216388797236e-05, "loss": 0.0, "step": 76160 }, { "epoch": 1.48, "learning_rate": 2.5371516927715132e-05, "loss": 0.0, "step": 76162 }, { "epoch": 1.48, "learning_rate": 2.53708699674579e-05, "loss": 0.0, "step": 76164 }, { "epoch": 1.48, "learning_rate": 2.5370223007200667e-05, "loss": 0.0, "step": 76166 }, { "epoch": 1.48, "learning_rate": 2.536957604694344e-05, "loss": 0.0001, "step": 76168 }, { "epoch": 1.48, "learning_rate": 2.5368929086686205e-05, "loss": 0.0, "step": 76170 }, { "epoch": 1.48, "learning_rate": 2.5368282126428978e-05, "loss": 0.0001, "step": 76172 }, { "epoch": 1.48, "learning_rate": 2.5367635166171743e-05, "loss": 0.0, "step": 76174 }, { "epoch": 1.48, "learning_rate": 2.536698820591451e-05, "loss": 0.0, "step": 76176 }, { "epoch": 1.48, "learning_rate": 2.536634124565728e-05, "loss": 0.0, "step": 76178 }, { "epoch": 1.48, "learning_rate": 2.5365694285400047e-05, "loss": 0.0, "step": 76180 }, { "epoch": 1.48, "learning_rate": 2.5365047325142816e-05, "loss": 0.0011, "step": 76182 }, { "epoch": 1.48, "learning_rate": 2.536440036488559e-05, "loss": 0.0, "step": 76184 }, { "epoch": 1.48, "learning_rate": 2.5363753404628355e-05, "loss": 0.0, "step": 76186 }, { "epoch": 1.48, "learning_rate": 2.536310644437112e-05, "loss": 0.0, "step": 76188 }, { "epoch": 1.48, "learning_rate": 2.5362459484113893e-05, "loss": 0.0033, "step": 76190 }, { "epoch": 1.48, "learning_rate": 2.536181252385666e-05, "loss": 0.0044, "step": 76192 }, { "epoch": 1.48, "learning_rate": 2.536116556359943e-05, "loss": 0.0, "step": 76194 }, { "epoch": 1.48, "learning_rate": 2.5360518603342197e-05, "loss": 0.0, "step": 76196 }, { "epoch": 1.48, "learning_rate": 2.5359871643084966e-05, "loss": 0.0002, "step": 76198 }, { "epoch": 1.48, "learning_rate": 2.5359224682827738e-05, "loss": 0.0017, "step": 76200 }, { "epoch": 1.48, "learning_rate": 2.5358577722570504e-05, "loss": 0.0, "step": 76202 }, { "epoch": 1.48, "learning_rate": 2.535793076231327e-05, "loss": 0.0001, "step": 76204 }, { "epoch": 1.48, "learning_rate": 2.5357283802056042e-05, "loss": 0.0, "step": 76206 }, { "epoch": 1.48, "learning_rate": 2.5356636841798808e-05, "loss": 0.0044, "step": 76208 }, { "epoch": 1.48, "learning_rate": 2.535598988154158e-05, "loss": 0.0, "step": 76210 }, { "epoch": 1.48, "learning_rate": 2.5355342921284346e-05, "loss": 0.0052, "step": 76212 }, { "epoch": 1.48, "learning_rate": 2.5354695961027115e-05, "loss": 0.0, "step": 76214 }, { "epoch": 1.48, "learning_rate": 2.5354049000769887e-05, "loss": 0.0, "step": 76216 }, { "epoch": 1.48, "learning_rate": 2.5353402040512653e-05, "loss": 0.0, "step": 76218 }, { "epoch": 1.48, "learning_rate": 2.535275508025542e-05, "loss": 0.0, "step": 76220 }, { "epoch": 1.48, "learning_rate": 2.535210811999819e-05, "loss": 0.0014, "step": 76222 }, { "epoch": 1.48, "learning_rate": 2.5351461159740957e-05, "loss": 0.0, "step": 76224 }, { "epoch": 1.48, "learning_rate": 2.5350814199483726e-05, "loss": 0.0007, "step": 76226 }, { "epoch": 1.48, "learning_rate": 2.5350167239226495e-05, "loss": 0.0001, "step": 76228 }, { "epoch": 1.48, "learning_rate": 2.5349520278969264e-05, "loss": 0.0, "step": 76230 }, { "epoch": 1.48, "learning_rate": 2.5348873318712037e-05, "loss": 0.0, "step": 76232 }, { "epoch": 1.48, "learning_rate": 2.5348226358454802e-05, "loss": 0.0024, "step": 76234 }, { "epoch": 1.48, "learning_rate": 2.5347579398197568e-05, "loss": 0.0046, "step": 76236 }, { "epoch": 1.48, "learning_rate": 2.534693243794034e-05, "loss": 0.0, "step": 76238 }, { "epoch": 1.48, "learning_rate": 2.5346285477683106e-05, "loss": 0.0055, "step": 76240 }, { "epoch": 1.48, "learning_rate": 2.5345638517425875e-05, "loss": 0.0002, "step": 76242 }, { "epoch": 1.48, "learning_rate": 2.5344991557168644e-05, "loss": 0.0002, "step": 76244 }, { "epoch": 1.48, "learning_rate": 2.5344344596911414e-05, "loss": 0.0002, "step": 76246 }, { "epoch": 1.48, "learning_rate": 2.534369763665418e-05, "loss": 0.0, "step": 76248 }, { "epoch": 1.48, "learning_rate": 2.5343050676396952e-05, "loss": 0.0, "step": 76250 }, { "epoch": 1.48, "learning_rate": 2.5342403716139717e-05, "loss": 0.0, "step": 76252 }, { "epoch": 1.48, "learning_rate": 2.534175675588249e-05, "loss": 0.0, "step": 76254 }, { "epoch": 1.48, "learning_rate": 2.5341109795625256e-05, "loss": 0.0, "step": 76256 }, { "epoch": 1.48, "learning_rate": 2.534046283536802e-05, "loss": 0.0, "step": 76258 }, { "epoch": 1.48, "learning_rate": 2.5339815875110794e-05, "loss": 0.0, "step": 76260 }, { "epoch": 1.48, "learning_rate": 2.5339168914853563e-05, "loss": 0.0, "step": 76262 }, { "epoch": 1.48, "learning_rate": 2.533852195459633e-05, "loss": 0.0001, "step": 76264 }, { "epoch": 1.48, "learning_rate": 2.53378749943391e-05, "loss": 0.008, "step": 76266 }, { "epoch": 1.48, "learning_rate": 2.5337228034081867e-05, "loss": 0.0, "step": 76268 }, { "epoch": 1.48, "learning_rate": 2.5336581073824632e-05, "loss": 0.0017, "step": 76270 }, { "epoch": 1.48, "learning_rate": 2.5335934113567405e-05, "loss": 0.0, "step": 76272 }, { "epoch": 1.48, "learning_rate": 2.533528715331017e-05, "loss": 0.0126, "step": 76274 }, { "epoch": 1.48, "learning_rate": 2.5334640193052943e-05, "loss": 0.0, "step": 76276 }, { "epoch": 1.48, "learning_rate": 2.5333993232795712e-05, "loss": 0.0, "step": 76278 }, { "epoch": 1.48, "learning_rate": 2.5333346272538478e-05, "loss": 0.0, "step": 76280 }, { "epoch": 1.48, "learning_rate": 2.533269931228125e-05, "loss": 0.0004, "step": 76282 }, { "epoch": 1.48, "learning_rate": 2.5332052352024016e-05, "loss": 0.0, "step": 76284 }, { "epoch": 1.48, "learning_rate": 2.533140539176678e-05, "loss": 0.0044, "step": 76286 }, { "epoch": 1.48, "learning_rate": 2.5330758431509554e-05, "loss": 0.0, "step": 76288 }, { "epoch": 1.48, "learning_rate": 2.533011147125232e-05, "loss": 0.0, "step": 76290 }, { "epoch": 1.48, "learning_rate": 2.5329464510995092e-05, "loss": 0.0, "step": 76292 }, { "epoch": 1.48, "learning_rate": 2.532881755073786e-05, "loss": 0.0, "step": 76294 }, { "epoch": 1.48, "learning_rate": 2.5328170590480627e-05, "loss": 0.0002, "step": 76296 }, { "epoch": 1.48, "learning_rate": 2.53275236302234e-05, "loss": 0.0, "step": 76298 }, { "epoch": 1.48, "learning_rate": 2.5326876669966165e-05, "loss": 0.0, "step": 76300 }, { "epoch": 1.48, "learning_rate": 2.532622970970893e-05, "loss": 0.0, "step": 76302 }, { "epoch": 1.48, "learning_rate": 2.5325582749451703e-05, "loss": 0.0, "step": 76304 }, { "epoch": 1.48, "learning_rate": 2.532493578919447e-05, "loss": 0.0, "step": 76306 }, { "epoch": 1.48, "learning_rate": 2.5324288828937238e-05, "loss": 0.0, "step": 76308 }, { "epoch": 1.48, "learning_rate": 2.532364186868001e-05, "loss": 0.0, "step": 76310 }, { "epoch": 1.48, "learning_rate": 2.5322994908422776e-05, "loss": 0.0001, "step": 76312 }, { "epoch": 1.48, "learning_rate": 2.532234794816555e-05, "loss": 0.0, "step": 76314 }, { "epoch": 1.48, "learning_rate": 2.5321700987908315e-05, "loss": 0.0, "step": 76316 }, { "epoch": 1.48, "learning_rate": 2.532105402765108e-05, "loss": 0.0, "step": 76318 }, { "epoch": 1.48, "learning_rate": 2.5320407067393853e-05, "loss": 0.0, "step": 76320 }, { "epoch": 1.48, "learning_rate": 2.531976010713662e-05, "loss": 0.0, "step": 76322 }, { "epoch": 1.48, "learning_rate": 2.5319113146879388e-05, "loss": 0.0004, "step": 76324 }, { "epoch": 1.48, "learning_rate": 2.531846618662216e-05, "loss": 0.0, "step": 76326 }, { "epoch": 1.48, "learning_rate": 2.5317819226364926e-05, "loss": 0.0, "step": 76328 }, { "epoch": 1.48, "learning_rate": 2.531717226610769e-05, "loss": 0.0, "step": 76330 }, { "epoch": 1.48, "learning_rate": 2.5316525305850464e-05, "loss": 0.0, "step": 76332 }, { "epoch": 1.48, "learning_rate": 2.531587834559323e-05, "loss": 0.0, "step": 76334 }, { "epoch": 1.48, "learning_rate": 2.5315231385336002e-05, "loss": 0.0088, "step": 76336 }, { "epoch": 1.48, "learning_rate": 2.5314584425078768e-05, "loss": 0.0, "step": 76338 }, { "epoch": 1.48, "learning_rate": 2.5313937464821537e-05, "loss": 0.0, "step": 76340 }, { "epoch": 1.48, "learning_rate": 2.531329050456431e-05, "loss": 0.0001, "step": 76342 }, { "epoch": 1.48, "learning_rate": 2.5312643544307075e-05, "loss": 0.0, "step": 76344 }, { "epoch": 1.48, "learning_rate": 2.531199658404984e-05, "loss": 0.0, "step": 76346 }, { "epoch": 1.48, "learning_rate": 2.5311349623792613e-05, "loss": 0.0, "step": 76348 }, { "epoch": 1.48, "learning_rate": 2.531070266353538e-05, "loss": 0.0, "step": 76350 }, { "epoch": 1.48, "learning_rate": 2.531005570327815e-05, "loss": 0.0021, "step": 76352 }, { "epoch": 1.48, "learning_rate": 2.5309408743020917e-05, "loss": 0.0, "step": 76354 }, { "epoch": 1.48, "learning_rate": 2.5308761782763686e-05, "loss": 0.0, "step": 76356 }, { "epoch": 1.48, "learning_rate": 2.5308114822506455e-05, "loss": 0.0, "step": 76358 }, { "epoch": 1.48, "learning_rate": 2.5307467862249224e-05, "loss": 0.0049, "step": 76360 }, { "epoch": 1.48, "learning_rate": 2.530682090199199e-05, "loss": 0.0, "step": 76362 }, { "epoch": 1.48, "learning_rate": 2.5306173941734762e-05, "loss": 0.0, "step": 76364 }, { "epoch": 1.48, "learning_rate": 2.5305526981477528e-05, "loss": 0.0, "step": 76366 }, { "epoch": 1.48, "learning_rate": 2.5304880021220294e-05, "loss": 0.0, "step": 76368 }, { "epoch": 1.48, "learning_rate": 2.5304233060963066e-05, "loss": 0.0, "step": 76370 }, { "epoch": 1.48, "learning_rate": 2.5303586100705835e-05, "loss": 0.0, "step": 76372 }, { "epoch": 1.48, "learning_rate": 2.5302939140448604e-05, "loss": 0.0, "step": 76374 }, { "epoch": 1.48, "learning_rate": 2.5302292180191374e-05, "loss": 0.0, "step": 76376 }, { "epoch": 1.48, "learning_rate": 2.530164521993414e-05, "loss": 0.0, "step": 76378 }, { "epoch": 1.48, "learning_rate": 2.5300998259676912e-05, "loss": 0.0, "step": 76380 }, { "epoch": 1.48, "learning_rate": 2.5300351299419677e-05, "loss": 0.0, "step": 76382 }, { "epoch": 1.48, "learning_rate": 2.5299704339162443e-05, "loss": 0.0, "step": 76384 }, { "epoch": 1.48, "learning_rate": 2.5299057378905216e-05, "loss": 0.0, "step": 76386 }, { "epoch": 1.48, "learning_rate": 2.5298410418647985e-05, "loss": 0.0, "step": 76388 }, { "epoch": 1.48, "learning_rate": 2.529776345839075e-05, "loss": 0.0001, "step": 76390 }, { "epoch": 1.48, "learning_rate": 2.5297116498133523e-05, "loss": 0.0, "step": 76392 }, { "epoch": 1.48, "learning_rate": 2.529646953787629e-05, "loss": 0.0, "step": 76394 }, { "epoch": 1.48, "learning_rate": 2.529582257761906e-05, "loss": 0.0, "step": 76396 }, { "epoch": 1.48, "learning_rate": 2.5295175617361827e-05, "loss": 0.0, "step": 76398 }, { "epoch": 1.48, "learning_rate": 2.5294528657104592e-05, "loss": 0.0001, "step": 76400 }, { "epoch": 1.48, "learning_rate": 2.5293881696847365e-05, "loss": 0.0, "step": 76402 }, { "epoch": 1.48, "learning_rate": 2.5293234736590134e-05, "loss": 0.0004, "step": 76404 }, { "epoch": 1.48, "learning_rate": 2.52925877763329e-05, "loss": 0.0, "step": 76406 }, { "epoch": 1.48, "learning_rate": 2.5291940816075672e-05, "loss": 0.0, "step": 76408 }, { "epoch": 1.48, "learning_rate": 2.5291293855818438e-05, "loss": 0.0, "step": 76410 }, { "epoch": 1.48, "learning_rate": 2.5290646895561203e-05, "loss": 0.0083, "step": 76412 }, { "epoch": 1.48, "learning_rate": 2.5289999935303976e-05, "loss": 0.0001, "step": 76414 }, { "epoch": 1.48, "learning_rate": 2.528935297504674e-05, "loss": 0.0, "step": 76416 }, { "epoch": 1.48, "learning_rate": 2.5288706014789514e-05, "loss": 0.0, "step": 76418 }, { "epoch": 1.48, "learning_rate": 2.528805905453228e-05, "loss": 0.0, "step": 76420 }, { "epoch": 1.48, "learning_rate": 2.528741209427505e-05, "loss": 0.0001, "step": 76422 }, { "epoch": 1.48, "learning_rate": 2.528676513401782e-05, "loss": 0.0, "step": 76424 }, { "epoch": 1.48, "learning_rate": 2.5286118173760587e-05, "loss": 0.0, "step": 76426 }, { "epoch": 1.48, "learning_rate": 2.5285471213503353e-05, "loss": 0.0002, "step": 76428 }, { "epoch": 1.48, "learning_rate": 2.5284824253246125e-05, "loss": 0.0, "step": 76430 }, { "epoch": 1.48, "learning_rate": 2.528417729298889e-05, "loss": 0.0004, "step": 76432 }, { "epoch": 1.48, "learning_rate": 2.5283530332731663e-05, "loss": 0.0001, "step": 76434 }, { "epoch": 1.48, "learning_rate": 2.528288337247443e-05, "loss": 0.0, "step": 76436 }, { "epoch": 1.48, "learning_rate": 2.5282236412217198e-05, "loss": 0.0, "step": 76438 }, { "epoch": 1.48, "learning_rate": 2.528158945195997e-05, "loss": 0.0, "step": 76440 }, { "epoch": 1.48, "learning_rate": 2.5280942491702736e-05, "loss": 0.0, "step": 76442 }, { "epoch": 1.48, "learning_rate": 2.5280295531445502e-05, "loss": 0.0, "step": 76444 }, { "epoch": 1.48, "learning_rate": 2.5279648571188275e-05, "loss": 0.0109, "step": 76446 }, { "epoch": 1.48, "learning_rate": 2.527900161093104e-05, "loss": 0.0, "step": 76448 }, { "epoch": 1.48, "learning_rate": 2.527835465067381e-05, "loss": 0.0001, "step": 76450 }, { "epoch": 1.48, "learning_rate": 2.527770769041658e-05, "loss": 0.0006, "step": 76452 }, { "epoch": 1.48, "learning_rate": 2.5277060730159347e-05, "loss": 0.0, "step": 76454 }, { "epoch": 1.48, "learning_rate": 2.527641376990212e-05, "loss": 0.0, "step": 76456 }, { "epoch": 1.48, "learning_rate": 2.5275766809644886e-05, "loss": 0.0, "step": 76458 }, { "epoch": 1.48, "learning_rate": 2.527511984938765e-05, "loss": 0.0, "step": 76460 }, { "epoch": 1.48, "learning_rate": 2.5274472889130424e-05, "loss": 0.0, "step": 76462 }, { "epoch": 1.48, "learning_rate": 2.527382592887319e-05, "loss": 0.0, "step": 76464 }, { "epoch": 1.48, "learning_rate": 2.527317896861596e-05, "loss": 0.0, "step": 76466 }, { "epoch": 1.48, "learning_rate": 2.5272532008358728e-05, "loss": 0.0, "step": 76468 }, { "epoch": 1.48, "learning_rate": 2.5271885048101497e-05, "loss": 0.0, "step": 76470 }, { "epoch": 1.48, "learning_rate": 2.5271238087844262e-05, "loss": 0.0, "step": 76472 }, { "epoch": 1.48, "learning_rate": 2.5270591127587035e-05, "loss": 0.0, "step": 76474 }, { "epoch": 1.48, "learning_rate": 2.52699441673298e-05, "loss": 0.0, "step": 76476 }, { "epoch": 1.48, "learning_rate": 2.5269297207072573e-05, "loss": 0.0, "step": 76478 }, { "epoch": 1.48, "learning_rate": 2.526865024681534e-05, "loss": 0.0033, "step": 76480 }, { "epoch": 1.48, "learning_rate": 2.5268003286558108e-05, "loss": 0.0, "step": 76482 }, { "epoch": 1.48, "learning_rate": 2.5267356326300877e-05, "loss": 0.0, "step": 76484 }, { "epoch": 1.48, "learning_rate": 2.5266709366043646e-05, "loss": 0.0, "step": 76486 }, { "epoch": 1.48, "learning_rate": 2.5266062405786412e-05, "loss": 0.0, "step": 76488 }, { "epoch": 1.48, "learning_rate": 2.5265415445529184e-05, "loss": 0.0079, "step": 76490 }, { "epoch": 1.48, "learning_rate": 2.526476848527195e-05, "loss": 0.0, "step": 76492 }, { "epoch": 1.48, "learning_rate": 2.5264121525014722e-05, "loss": 0.0, "step": 76494 }, { "epoch": 1.48, "learning_rate": 2.5263474564757488e-05, "loss": 0.0, "step": 76496 }, { "epoch": 1.48, "learning_rate": 2.5262827604500254e-05, "loss": 0.0054, "step": 76498 }, { "epoch": 1.48, "learning_rate": 2.5262180644243026e-05, "loss": 0.0, "step": 76500 }, { "epoch": 1.48, "learning_rate": 2.5261533683985795e-05, "loss": 0.0, "step": 76502 }, { "epoch": 1.48, "learning_rate": 2.526088672372856e-05, "loss": 0.0, "step": 76504 }, { "epoch": 1.48, "learning_rate": 2.5260239763471334e-05, "loss": 0.0, "step": 76506 }, { "epoch": 1.48, "learning_rate": 2.52595928032141e-05, "loss": 0.0, "step": 76508 }, { "epoch": 1.48, "learning_rate": 2.5258945842956865e-05, "loss": 0.0, "step": 76510 }, { "epoch": 1.49, "learning_rate": 2.5258298882699637e-05, "loss": 0.0, "step": 76512 }, { "epoch": 1.49, "learning_rate": 2.5257651922442403e-05, "loss": 0.0, "step": 76514 }, { "epoch": 1.49, "learning_rate": 2.5257004962185176e-05, "loss": 0.0001, "step": 76516 }, { "epoch": 1.49, "learning_rate": 2.5256358001927945e-05, "loss": 0.0, "step": 76518 }, { "epoch": 1.49, "learning_rate": 2.525571104167071e-05, "loss": 0.0, "step": 76520 }, { "epoch": 1.49, "learning_rate": 2.5255064081413483e-05, "loss": 0.0001, "step": 76522 }, { "epoch": 1.49, "learning_rate": 2.525441712115625e-05, "loss": 0.0, "step": 76524 }, { "epoch": 1.49, "learning_rate": 2.5253770160899014e-05, "loss": 0.0, "step": 76526 }, { "epoch": 1.49, "learning_rate": 2.5253123200641787e-05, "loss": 0.0, "step": 76528 }, { "epoch": 1.49, "learning_rate": 2.5252476240384552e-05, "loss": 0.0, "step": 76530 }, { "epoch": 1.49, "learning_rate": 2.525182928012732e-05, "loss": 0.0061, "step": 76532 }, { "epoch": 1.49, "learning_rate": 2.5251182319870094e-05, "loss": 0.0003, "step": 76534 }, { "epoch": 1.49, "learning_rate": 2.525053535961286e-05, "loss": 0.0001, "step": 76536 }, { "epoch": 1.49, "learning_rate": 2.5249888399355632e-05, "loss": 0.0, "step": 76538 }, { "epoch": 1.49, "learning_rate": 2.5249241439098398e-05, "loss": 0.0, "step": 76540 }, { "epoch": 1.49, "learning_rate": 2.5248594478841163e-05, "loss": 0.0, "step": 76542 }, { "epoch": 1.49, "learning_rate": 2.5247947518583936e-05, "loss": 0.0, "step": 76544 }, { "epoch": 1.49, "learning_rate": 2.52473005583267e-05, "loss": 0.0, "step": 76546 }, { "epoch": 1.49, "learning_rate": 2.524665359806947e-05, "loss": 0.0, "step": 76548 }, { "epoch": 1.49, "learning_rate": 2.5246006637812243e-05, "loss": 0.0, "step": 76550 }, { "epoch": 1.49, "learning_rate": 2.524535967755501e-05, "loss": 0.0001, "step": 76552 }, { "epoch": 1.49, "learning_rate": 2.5244712717297775e-05, "loss": 0.0038, "step": 76554 }, { "epoch": 1.49, "learning_rate": 2.5244065757040547e-05, "loss": 0.0, "step": 76556 }, { "epoch": 1.49, "learning_rate": 2.5243418796783313e-05, "loss": 0.0, "step": 76558 }, { "epoch": 1.49, "learning_rate": 2.5242771836526085e-05, "loss": 0.0, "step": 76560 }, { "epoch": 1.49, "learning_rate": 2.524212487626885e-05, "loss": 0.0221, "step": 76562 }, { "epoch": 1.49, "learning_rate": 2.524147791601162e-05, "loss": 0.0, "step": 76564 }, { "epoch": 1.49, "learning_rate": 2.5240830955754393e-05, "loss": 0.0, "step": 76566 }, { "epoch": 1.49, "learning_rate": 2.5240183995497158e-05, "loss": 0.0019, "step": 76568 }, { "epoch": 1.49, "learning_rate": 2.5239537035239924e-05, "loss": 0.0002, "step": 76570 }, { "epoch": 1.49, "learning_rate": 2.5238890074982696e-05, "loss": 0.0054, "step": 76572 }, { "epoch": 1.49, "learning_rate": 2.5238243114725462e-05, "loss": 0.0, "step": 76574 }, { "epoch": 1.49, "learning_rate": 2.5237596154468235e-05, "loss": 0.0, "step": 76576 }, { "epoch": 1.49, "learning_rate": 2.5236949194211e-05, "loss": 0.0, "step": 76578 }, { "epoch": 1.49, "learning_rate": 2.523630223395377e-05, "loss": 0.0023, "step": 76580 }, { "epoch": 1.49, "learning_rate": 2.5235655273696542e-05, "loss": 0.0005, "step": 76582 }, { "epoch": 1.49, "learning_rate": 2.5235008313439307e-05, "loss": 0.0, "step": 76584 }, { "epoch": 1.49, "learning_rate": 2.5234361353182073e-05, "loss": 0.0012, "step": 76586 }, { "epoch": 1.49, "learning_rate": 2.5233714392924846e-05, "loss": 0.0, "step": 76588 }, { "epoch": 1.49, "learning_rate": 2.523306743266761e-05, "loss": 0.0, "step": 76590 }, { "epoch": 1.49, "learning_rate": 2.5232420472410377e-05, "loss": 0.0, "step": 76592 }, { "epoch": 1.49, "learning_rate": 2.523177351215315e-05, "loss": 0.0, "step": 76594 }, { "epoch": 1.49, "learning_rate": 2.523112655189592e-05, "loss": 0.0, "step": 76596 }, { "epoch": 1.49, "learning_rate": 2.5230479591638688e-05, "loss": 0.0, "step": 76598 }, { "epoch": 1.49, "learning_rate": 2.5229832631381457e-05, "loss": 0.0, "step": 76600 }, { "epoch": 1.49, "learning_rate": 2.5229185671124222e-05, "loss": 0.0, "step": 76602 }, { "epoch": 1.49, "learning_rate": 2.5228538710866995e-05, "loss": 0.0, "step": 76604 }, { "epoch": 1.49, "learning_rate": 2.522789175060976e-05, "loss": 0.0161, "step": 76606 }, { "epoch": 1.49, "learning_rate": 2.5227244790352526e-05, "loss": 0.0, "step": 76608 }, { "epoch": 1.49, "learning_rate": 2.52265978300953e-05, "loss": 0.0, "step": 76610 }, { "epoch": 1.49, "learning_rate": 2.5225950869838068e-05, "loss": 0.0, "step": 76612 }, { "epoch": 1.49, "learning_rate": 2.5225303909580834e-05, "loss": 0.0, "step": 76614 }, { "epoch": 1.49, "learning_rate": 2.5224656949323606e-05, "loss": 0.0002, "step": 76616 }, { "epoch": 1.49, "learning_rate": 2.5224009989066372e-05, "loss": 0.0001, "step": 76618 }, { "epoch": 1.49, "learning_rate": 2.5223363028809144e-05, "loss": 0.0, "step": 76620 }, { "epoch": 1.49, "learning_rate": 2.522271606855191e-05, "loss": 0.0, "step": 76622 }, { "epoch": 1.49, "learning_rate": 2.5222069108294676e-05, "loss": 0.0014, "step": 76624 }, { "epoch": 1.49, "learning_rate": 2.5221422148037448e-05, "loss": 0.0, "step": 76626 }, { "epoch": 1.49, "learning_rate": 2.5220775187780217e-05, "loss": 0.0, "step": 76628 }, { "epoch": 1.49, "learning_rate": 2.5220128227522983e-05, "loss": 0.0, "step": 76630 }, { "epoch": 1.49, "learning_rate": 2.5219481267265755e-05, "loss": 0.0, "step": 76632 }, { "epoch": 1.49, "learning_rate": 2.521883430700852e-05, "loss": 0.0, "step": 76634 }, { "epoch": 1.49, "learning_rate": 2.5218187346751294e-05, "loss": 0.0, "step": 76636 }, { "epoch": 1.49, "learning_rate": 2.521754038649406e-05, "loss": 0.0, "step": 76638 }, { "epoch": 1.49, "learning_rate": 2.5216893426236825e-05, "loss": 0.0, "step": 76640 }, { "epoch": 1.49, "learning_rate": 2.5216246465979597e-05, "loss": 0.0, "step": 76642 }, { "epoch": 1.49, "learning_rate": 2.5215599505722366e-05, "loss": 0.0, "step": 76644 }, { "epoch": 1.49, "learning_rate": 2.5214952545465132e-05, "loss": 0.0, "step": 76646 }, { "epoch": 1.49, "learning_rate": 2.5214305585207905e-05, "loss": 0.0019, "step": 76648 }, { "epoch": 1.49, "learning_rate": 2.521365862495067e-05, "loss": 0.0105, "step": 76650 }, { "epoch": 1.49, "learning_rate": 2.5213011664693436e-05, "loss": 0.0, "step": 76652 }, { "epoch": 1.49, "learning_rate": 2.521236470443621e-05, "loss": 0.0, "step": 76654 }, { "epoch": 1.49, "learning_rate": 2.5211717744178974e-05, "loss": 0.0, "step": 76656 }, { "epoch": 1.49, "learning_rate": 2.5211070783921747e-05, "loss": 0.0197, "step": 76658 }, { "epoch": 1.49, "learning_rate": 2.5210423823664512e-05, "loss": 0.0024, "step": 76660 }, { "epoch": 1.49, "learning_rate": 2.520977686340728e-05, "loss": 0.0, "step": 76662 }, { "epoch": 1.49, "learning_rate": 2.5209129903150054e-05, "loss": 0.0032, "step": 76664 }, { "epoch": 1.49, "learning_rate": 2.520848294289282e-05, "loss": 0.003, "step": 76666 }, { "epoch": 1.49, "learning_rate": 2.5207835982635585e-05, "loss": 0.0, "step": 76668 }, { "epoch": 1.49, "learning_rate": 2.5207189022378358e-05, "loss": 0.0, "step": 76670 }, { "epoch": 1.49, "learning_rate": 2.5206542062121123e-05, "loss": 0.0, "step": 76672 }, { "epoch": 1.49, "learning_rate": 2.5205895101863893e-05, "loss": 0.0, "step": 76674 }, { "epoch": 1.49, "learning_rate": 2.520524814160666e-05, "loss": 0.0, "step": 76676 }, { "epoch": 1.49, "learning_rate": 2.520460118134943e-05, "loss": 0.0, "step": 76678 }, { "epoch": 1.49, "learning_rate": 2.5203954221092203e-05, "loss": 0.0015, "step": 76680 }, { "epoch": 1.49, "learning_rate": 2.520330726083497e-05, "loss": 0.0001, "step": 76682 }, { "epoch": 1.49, "learning_rate": 2.5202660300577735e-05, "loss": 0.0, "step": 76684 }, { "epoch": 1.49, "learning_rate": 2.5202013340320507e-05, "loss": 0.0, "step": 76686 }, { "epoch": 1.49, "learning_rate": 2.5201366380063273e-05, "loss": 0.0, "step": 76688 }, { "epoch": 1.49, "learning_rate": 2.5200719419806042e-05, "loss": 0.0, "step": 76690 }, { "epoch": 1.49, "learning_rate": 2.520007245954881e-05, "loss": 0.0, "step": 76692 }, { "epoch": 1.49, "learning_rate": 2.519942549929158e-05, "loss": 0.0, "step": 76694 }, { "epoch": 1.49, "learning_rate": 2.5198778539034346e-05, "loss": 0.0, "step": 76696 }, { "epoch": 1.49, "learning_rate": 2.5198131578777118e-05, "loss": 0.0001, "step": 76698 }, { "epoch": 1.49, "learning_rate": 2.5197484618519884e-05, "loss": 0.0001, "step": 76700 }, { "epoch": 1.49, "learning_rate": 2.5196837658262656e-05, "loss": 0.0, "step": 76702 }, { "epoch": 1.49, "learning_rate": 2.5196190698005422e-05, "loss": 0.0, "step": 76704 }, { "epoch": 1.49, "learning_rate": 2.519554373774819e-05, "loss": 0.0, "step": 76706 }, { "epoch": 1.49, "learning_rate": 2.519489677749096e-05, "loss": 0.0, "step": 76708 }, { "epoch": 1.49, "learning_rate": 2.519424981723373e-05, "loss": 0.0, "step": 76710 }, { "epoch": 1.49, "learning_rate": 2.5193602856976495e-05, "loss": 0.0, "step": 76712 }, { "epoch": 1.49, "learning_rate": 2.5192955896719267e-05, "loss": 0.0, "step": 76714 }, { "epoch": 1.49, "learning_rate": 2.5192308936462033e-05, "loss": 0.0014, "step": 76716 }, { "epoch": 1.49, "learning_rate": 2.5191661976204806e-05, "loss": 0.0, "step": 76718 }, { "epoch": 1.49, "learning_rate": 2.519101501594757e-05, "loss": 0.0, "step": 76720 }, { "epoch": 1.49, "learning_rate": 2.519036805569034e-05, "loss": 0.0, "step": 76722 }, { "epoch": 1.49, "learning_rate": 2.518972109543311e-05, "loss": 0.0011, "step": 76724 }, { "epoch": 1.49, "learning_rate": 2.518907413517588e-05, "loss": 0.0111, "step": 76726 }, { "epoch": 1.49, "learning_rate": 2.5188427174918644e-05, "loss": 0.0, "step": 76728 }, { "epoch": 1.49, "learning_rate": 2.5187780214661417e-05, "loss": 0.0072, "step": 76730 }, { "epoch": 1.49, "learning_rate": 2.5187133254404182e-05, "loss": 0.0, "step": 76732 }, { "epoch": 1.49, "learning_rate": 2.5186486294146948e-05, "loss": 0.0, "step": 76734 }, { "epoch": 1.49, "learning_rate": 2.518583933388972e-05, "loss": 0.0, "step": 76736 }, { "epoch": 1.49, "learning_rate": 2.5185192373632486e-05, "loss": 0.0, "step": 76738 }, { "epoch": 1.49, "learning_rate": 2.518454541337526e-05, "loss": 0.0, "step": 76740 }, { "epoch": 1.49, "learning_rate": 2.5183898453118028e-05, "loss": 0.0, "step": 76742 }, { "epoch": 1.49, "learning_rate": 2.5183251492860794e-05, "loss": 0.0057, "step": 76744 }, { "epoch": 1.49, "learning_rate": 2.5182604532603566e-05, "loss": 0.0001, "step": 76746 }, { "epoch": 1.49, "learning_rate": 2.5181957572346332e-05, "loss": 0.0, "step": 76748 }, { "epoch": 1.49, "learning_rate": 2.5181310612089097e-05, "loss": 0.0, "step": 76750 }, { "epoch": 1.49, "learning_rate": 2.518066365183187e-05, "loss": 0.0, "step": 76752 }, { "epoch": 1.49, "learning_rate": 2.5180016691574636e-05, "loss": 0.0, "step": 76754 }, { "epoch": 1.49, "learning_rate": 2.5179369731317405e-05, "loss": 0.0, "step": 76756 }, { "epoch": 1.49, "learning_rate": 2.5178722771060177e-05, "loss": 0.0, "step": 76758 }, { "epoch": 1.49, "learning_rate": 2.5178075810802943e-05, "loss": 0.0, "step": 76760 }, { "epoch": 1.49, "learning_rate": 2.5177428850545715e-05, "loss": 0.0, "step": 76762 }, { "epoch": 1.49, "learning_rate": 2.517678189028848e-05, "loss": 0.0, "step": 76764 }, { "epoch": 1.49, "learning_rate": 2.5176134930031247e-05, "loss": 0.0, "step": 76766 }, { "epoch": 1.49, "learning_rate": 2.517548796977402e-05, "loss": 0.0, "step": 76768 }, { "epoch": 1.49, "learning_rate": 2.5174841009516785e-05, "loss": 0.0, "step": 76770 }, { "epoch": 1.49, "learning_rate": 2.5174194049259554e-05, "loss": 0.0, "step": 76772 }, { "epoch": 1.49, "learning_rate": 2.5173547089002326e-05, "loss": 0.0, "step": 76774 }, { "epoch": 1.49, "learning_rate": 2.5172900128745092e-05, "loss": 0.0, "step": 76776 }, { "epoch": 1.49, "learning_rate": 2.5172253168487865e-05, "loss": 0.0, "step": 76778 }, { "epoch": 1.49, "learning_rate": 2.517160620823063e-05, "loss": 0.0, "step": 76780 }, { "epoch": 1.49, "learning_rate": 2.5170959247973396e-05, "loss": 0.0001, "step": 76782 }, { "epoch": 1.49, "learning_rate": 2.517031228771617e-05, "loss": 0.0004, "step": 76784 }, { "epoch": 1.49, "learning_rate": 2.5169665327458934e-05, "loss": 0.0, "step": 76786 }, { "epoch": 1.49, "learning_rate": 2.5169018367201703e-05, "loss": 0.0033, "step": 76788 }, { "epoch": 1.49, "learning_rate": 2.5168371406944476e-05, "loss": 0.0, "step": 76790 }, { "epoch": 1.49, "learning_rate": 2.516772444668724e-05, "loss": 0.0, "step": 76792 }, { "epoch": 1.49, "learning_rate": 2.5167077486430007e-05, "loss": 0.0009, "step": 76794 }, { "epoch": 1.49, "learning_rate": 2.516643052617278e-05, "loss": 0.0, "step": 76796 }, { "epoch": 1.49, "learning_rate": 2.5165783565915545e-05, "loss": 0.0, "step": 76798 }, { "epoch": 1.49, "learning_rate": 2.5165136605658318e-05, "loss": 0.0, "step": 76800 }, { "epoch": 1.49, "learning_rate": 2.5164489645401083e-05, "loss": 0.0, "step": 76802 }, { "epoch": 1.49, "learning_rate": 2.5163842685143853e-05, "loss": 0.0001, "step": 76804 }, { "epoch": 1.49, "learning_rate": 2.5163195724886625e-05, "loss": 0.0, "step": 76806 }, { "epoch": 1.49, "learning_rate": 2.516254876462939e-05, "loss": 0.0, "step": 76808 }, { "epoch": 1.49, "learning_rate": 2.5161901804372156e-05, "loss": 0.0, "step": 76810 }, { "epoch": 1.49, "learning_rate": 2.516125484411493e-05, "loss": 0.0, "step": 76812 }, { "epoch": 1.49, "learning_rate": 2.5160607883857695e-05, "loss": 0.0, "step": 76814 }, { "epoch": 1.49, "learning_rate": 2.515996092360046e-05, "loss": 0.0, "step": 76816 }, { "epoch": 1.49, "learning_rate": 2.5159313963343233e-05, "loss": 0.0, "step": 76818 }, { "epoch": 1.49, "learning_rate": 2.5158667003086002e-05, "loss": 0.0, "step": 76820 }, { "epoch": 1.49, "learning_rate": 2.5158020042828774e-05, "loss": 0.0, "step": 76822 }, { "epoch": 1.49, "learning_rate": 2.515737308257154e-05, "loss": 0.0012, "step": 76824 }, { "epoch": 1.49, "learning_rate": 2.5156726122314306e-05, "loss": 0.0, "step": 76826 }, { "epoch": 1.49, "learning_rate": 2.5156079162057078e-05, "loss": 0.0033, "step": 76828 }, { "epoch": 1.49, "learning_rate": 2.5155432201799844e-05, "loss": 0.0001, "step": 76830 }, { "epoch": 1.49, "learning_rate": 2.515478524154261e-05, "loss": 0.0001, "step": 76832 }, { "epoch": 1.49, "learning_rate": 2.5154138281285382e-05, "loss": 0.0124, "step": 76834 }, { "epoch": 1.49, "learning_rate": 2.515349132102815e-05, "loss": 0.0, "step": 76836 }, { "epoch": 1.49, "learning_rate": 2.5152844360770917e-05, "loss": 0.0, "step": 76838 }, { "epoch": 1.49, "learning_rate": 2.515219740051369e-05, "loss": 0.0, "step": 76840 }, { "epoch": 1.49, "learning_rate": 2.5151550440256455e-05, "loss": 0.0, "step": 76842 }, { "epoch": 1.49, "learning_rate": 2.5150903479999227e-05, "loss": 0.0, "step": 76844 }, { "epoch": 1.49, "learning_rate": 2.5150256519741993e-05, "loss": 0.0001, "step": 76846 }, { "epoch": 1.49, "learning_rate": 2.514960955948476e-05, "loss": 0.0, "step": 76848 }, { "epoch": 1.49, "learning_rate": 2.514896259922753e-05, "loss": 0.0, "step": 76850 }, { "epoch": 1.49, "learning_rate": 2.51483156389703e-05, "loss": 0.0, "step": 76852 }, { "epoch": 1.49, "learning_rate": 2.5147668678713066e-05, "loss": 0.0, "step": 76854 }, { "epoch": 1.49, "learning_rate": 2.514702171845584e-05, "loss": 0.0092, "step": 76856 }, { "epoch": 1.49, "learning_rate": 2.5146374758198604e-05, "loss": 0.0001, "step": 76858 }, { "epoch": 1.49, "learning_rate": 2.5145727797941377e-05, "loss": 0.0021, "step": 76860 }, { "epoch": 1.49, "learning_rate": 2.5145080837684142e-05, "loss": 0.001, "step": 76862 }, { "epoch": 1.49, "learning_rate": 2.5144433877426908e-05, "loss": 0.0, "step": 76864 }, { "epoch": 1.49, "learning_rate": 2.514378691716968e-05, "loss": 0.0, "step": 76866 }, { "epoch": 1.49, "learning_rate": 2.514313995691245e-05, "loss": 0.0, "step": 76868 }, { "epoch": 1.49, "learning_rate": 2.5142492996655215e-05, "loss": 0.0006, "step": 76870 }, { "epoch": 1.49, "learning_rate": 2.5141846036397988e-05, "loss": 0.0, "step": 76872 }, { "epoch": 1.49, "learning_rate": 2.5141199076140754e-05, "loss": 0.0, "step": 76874 }, { "epoch": 1.49, "learning_rate": 2.514055211588352e-05, "loss": 0.0, "step": 76876 }, { "epoch": 1.49, "learning_rate": 2.5139905155626292e-05, "loss": 0.0112, "step": 76878 }, { "epoch": 1.49, "learning_rate": 2.5139258195369057e-05, "loss": 0.0002, "step": 76880 }, { "epoch": 1.49, "learning_rate": 2.513861123511183e-05, "loss": 0.0, "step": 76882 }, { "epoch": 1.49, "learning_rate": 2.51379642748546e-05, "loss": 0.0, "step": 76884 }, { "epoch": 1.49, "learning_rate": 2.5137317314597365e-05, "loss": 0.0, "step": 76886 }, { "epoch": 1.49, "learning_rate": 2.5136670354340137e-05, "loss": 0.0, "step": 76888 }, { "epoch": 1.49, "learning_rate": 2.5136023394082903e-05, "loss": 0.0, "step": 76890 }, { "epoch": 1.49, "learning_rate": 2.513537643382567e-05, "loss": 0.0, "step": 76892 }, { "epoch": 1.49, "learning_rate": 2.513472947356844e-05, "loss": 0.0, "step": 76894 }, { "epoch": 1.49, "learning_rate": 2.5134082513311207e-05, "loss": 0.0, "step": 76896 }, { "epoch": 1.49, "learning_rate": 2.5133435553053976e-05, "loss": 0.0001, "step": 76898 }, { "epoch": 1.49, "learning_rate": 2.5132788592796745e-05, "loss": 0.0, "step": 76900 }, { "epoch": 1.49, "learning_rate": 2.5132141632539514e-05, "loss": 0.0, "step": 76902 }, { "epoch": 1.49, "learning_rate": 2.5131494672282286e-05, "loss": 0.0, "step": 76904 }, { "epoch": 1.49, "learning_rate": 2.5130847712025052e-05, "loss": 0.0, "step": 76906 }, { "epoch": 1.49, "learning_rate": 2.5130200751767818e-05, "loss": 0.0002, "step": 76908 }, { "epoch": 1.49, "learning_rate": 2.512955379151059e-05, "loss": 0.0002, "step": 76910 }, { "epoch": 1.49, "learning_rate": 2.5128906831253356e-05, "loss": 0.0, "step": 76912 }, { "epoch": 1.49, "learning_rate": 2.5128259870996125e-05, "loss": 0.0, "step": 76914 }, { "epoch": 1.49, "learning_rate": 2.5127612910738894e-05, "loss": 0.0, "step": 76916 }, { "epoch": 1.49, "learning_rate": 2.5126965950481663e-05, "loss": 0.0, "step": 76918 }, { "epoch": 1.49, "learning_rate": 2.5126318990224436e-05, "loss": 0.0, "step": 76920 }, { "epoch": 1.49, "learning_rate": 2.51256720299672e-05, "loss": 0.0, "step": 76922 }, { "epoch": 1.49, "learning_rate": 2.5125025069709967e-05, "loss": 0.0, "step": 76924 }, { "epoch": 1.49, "learning_rate": 2.512437810945274e-05, "loss": 0.0, "step": 76926 }, { "epoch": 1.49, "learning_rate": 2.5123731149195505e-05, "loss": 0.0, "step": 76928 }, { "epoch": 1.49, "learning_rate": 2.5123084188938274e-05, "loss": 0.0045, "step": 76930 }, { "epoch": 1.49, "learning_rate": 2.5122437228681043e-05, "loss": 0.0035, "step": 76932 }, { "epoch": 1.49, "learning_rate": 2.5121790268423813e-05, "loss": 0.0, "step": 76934 }, { "epoch": 1.49, "learning_rate": 2.5121143308166578e-05, "loss": 0.0, "step": 76936 }, { "epoch": 1.49, "learning_rate": 2.512049634790935e-05, "loss": 0.0063, "step": 76938 }, { "epoch": 1.49, "learning_rate": 2.5119849387652116e-05, "loss": 0.0067, "step": 76940 }, { "epoch": 1.49, "learning_rate": 2.511920242739489e-05, "loss": 0.0, "step": 76942 }, { "epoch": 1.49, "learning_rate": 2.5118555467137655e-05, "loss": 0.0, "step": 76944 }, { "epoch": 1.49, "learning_rate": 2.5117908506880424e-05, "loss": 0.0001, "step": 76946 }, { "epoch": 1.49, "learning_rate": 2.5117261546623193e-05, "loss": 0.0, "step": 76948 }, { "epoch": 1.49, "learning_rate": 2.5116614586365962e-05, "loss": 0.0, "step": 76950 }, { "epoch": 1.49, "learning_rate": 2.5115967626108728e-05, "loss": 0.0, "step": 76952 }, { "epoch": 1.49, "learning_rate": 2.51153206658515e-05, "loss": 0.0016, "step": 76954 }, { "epoch": 1.49, "learning_rate": 2.5114673705594266e-05, "loss": 0.0, "step": 76956 }, { "epoch": 1.49, "learning_rate": 2.511402674533703e-05, "loss": 0.0, "step": 76958 }, { "epoch": 1.49, "learning_rate": 2.5113379785079804e-05, "loss": 0.0002, "step": 76960 }, { "epoch": 1.49, "learning_rate": 2.5112732824822573e-05, "loss": 0.0, "step": 76962 }, { "epoch": 1.49, "learning_rate": 2.5112085864565342e-05, "loss": 0.0, "step": 76964 }, { "epoch": 1.49, "learning_rate": 2.511143890430811e-05, "loss": 0.0, "step": 76966 }, { "epoch": 1.49, "learning_rate": 2.5110791944050877e-05, "loss": 0.0, "step": 76968 }, { "epoch": 1.49, "learning_rate": 2.511014498379365e-05, "loss": 0.0, "step": 76970 }, { "epoch": 1.49, "learning_rate": 2.5109498023536415e-05, "loss": 0.0, "step": 76972 }, { "epoch": 1.49, "learning_rate": 2.510885106327918e-05, "loss": 0.0, "step": 76974 }, { "epoch": 1.49, "learning_rate": 2.5108204103021953e-05, "loss": 0.0, "step": 76976 }, { "epoch": 1.49, "learning_rate": 2.510755714276472e-05, "loss": 0.0, "step": 76978 }, { "epoch": 1.49, "learning_rate": 2.5106910182507488e-05, "loss": 0.0, "step": 76980 }, { "epoch": 1.49, "learning_rate": 2.510626322225026e-05, "loss": 0.0, "step": 76982 }, { "epoch": 1.49, "learning_rate": 2.5105616261993026e-05, "loss": 0.0001, "step": 76984 }, { "epoch": 1.49, "learning_rate": 2.51049693017358e-05, "loss": 0.0001, "step": 76986 }, { "epoch": 1.49, "learning_rate": 2.5104322341478564e-05, "loss": 0.0, "step": 76988 }, { "epoch": 1.49, "learning_rate": 2.510367538122133e-05, "loss": 0.0, "step": 76990 }, { "epoch": 1.49, "learning_rate": 2.5103028420964102e-05, "loss": 0.0, "step": 76992 }, { "epoch": 1.49, "learning_rate": 2.5102381460706868e-05, "loss": 0.0, "step": 76994 }, { "epoch": 1.49, "learning_rate": 2.5101734500449637e-05, "loss": 0.0, "step": 76996 }, { "epoch": 1.49, "learning_rate": 2.510108754019241e-05, "loss": 0.0005, "step": 76998 }, { "epoch": 1.49, "learning_rate": 2.5100440579935175e-05, "loss": 0.0, "step": 77000 } ], "max_steps": 154569, "num_train_epochs": 3, "total_flos": 5.674080833304576e+16, "trial_name": null, "trial_params": null }